For years I have been writing about how HIV/AIDS was deliberately seeded into
the American gay community via the government-sponsored contaminated
hepatitis B experiments (1978-1981) ---- and how this was covered-up by the
AIDS establishment.

Now we have statistical proof to show that HIV was indeed planted in the gay
community in those "pre-AIDS epidemic" years.

Please read Tom Keske's report in this file attachment.   Much of it is
statistical data which proves his case. However, there is also much in his
report that is very easily read --- and which provides you with
historic/scientific facts that show, without doubt, that HIV/AIDS is
connected to government experiments using gay men as guinea pigs.

Please pass on to other interested parties who might be interested in the
origin of HIV/AIDS as it explains why AIDS was originally a "gay disease" in
the United States. This report is also if interest to groups interested in
vaccine dangers and contamination problems connected with vaccines.

Author Tom Keske can be reached at :

Personally, I believe Tom has done an amazing job with his research.
For those of us who have been saying for many years that AIDS is a man-made
disease, Tom Keske's hepatitis B vaccine report is required reading.

Alan Cantwell Jr MD


Statistical Analysis Linking U.S. AIDS  Outbreak to Hepatitis Experiments---Thomas R. Keske
           Version 4.2                                 
           Author:        Thomas R. Keske

TABLE OF CONTENTS  ------------------------------------------------- Page 2

   1 Abstract                                                           4
   2 Introduction                                                       5
   3 HIV Correlation to Vaccine Trial                                   7
     3.1 Methodology Overview                                           7
     3.2 Lemp 1990 Study                                                8
     3.3 Estimation of High Risk Population                             9
     3.4 Vaccine Correlation Analysis                                  12
     3.5 Lemp Study, Odds Per Year                                     14

   4 HIV Correlation to Non-vaccine Participation                      15
     4.1 Over-representation in earliest AIDS cases                    15
     4.2 High Rate of Seroconversion During Recruitment Years          20
     4.3 Unreasonable Study Size/Selection                             22
   5 Epidemiological Anomalies                                         24
     5.1 Unreasonable Delay of HIV in IV Drug Community                24
     5.2 Anomalies Revealed by Computer Modeling                       24
   6 Unreasonable Approval of the Vaccine                              26
   7 Historical Context                                                27
   8 Conclusions                                                       30
   9 Refuting Counter-Arguments                                        33
   10 About the Author                                                 37
   11 References                                                       38
   12 Acknowledgments                                                  41
   13 Document Reproduction                                            41
   Appendix A Demonstrating the Validity of the Statistical Approach  42
   Appendix B  Letter From Dr. George Lemp                             44
   Appendix C Error Analysis for Lemp Data Calculations                46
     C.1 Effect of Variation in High Risk Population Estimate          46
     C.2 Effect of Errors in HIV Infection Rate Figures                47
   Appendix D  Letter from Case Western Reserve Statistics Department 48

Table of Contents  ------------------------------------------------- Page 3

   Appendix E Software Epidemic Modeling Analysis                      50
     E.1 Per-Contact Infection Rates                                   51
     E.2 Evidence of Program Accuracy                                  52
       E.2.1 Consistency with Independent Mathematical Test            52
       E.2.2 Consistency with Real-Life Experimental Results           52
     E.3 First Year, SFHBVCS                                           54
     E.4 Patient Zero Scenario                                         56
     E.5 Estimated Seed Size in SF                                     57
     E.6 From Where Comes the Seed?                                    58
     E.7 Variable Infectivity Per Stage                                59
   Appendix F General Statistical Primer                               61

Section 1 Abstract   ------------------------------------------------ Page 4
   1  Abstract
   This statistical study concerns what is probably one of the most
   significant and overlooked issues of our time.  It  demonstrates proof of
   a strong link between the U.S. outbreak of AIDS, and hepatitis studies
   that were performed on gay males, starting in the late 1970s.     The
   analysis refutes explanations that attribute the connection simply to
   sexual risk behavior on the part of the study participants.

   The analysis also presents evidence suggesting that HIV infections
   occurring in the studies were more likely to have been intentional rather
   than accidental.   This raises the question of whether the men in these
   studies might have been used as guinea pigs for covert experimentation, or
   whether a sexually-transmitted epidemic might have been deliberately
   induced, as a means to rid society of "undesirables".    Regardless of
   whether the virus itself came into existence naturally, its initial spread
   was clearly unnatural.

   The methodology used in this document is highly similar to that which is
   typically used to evaluate the effectiveness and safety of vaccines.   The
   analysis evaluates differences in infection rates between a suitable
   control group, versus a vaccine test group.

   In the first two years of the epidemic in San Francisco, between 50 and 60
   percent of the earliest known AIDS cases were from persons involved in the
   hepatitis studies.  A goal of this analysis is to calculate specific
   probabilities for these and other similar figures.  It demonstrates that
   such figures cannot credibly be attributed merely to chance, or to
   differences in risk behaviors.

   Odds of the disproportionate levels of HIV infection among men in the
   vaccine trial, relative to other men of similar risk behaviors, are shown
   to be as little as 1 in a trillion.

   A statistical link exists not only to experimental vaccines, but also
   simply to the fact of participation in the hepatitis studies, such as
   simply to have blood drawn for purposes of monitoring  hepatitis
   prevalence.   Few logical or benign possibilities exist to explain why
   there should be such a connection, yet it exists.  Odds against the higher
   initial rate of AIDS among study participants was as little as 1 in 300,
   000, when compared to men of equal or higher risk.

   Various epidemiological anomalies also suggest that an artificial,
   simultaneous, mass-infection would have been necessary in order to produce
   the type of explosion in HIV that was observed in the early 1980s.    Full-
   blown AIDS should have been evident many years earlier, before HIV  was
   nearly so widespread   Thousands of infections would have been necessary
   to fuel the levels of HIV growth that were observed, during years in
   which no retroactive evidence of HIV exists.

   These anomalies are analyzed using computer modeling software.

Section 2 Introduction --------------------------------------------- Page 5

   2  Introduction

   Recently, there has been renewed interest as to whether massive polio and
   smallpox vaccine programs in Africa may have initiated, or at least
   accelerated, the spread of AIDS on that continent [1].  The possibility
   that a similar vaccine phenomenon may have occurred in the United States
   should heighten concern.

   For two decades, there has been some concern about a possible connection
   between the outbreak of AIDS in America and a government-sponsored
   experimental hepatitis B vaccine which was injected into gay men in New
   York City, San Francisco, Los Angeles, Chicago, St. Louis, and Denver,
   between the years 1978-1982.

   Several reports in the medical literature attest to the safety of the
   experimental vaccine.  Thus, the connection between the gay experiments
   and AIDS has been largely dismissed as unworthy of investigation. There is
   a possible element of bias in this hasty dismissal.   The question of
   possible vaccine contamination is too important to allow political
   concerns or emotional reactions to interfere with an objective
   investigation and discussion.

   The purpose of this document is to reinvestigate the connection of the
   original AIDS outbreak in San Francisco gay men not only with the
   experimental hepatitis vaccine, but also simply with the act of
   participation in the government-sponsored hepatitis study.  This is a new
   analysis of data collected from various published studies.

   This document will demonstrate how gay men who volunteered for government
   hepatitis experiments were far more likely to become infected with HIV
   than those who did not take part in such experiments, to a degree not
   credibly explainable by chance or by life-style.

   This document does not speculate whether HIV is an old or a new virus, nor
   does it explore the outbreak in Africa.  It does not focus on whether HIV
   is a natural virus or a genetically engineered virus that could have
   resulted from laboratory experimentation.

   None of these scenarios preclude the possibility of HIV contamination of
   the hepatitis vaccines.

   The analysis will attempt to show that there is a significant statistical
   correlation between vaccine volunteers and HIV infection, that is not
   merely the result of "high-risk" sexual behaviors.

   This document is being distributed to AIDS activists, virologists,
   biostaticians, journalists and others, in an effort to promote further
   research and dialog into the proposed connection of AIDS to the government
   hepatitis studies.

Section 2 Introduction --------------------------------------------- Page 6

   This document is intended for a broad target audience, including persons
   of varying backgrounds and levels of knowledge about statistics.  For
   those who have more questions about the statistical computation, a primer
   and a more extended discussion is provided in Appendix F.  Persons who
   have background in statistics can simply skip this section.

   As will be explained, there is good reason to believe that HIV was in the
   vaccines.  There is also reason to believe that the presence of HIV was
   not likely to have been accidental.

   Furthermore, there is an even more peculiar connection of HIV infection
   simply with participation in the government studies, even for gays who
   received no experimental vaccine.  This connection is not credibly
   explained simply by the "risk" status of the men involved.    This
   connection is even more disturbing, because there are no vaccines involved
   for which a possible accident could have occurred in the production.

   During World War II, a U.S. State Department official once dismissed
   allegations about the Holocaust as being of too "fantastic" of a nature to
   be worthy of forwarding.  It may be fashionable in current times to make
   caricatures of every allegation concerning cover-ups, conspiracies, or
   secret experiments.   However, this fashion does not represent wisdom
   today, any more than it did in World War II.

   There are already more than 33 million people estimated to be infected
   with HIV/AIDS.

   It is perhaps the single most significant incident that has occurred in
   human history.  The number of lives claimed can be expected to exceed the
   6 million killed in the Holocaust, or even to exceed the total global
   battle deaths of World War II.   If there is any chance of human agency
   involved in the genesis of the AIDS epidemic, it is perhaps the most
   important question of our time.

   If there is even the slightest chance of negligence or malfeasance, it
   would deserve to be investigated.  This document will put a quantitative
   number on that chance, and show that it is more than slight.

   The purpose of raising this question is not simply to cast accusation or
   blame.  If a vaccine accident occurred, it is important to determine why,
   so that such accidents do not happen again.  If there is any chance that
   vaccines could have been contaminated through carelessness or intention,
   then there must be accountability.   If there is a plausible chance that
   there were further factors causing HIV infection, in addition to the use
   of vaccines, then those factors must also be identified.

Section 3: HIV Correlation to Vaccine Trial ------------------------ Page 7

   3  HIV Correlation to Vaccine Trial

   3.1  Methodology Overview

   The primary focus for the statistical analysis of vaccine involvement is
   based on two distinct groups (cohorts) of San Francisco gay men.  One
   group received the experimental hepatitis B vaccine; the other did not.

   In 1978, a research group from the San Francisco Department of Public
   Health began epidemiological studies of gay and bisexual men attending the
   City Clinic, a public health clinic for treatment of sexually transmitted

   The San Francisco City Clinic Cohort Study (SFCCC) involved over 6700 gay
   men.   Most of these men participated by donating blood samples for
   hepatitis study, but did not receive experimental vaccines.

   A smaller cohort of 359 homosexual and bisexual men, selected from the
   larger group of 6700+ men in the San Francisco City Clinic, participated
   in a clinical trial of a vaccine to prevent Hepatitis B [2].  In this
   document, this group is known as the San Francisco Hepatitis B Vaccine
   Cohort Study (SFHBVCS).

   The second group referenced in this document for purposes of
   retrospectively analyzing HIV incidence is the San Francisco Men's Health
   Study (SFMHS), a cohort of 799 homosexual and bisexual men sampled from 19
   high-risk census tracts in San Francisco [2].  The SFMHS began in June
   1984.   It also included 204 HIV-negative heterosexual men, who are not
   relevant to this analysis.

   Different studies sometimes refer to slightly differing numbers of men in
   the SFCCC and SFMHS, depending on the date of the study.  For example,
   some cite 6875 for SFCCC and 809 for SFMHS.  This document will use values
   as cited in the context of specific, referenced studies, or will otherwise
   prefer the higher values.

   The SFHBVCS group is composed of 359 San Francisco gay men who received
   the experimental vaccine.  The SFMHS group consists of the 799 high-risk
   San Francisco gay/bisexual men who did not receive the vaccine.  Both
   groups were at equal risk of acquiring HIV.  Further details on this point
   are addressed in section 9.

   If "equal risk" can be adequately demonstrated, then any differences in
   the HIV rates between the two groups should be attributable merely to
   random chance.   The analysis will demonstrate that the hypothesis of
   "random chance" can be rejected.

Section 3: HIV Correlation to Vaccine Trial ------------------------ Page 8

   The statistical comparison of the two groups is a very commonplace type of
   problem, that is easily computed.

   3.2  Lemp 1990 Study

   One of the sources of data for this analysis is a study headed by Dr.
   George Lemp [3]: "Projections of AIDS morbidity and mortality in San

   Dr. Lemp, formerly with the AIDS Office of The City's Department of Public
   Health, now serves as director of the University-wide AIDS Research
   Program at the University of California (starting 1997) [4]

   The purpose of Lemp's study was not to evaluate the hepatitis vaccine or
   to compare the men in the trial with men who did not receive the
   experimental vaccine.  Its purpose was to develop a model for predicting
   the growth over time of the AIDS epidemic.  These projections required
   tracking of HIV seroconversion in high-risk men.  Stored blood samples,
   taken from the men in the San Francisco Clinic studies, were useful for
   this purpose.  Testing of the blood samples determined exactly when the
   men showed their first indications of exposure to HIV.

   For these purposes of his own, it happened that Lemp collected data which
   compares the SFHBVCS vaccine group with the SFMHS non-vaccine group. This
   same data is also useful for the further analysis in this document.

   This document does not claim any endorsement from Dr. Lemp.  It merely
   uses the Lemp data for a different purpose.   I contacted Dr. Lemp to
   verify that his data was correct as quoted (see Appendix B).

   The data in question, which I verified with Dr. Lemp, was derived from
   charts contained in the study.   It was previously posted to
   by Billie Goldberg, a San Francisco lay scientist and AIDS researcher:

   SFHBVCS: 1978 - 0.3%, 1979 - 4%, 1980 - 15%, 1981- 28%, 1982 - 40%,
   1983 - 46%, 1984 - 47%, 1985 - 48%, 1986 - 48%, 1987 - 49.3%

   SFMHS: 1978 - 0%, 1979 - 2%, 1980 - 4%, 1981- 10%, 1982 - 23%,
   1983 - 42%, 1984 - 48%, 1985 - 49%, 1986 - 49.3%, 1987 - 49.3%

   The above two lines show the total percent of HIV infection, in each year,
   for the SFHBVCS (vaccine) group, and the SFMHS (high-risk gay men who did
   not receive vaccine).

Section 3: HIV Correlation to Vaccine Trial ------------------------ Page 9

   The SFHBVCS group had blood samples taken in each year shown, used to
   estimate HIV prevalence.  The SFMHS had an estimate derived from a
   subgroup in 1982, and complete samples from 1984, on.  The Lemp study used
   curve-fitting to fill in values for years that did not have actual samples.
   The analysis in this document will focus primarily on the year 1982, since
   that is the first figure based on actual measurement, and is therefore
   likely to be the most accurate.

   The rate of HIV infection in the SFMHS group will be taken as a measure of
   the "expected" rate of HIV for the high-risk population of San Francisco,
   for this analysis.

   The justification for this is based on the comparative patterns of HIV
   growth in the SFMHS and the SFHBVCS, as shown in the Lemp data.

   Both groups start out nearly the same.   When each group hit a level of 45+
   percent infection, it abruptly ceased the high rate of growth.

   Even though the SFHBVCS vaccine group had a slight "head start" in
   infection, the SFMHS group "caught up" and actually hit the wall of
   saturation at 49.3% infection, in 1986, a year ahead of SFCSS vaccine
   group.  This suggests that on average, the SFMHS, non-vaccine group may
   have been even more promiscuous and high-risk than their vaccine

   What most distinguishes the two groups is the large, initial level of HIV
   seroconversion in the vaccine group, shortly after they received the
   vaccine.  Based on the fact that SFMHS appears to be at least at equal
   risk for HIV infection, the two groups should have been roughly equal in
   HIV prevalence during the earliest years, as well.  What the statistical
   analysis examines is the likelihood for this initial deviation.

   3.3  Estimation of High Risk Population

   In order to compute probabilities for the vaccine group to have exhibited
   their high rate

   of HIV infection by random chance, it is necessary to have an estimate of
   the total high-risk, gay male population in San Francisco, in 1978.  It is
   important to note that this is an estimate of only the high-risk 
   population, not the total gay male population.  The intent is to eliminate
   the high-risk status of the vaccine group as the postulated reason for
   their exhibiting a higher rate of HIV infection.

   As it happens, the conclusion of the statistical analysis is not highly
   dependent on having more than a rough count of the "high risk" population.  
   This question is analyzed further in section C.1.

Section 3: HIV Correlation to Vaccine Trial ------------------------ Page 10

   The estimate used is based on statistics for the city of San Francisco,
   for total HIV infections that occurred between 1978 and 1999.   It is
   reasonable to imagine that the numbers of gay men who did in fact become
   HIV+ in San Francisco, in the two decades to follow, would roughly reflect
   the numbers who were at risk in 1978 (not accounting for immigration and

   Following  is data from the city of San Francisco [5].

        Reported AIDS cases since 1981:  26,398
        AIDS deaths to date since 1981:  18,066
        Persons currently living with AIDS: 8,332

        Estimated HIV Infected to date (since 1981): 15,250
          (approx. 1 in every 50 San Franciscans (2.1%)


           SAN FRANCISCO: 79% MSM, 11% MSM+drug,  7% iv-drug, 1% heterosexual
           CALIF:          71% MSM,  9% MSM+drug, 10% iv-drug, 4% heterosexual
           US:             49% MSM,  6% MSM+drug, 25% iv-drug, 9% heterosexual

                     ("MSM" = Men having sex with Men")

   The total HIV infections in San Francisco is the total number of
   cumulative AIDS cases, (26398) plus the total number of HIV infections
   that are not yet progressed to AIDS (15250).

   In San Francisco, 79% of the AIDS/HIV cases are gay men (MSM).

   An additional 11% is combined "MSM plus IV drug" category.    For the sake
   of this estimate, half of the 11% will be counted as attributable to MSM
   (this is a small factor, anyway).

   The total percent attributed to MSM is then  .79 + (.5 * .11) = .84

   The total estimate of high-risk gay men is then
   (26398+15250) * (.84), or roughly 35000.

Section 3: HIV Correlation to Vaccine Trial ------------------------ Page 11

   A factor that could possibly tend to make this estimate too high is that
   the estimate does not account for possible immigration into the city.   
   Also, people who acquired HIV over a longer period of time may not have
   been as high-risk as the people who acquired AIDS in the early years.

   However, there are also significant  factors that could tend to make the
   estimate too low.  It does not account for men who are HIV+, but simply
   have not been tested and counted.  It does not account for men who adopted
   "safe sex" practices after the AIDS epidemic became publicized.   This
   would tend to significantly reduce the risk level of men who were
   previously "high risk."

   Thus, some factors would tend to make the estimate err on the side of
   being too high, while other factors would make it err on the side of being
   too low.   This mix of factors has some tendency to cancel each other out,
   in overall effect.

   The estimate of 35000 high-risk gay men was made independently of the Lemp
   study.  Interestingly, the Lemp study needed to make an estimate of the
   entire gay male population of San Francisco.   At a minimum, the estimate
   of 35000 "high risk" gay men should be less than the estimated total of
   gay males in the city.

   In a random phone surveys, Lemp's study estimated 69,122 openly gay males.  
   In another, more conservative estimate,  based on extrapolating   SFMHS
   rates of AIDS to the entire city, produced a figure of 42,509.  Lemp
   compromised on a middle figure of 55,816 gay males in San Francisco.

   Thus, the 35000 figure for "high risk" seems reasonable.  Note that Lemp's
   phone estimate of 69000+ gay males does not account for persons who would
   refuse to discuss their sexual orientation over the phone, which may have
   been significant.

   A case could be made for counting most of the gay males in the city as
   being "high risk".  Consider, for example two men who play Russian
   roulette, one using 4 of 6 empty chambers on each play, and another who
   uses only 2 of 6 empty chambers.  One has twice the risk of death, per
   trial.   After 25 trials, the "lower risk" man has a 99.996% chance of
   being killed, versus a virtual 100% for the "higher risk" man.

   A similar phenomenon appears to apply to the lion's share of men in San
   Francisco- within less than a decade, infection rates were nearly 50% in
   many areas.   Thus, the distinction between "high risk" and "low risk"
   among gay males may be a false dichotomy.

   As it turns out, the accuracy of the figure for "high risk" gay men is
   somewhat of a moot point.  Later analysis will show that the net result of
   the statistical comparison is relatively insensitive to the effect of
   varying the estimated number of high risk men (Section C.1)

Section 3: HIV Correlation to Vaccine Trial ------------------------ Page 12

   3.4  Vaccine Correlation Analysis

   Referring again to the Lemp study data:

        SFHBVCS: 1978 - 0.3%, 1979 - 4%, 1980 - 15%, 1981- 28%, 1982 - 40%,
        1983 - 46%, 1984 - 47%, 1985 - 48%, 1986 - 48%, 1987 - 49.3%

        SFMHS: 1978 - 0%, 1979 - 2%, 1980 - 4%, 1981- 10%, 1982 - 23%,
        1983 - 42%, 1984 - 48%, 1985 - 49%, 1986 - 49.3%, 1987 - 49.3%

   Both groups start out with zero or near-zero HIV exposure in 1978.    HIV
   infection exploded in both groups from about 1980, onward.

   By 1987, both groups are nearly identical once again, with a nearly 50%
   infection level.  New growth in both groups has slowed dramatically, to
   little or no new increase.

   The behavior is like an "explosion", starting from next to nothing,
   spreading very quickly, and infecting a sizeable percentage of both groups.   
   It is not highly meaningful to examine either the very late period (1983-
   1987), when both groups reached a near-saturation, or to examine the very
   beginning (1979-1980), when hardly anyone was infected.

   The early period from about 1980-1982 shows a more tell-tale difference.  
   In 1982, some 40% of the SFHBVCS shows HIV infection, while only 23% of
   the SFMHS shows infection.

   My question is:  What is the exact statistical probability of this
   difference occurring by "random chance" alone?   This is the essence of
   the computation to follow.

   We have estimated a pool of 35000, equally high-risk gay men.  From these,
   we play "God" and draw a random sample of 23% (using the SFMHS infection
   rate), designating these as the men who will have become HIV+ by 1980.   
   This equals (23% x 35000) = 8050 men.   Of these 8050, some (40% x 359) =
   144 are from the SFHBVCS vaccine group (the actual result for that group).

   What we would have normally expected, on average, would have been
   (23% x 359) = 83 men,  rather than our 144.    The probability for this
   difference between the expected and actual results, by random chance alone,
   is computed by a standard formula:

            T = total population size (=35000)
            v = vaccine subgroup size (=359)
            s = sample size - no. of HIV infected men drawn at random from total (=8050)
            n = no. of men from the vaccine group who were found in that sample (=144)

Section 3: HIV Correlation to Vaccine Trial ------------------------ Page 13

   The odds for drawing exactly "n" men would be given by the formula

         (vCn * (T-v)C(s-n))  /  sCT

   where the notation "xCy" is the computation for total combinations in
   choosing "y objects from a group of x objects":

             xCy = x!/((x-y)!y!)    

   where "!" is "factorial".   E.g.  x! = x * (x-1) * (x-2) * (x-3) ... * 2 * 1

   The odds for drawing "n" or more is computed iteratively.    After
   computing the odds for "n", we then compute the odds for n+1, n+2, ... s,
   and sum these probabilities.

   This computation was performed using a computer program written by the
   author, called "comb.c".  The program source is not listed here, due to
   length, but is available upon request from the author.  It is written in C
   programming language, for Unix or for Microsoft Visual C++.

   The program output for this problem is as follows:.

        Subgroup size = 359
        Total group size = 35000
        Sample size = 8050
        n = 144

        PROBABILITY IS: 2.72128e-13

   This is 2.7 times 10 to the -13th power, which is a unimaginably small
   probability (roughly 1 in 3,700,000,000,000).

   The methodology for the above calculation is very analogous to what the
   researchers themselves used in order to prove that their vaccine prevented
   hepatitis (explained further in Appendix A).  The validity of the logical
   and mathematical approach is easily demonstrated.

   Accounting for margins of error does not alter the conclusion, as
   discussed in Appendix C.

Section 3: HIV Correlation to Vaccine Trial ------------------------ Page 14

   3.5  Lemp Study, Odds Per Year

   Below are computations of the probabilities for all of the years from 1979-
   1982, that the higher proportion of HIV infection in the vaccine group
   might have been random chance.

   In the earliest year of 1979, at the beginning of the vaccine trial, there
   is not a statistically significant difference between the vaccine and non-
   vaccine group.  All of the other years show a significant difference.

   *  1979: SFMHS =2% =700 ; SFHBVCS =4% =14

      Subgroup size = 359
      Total group size = 35000
      Sample size = 700
      n = 14

      PROBABILITY IS: 0.014146      (1 in 71)

   *  1980: SFMHS =4% =1400 , SFHBVCS =15% =54

      Subgroup size = 359
      Total group size = 35000
      Sample size = 1400
      n = 54

      PROBABILITY IS: 5.62915e-17         ( 1 in 18 quadrillion)

   *  1981: SFMHS =10% =3500 , SFHBVCS =28% = 101

      Subgroup size = 359
      Total group size = 35000
      Sample size = 3500
      n = 101

      PROBABILITY IS: 2.2762e-22      (1 in 4 billion-trillion)

   *  1982: SFMHS =23% =8050 , SFHBVCS =40% = 144

      PROBABILITY IS:  2.72128e-13  (this is the previous example)

Section 4: HIV Correlation to Non-vaccine Participation ----------- Page 15

   4  HIV Correlation to Non-vaccine Participation

   4.1  Over-representation in earliest AIDS cases

   Further analysis shows that there is peculiar, unexplained correlation of
   early AIDS infection in the entire San Francisco City Clinic Cohort (SFCCC).   
   These 6875 men were studied by having blood drawn for purposes of
   determining prevalence and transmissibility of hepatitis B.  Only a much
   smaller subgroup of 359 men, drawn from entire SFCCC, had received
   experimental vaccine.

   A number of studies suggest that the SFCCC was over-represented among the
   earliest of AIDS cases, compared to men of equal risk.   These studies do
   not identify if the early AIDS cases from the SFCCC also happened to be
   the same men who had received the experimental vaccine.  If they had in
   fact received the vaccine, then the failure to mention this would have
   been a gross omission on the part of the AIDS researchers.

   If all or some of these disproportionate SFCCC AIDS cases were actually
   attributable to the subset of men in the vaccine trial, then it suggests
   an extremely strong link to the vaccines.

   If the AIDS cases were merely men in the SFCCC who did not receive the
   vaccine, then it raises the disturbing question of what other unknown
   factor could explain the disproportionate HIV infection.

   There is little in the study that should by rights have put men at
   significantly higher risk for getting HIV.     It is unlikely, for example,
   that unsterile or reused needles would have been involved in drawing blood.

   It must be demonstrated that the men in the SFCCC did not have a higher
   rate of early HIV infection simply because of such an obvious factor as
   higher-risk behavior.

   Consider how the shape of a graph should appear when a group at higher
   risk for AIDS infection is compared to a lower risk group.

   The higher-risk group will tend to produce AIDS cases earlier.  The
   numbers of cases will grow faster, thus producing a sharper, faster rising

Section 4  HIV Correlation to Non-vaccine Participation ----------- Page 16

   Below is a chart showing the growth of clinical AIDS in the SFCCC, versus
   the entire city of San Francisco, in the first years of the epidemic:

              --------------         -----------------------------------------

JUL-DEC 1981     6                      12                                     
JAN-JUN 1982     6                      24
JUL-DEC 1982    12                      58
JAN-JUN 1983    18                     100
JUL-DEC 1983    20                     146
JAN-JUN 1984    24                     210
JUL-DEC 1984    38                     280

   The preceding figure is taken from Jaffe, et al  [10].

Section 4  HIV Correlation to Non-vaccine Participation ----------- Page 17

   We would of course expect to find a larger absolute number of AIDS cases
   among gay men in the entire city than in the hepatitis trial, because
   there were approximately 8 times more gay males in the general population,
   than in the hepatitis trial.  However, the growth rate was also
   consistently higher for the city in general, even in relative terms,
   during the early epidemic years.  This fact is easily visible in the
   sharper curve, represented in the chart.  This suggests that these first
   victims from the city as a whole were at even higher-risk for HIV
   infection, than the men in the SFCCC.  Between 1981 and 1984, the numbers
   of new cases in the city increased 20-fold, while the numbers of new cases
   in the SFCCC increased by less than seven-fold.

   The overall gay population of the entire city should have been lower risk
   than the SFCCC, because the SFCCC was composed of men recruited at a VD
   clinic.  However, the very first of the AIDS cases in the city do not
   merely reflect the overall population.  They naturally tend to reflect a
   subgroup of the very highest-risk men in the entire city.  This
   characterization is evidenced by the very fact of their becoming the first
   to be infected, out of the entire population.  Men who have large numbers
   of partners are far more likely to be infected first.  Exceptions to the
   rule would exist, but would be small in comparative numbers.

   Thus, it is more than justified to treat the SFCCC and the total pool of
   men who contracted AIDS in these first few years, from 1981-1984 as being
   at least equal in risk for having acquired AIDS.  This assumption is
   similar to that made previously for the Lemp study, but is even more
   conservative, because the "high-risk" population is defined using a much
   shorter period of time.  The evidence of higher-risk in the general SF
   group is even more pronounced.

   The fact that the men in the city as a whole had a sharper growth curve
   might at first seem to exonerate the hepatitis studies.   However, this is
   contradicted by the significant over-representation of the SFCCC in the
   very first years.   This over-representation suggests that the AIDS
   epidemic was somehow seeded initially among the SFCCC, and then spread
   like wildfire to the rest of the city.

   There is clear evidence that the men from the overall gay population who
   became early AIDS cases were at least equal in risk for getting AIDS, in
   comparison to the men in the SFCCC. Thus, we can estimate statistically
   whether the higher representation of the SFCCC among the earliest AIDS
   cases could be attributed to random chance.

   By the end of 1984, 166 of the SFCCC were diagnosed with AIDS, including
   19 who had moved to other cities (Jaffe, et al [10]).   The remaining
   group of 147 men is our "subgroup size".

   A total of 898 men in the San Francisco Standard Metropolitan Statistical
   Area (SMSA), which includes the SFCCC, were diagnosed with AIDS by the end
   of 1984.  This represents our "total group size".

Section 4  HIV Correlation to Non-vaccine Participation ----------- Page 18

   We are interested in the sample of men who were the first cases, in 1981
   thru 1982,

   so we can compute the odds for the higher SFCCC representation to be the
   result of random chance.

   The SFCC represented 60% of the first victims in 1981 [11] , then it
   quickly plummeted to 38.5% by the last half of 1981, and to 14.6% in the
   last half of 1984.

   Six of the first 10 AIDS cases in San Francisco were members of the SFCCC
   [11], and 11 of the first 24  cases were also members [10].    By Jan 1,
   1983, there were 104 AIDS cases in the city [12].   Of these, 35 were from
   the SFCCC [11] (estimated from previous chart).

   *   6 of first 10 AIDS cases from SFCCC (1981, first half)

      Subgroup size = 147
      Total group size = 898
      Sample size = 10
      n = 6
      PROBABILITY IS:   0.00207447    (1 in 480)

   *  11 of first 24 from AIDS cases SFCCC (1981, last half):

      Subgroup size = 147
      Total group size = 898
      Sample size = 24
      n = 11
      PROBABILITY IS:   0.00057291  (1 in 1700)

   *  35 of first 104 AIDS cases from SFCCC (end of 1982):

      Subgroup size = 147
      Total group size = 898
      Sample size = 104
      n = 35
      PROBABILITY IS:   2.76209e-06   (1 in 362,000)

Section 4  HIV Correlation to Non-vaccine Participation ----------- Page 19

   For 1981, the sample sizes are small, so it becomes more difficult to
   demonstrate highly dramatic differences (for very small samples, nearly
   any outcome tends to be possible, within reason).  However, the
   probabilities are only in the range of a small fraction of a percent, so
   they are revealing enough, in themselves.

   As the years progress and sample sizes increase, the differences become
   more evident, tending to confirm that the pattern is genuine and

   By the end of 1982, it is the most revealing, with a probability of less
   than 1 in 360,000 that the differences between the two groups were due
   simply to chance.

   It should be noted also that 19 of the SFCCC men who developed AIDS are
   not reflected in the preceding statistics, because they had moved to other
   cities.  If any appreciable number of these men were dropped from the
   charts before 1983, then it could significantly raise the levels of
   improbability, to a degree as high as 10 to the -15 
   (1 in 1,000,000,000,000,000)

   It is important to note that men in the SFCCC were not  methodically
   screened for AIDS in the early years of the epidemic.  Thus, there was  
   not a disproportionately higher level of reporting that would distort the
   comparison to other gay men.  At the time of Jaffe's 6-year follow-up study
   [10], "No formal procedures were available to determine if patients who
   were reported to their health departments were cohort members."    For the
   earliest AIDS cases, membership in the SFCCC was determined after the fact
   of diagnosis.  These men had not received any experimental vaccine, so
   therefore had no special reason to be concerned about possible health
   effects.  Epidemiological screening of the SFCCC for AIDS did not begin
   until 1983 [14].

   Men in the SFCCC had the right to refrain from testing for AIDS, as some
   opted to do.  None were forcibly tested.

Section 4  HIV Correlation to Non-vaccine Participation ----------- Page 20

   4.2   High Rate of Seroconversion During Recruitment Years

   This analysis hypothesizes that enrollment in the SFCCC seemed to spark a
   sudden onset of new HIV infection during the time-frame of 1978-1980, when
   men were recruited for this study.

   Is this hypothesis born out by examination of the seroconversion dates for  
   SFCCC participants?

   Estimated dates are provided by Rutherford, et al [14]. in the following,
   reproduced table:

   Table 1, Estimated Year of HIV-1 Seroconversion in San Francisco City Clinic Cohort

   Estimated year                        New
   of seroconversion                    Seroconversions

   1977                                 3
   1978                               114
   1979                               143     
   1980                               121
   1981                                42
   1982                                30
   1983                                 7
   1984                                 9
   1985                                 4
   1986                                 5
   1987                                 3
   1988                                 4
   1989                                 4

   The new seroconversion figures represent a sample of 489 SFCCC men who had
   progressed to full-blown AIDS prior to the Rutherford study.  These men
   were used to study the incubation period for AIDS, from initial HIV

   The Rutherford study claims that 8% of a sample of 2877 men were HIV+ upon
   entry into the SFCCC study.  Of the above 489 subgroup, 64% were
   supposedly seropositive upon entry, and 36% seroconverted within 24 months
   of entry.

   The Rutherford study's estimate of the AIDS incubation period (about 11
   years) is in line with current thinking.   This suggests that the
   estimated dates of seroconversion are also accurate.

Section 4  HIV Correlation to Non-vaccine Participation ----------- Page 21

   The fact that there were men who were claimed to be HIV-positive upon
   entry into the study does not exonerate the study's possible connection to
   the outbreak of AIDS.

   Men were not recruited all at the same time for the SFCCC study.    Most
   were recruited over a two year period, between 1978 and 1980.  The first
   831 participants were recruited over a 5 month period, between June and
   May of 1978 [14].  If men were infected continuously as they were
   recruited, they could also have been quickly spreading HIV through the
   general population at the same time, also infecting new recruits.

   Even if you subtracted entirely from the graph all 312 men who were
   supposedly HIV+ upon entry, there would still be a noticeable bulge in the
   graph of seroconversions, around the years of recruitment.  As computed
   from the table, this would still leave 69 seroconversions through the end
   of the recruitment period in 1980, another 72 in the next two year period
   of 81-82, and only 36 in the entire 7 year period following that,
   from 83-89.

   This is particularly suspicious because rate of HIV spread should have
   continued to accelerate through much of the 1980s, because of the growing
   pool of already-infected men.  It is well documented, how there were
   considerable difficulties in convincing the gay community to close the
   baths, for example, and to change sexual habits.   Given the slowness and
   marginal results of this process, it is not reasonable to see such drastic
   fall-off in the rate of new infections in high-risk men, when the
   critical mass of already-infected men is now so much larger.

   Even if many of these men were supposedly seropositive upon entry into the
   SFCCC, what would explain the extremely high clustering around the
   recruitment years or 1978-1980?   This peculiarity requires explanation,
   in any case.

   Not only is the sharp drop-off of new HIV seroconversions in high-risk gay
   men suspicious, but the paucity of seroconversions before 1978, when the
   trials started, is also peculiar.  For example, there is no evidence that
   HIV infection existed in San Francisco before 1977.   This includes no
   retroactive evidence of transfusion-related AIDS, in the context of
   totally unprotected blood banks, preceding 1977.

   If there were accidental or intentional infection that had taken place in
   during the hepatitis study, stored blood samples could also have been
   contaminated at will, to obscure that fact.

   Regardless of the necessarily-suspect claim of HIV upon entry into the
   cohort, the huge burst of seroconversion matching precisely the
   recruitment period is clear.

   These statistical peculiarities suggest an unnatural, large-scale,
   simultaneous mass-infection that seeded HIV into the gay community of San

Section 4  HIV Correlation to Non-vaccine Participation ----------- Page 22

   4.3  Unreasonable Study Size/Selection

   The government's criteria for choosing its hepatitis study participants
   and its sample sizes ought to seem questionable, in light of the basic
   rules of sampling.

   The 6875 men in the SFCCC represent between 1 in 8 and 1 in 10 all the gay
   men in San Francisco, according to the Lemp estimates.  Was such a large
   percentage of the gay population really necessary for this study?

   If the government had chosen a sample of 1000 gay men to study, its margin
   of error in representing the total population would be 3.1% (95%
   confidence interval).

   If they had chosen a sample of 3438 men (half the size of the SFCCC), the
   margin of error would be about 1.7% (95% confidence interval).  By
   doubling the size of the sample, to 6875 men, reduces the margin of error
   only to 1.2%.  This would represent a classic mistake in choosing sample
   size, by doubling the size and expense of your study, in return for only a
   slight extra return in accuracy.

   This error would be compounded, however.   One of the cardinal rules of
   sampling is that you must select a sample that is representative of the
   larger population that you wish to study.  The assumption for your margin
   of error, as calculated above, depends completely on this.  You cannot
   have "sampling bias", for example favoring young over old, rich over poor,
   urban over rural, etc, based on your method of picking study participants.

   The SFCC was exclusively chosen from urban gay males, by way of a VD
   clinic.    For monitoring purposes, the study would not have been
   representative of the general public, or of the general gay community, or
   even of the general gay urban male population.  It included no
   heterosexuals, no women, no lesbians, no rural or small town gay males,
   and few urban gay males having more average sexual habits.

   For "monitoring" purposes, what would be the point?  So that it could be
   announced, with highest possible degree of accuracy, that in order to
   avoid hepatitis, you should avoid the obvious factors of having thousands
   of sex partners, and the extensive use of dangerous drugs?   To announce
   that people who had already-known risk factors for hepatitis were indeed
   getting lots of hepatitis?

   Furthermore, these hepatitis studies involved not only the full 6.875 gay
   men in San Francisco, but also roughly  9000 gay men in New York City   [13].

   The most seemingly legitimate reason for the large screening program might
   be to identify a subgroup of high-risk men that had never been exposed to
   hepatitis B.  Such men were needed as suitable subjects for the hepatitis
   vaccine trials.

Section 4  HIV Correlation to Non-vaccine Participation ----------- Page 23

   However, even this does not quite make sense, because 3 to 5 times as many
   gay men were screened and found to be negative for hepatitis, than were
   actually used in the vaccine programs.  In NY,  about 10,000 men were
   screened, of which 3200 were HBV negative, of which only 1090 were
   enrolled for the vaccine tests.  In San Francisco, 6704 were screened, of
   which 1676 were HBV negative, of which only 359 were enrolled for vaccine
   tests.   Detailed questionnaires about sexual habits were issued after the
   vaccine selection was made  [18].   It would have been possible to screen
   far fewer men in order to find the necessary vaccine trial participants.

   It seems implausible that the government had an outpouring of concern for
   the health of promiscuous urban gay males, relating to a common disease
   (hepatitis) that involved comparatively few fatalities.  When the fatal
   AIDS epidemic began spreading like wildfire in the gay community, the
   government appeared to be relatively unconcerned, for a long period of

   It also does not appear likely that subjects were chosen for hepatitis
   studies merely to benefit those who were at highest risk for hepatitis.  
   Alaskan Eskimos, including the Dena' Ina Tribe, were also chosen for
   hepatitis vaccine experiments [24].  They also alleged that they suffered
   health effects.  Among their complaints was a charge that they were among
   the lowest risk for hepatitis, and were chosen merely for use as human
   guinea pigs, as an expendable population.

   The selection of such a large and biased population for the hepatitis
   experiment was not consistent with sound rules of sampling, but would have
   been consistent with an intent to infect an "undesirable" population.

Section 5  Epidemiological Anomalies ------------------------------ Page 24

   5  Epidemiological Anomalies

   The following sections discuss various reasons why the early AIDS epidemic
   does not fit a reasonable epidemiological model.

   5.1   Unreasonable Delay of HIV in IV Drug Community

   According to the city of San Francisco [5], the total number of AIDS cases
   in the city, as compiled in 1999,  showed about 79% gay males, 11% IV drug-
   using gay males, and 7% heterosexual  IV drug users.   Among the IV drug
   users, gay males were greater than half.

   Because HIV is a blood-borne virus, crossover of HIV infections into the
   gay and straight IV drug community should have been rapid.  Lemp estimates
   that 1120 gay males in the city were infected as early as 1979.  Yet Lemp
   claims a 0% rate of HIV infection among drug users as late as 1981.

   After even 100 HIV infections, the probability would be 99.9998% that at
   least one of those gay males would have been a drug user.

   Furthermore, the infectivity rate of HIV due to a contaminated needle is
   nearly as high as the infectivity rate of unprotected anal sex [16][17].   
   Once that crossover occurred into the IV community, it should have spread

   Lemp also estimated that 3% of the IV drug community would become infected,
   per year.  Thus, the delay of at least 3 years (from 1978 to 1981) for
   crossover to have occurred in the IV drug users defies normal
   epidemiological  expectations.

   5.2  Anomalies Revealed by Computer Modeling

   It might make intuitive sense to some observers that there are peculiar
   aspects to the manner in which the AIDS epidemic unfolded.

   Throughout the 1970s there was no awareness of any such problem as AIDS,
   in spite of a supposed presence of HIV in human beings, as early as 1930.  
   In the early 1980s, AIDS suddenly exploded with pronounced visible effect,
   almost simultaneously in far-flung locations round the globe: in Africa,
   in Haiti and the Caribbean,  in Europe, in North America.

   Retrospectively diagnosed AIDS cases from earlier in the 1970s exist at
   best anecdotally, in small handfuls.  In some cases, even the anecdotes
   are subject to questions about reliability.

   Epidemics cannot be modeled with great precision, because there are many
   variables that are complicated, or unknown, or unpredictable.  However,
   computer modeling can determine if there are profound inconsistencies in
   terms of parameters such as time, numbers of people infected, infectivity
   rates per sexual contact, numbers of sexual contacts, etc.

Section 5  Epidemiological Anomalies ------------------------------ Page 25

   Appendix E gives the details of an analysis using epidemic modeling
   software that was developed by the author (a software engineer of more
   than 25 years experience).  Included in this discussion are demonstrations
   that the software accurately duplicates real-life experimental results, as
   well as conforming with theoretical, mathematical testing.

   The major conclusions of this analysis are as follows:

   * The rate of HIV infection in the San Francisco Hepatitis B Vaccine
     Cohort is far higher than what could be reasonably expected

   * Full-blown AIDS cases should have been evident many years earlier in
     San Francisco, based on the numbers of individuals infected in the early

   * In order to produce the levels of new infections seen per year in
     San Francisco, it would have required nearly 2000 infections to have
     existed as early as 1976- a time when virtually no HIV has been
     retrospectively discovered in the city.

   * The numbers of men that would be required to appear suddenly in the
     late 1970s, in order to account for the subsequent level of HIV growth,
     is far more than can be reasonably accounted for by natural
     explanations such as vacationing in high-risk areas.

   * Models attempting to explain the HIV growth curve by postulating
     variable rates of HIV infectivity are also inadequate.

   The number of new HIV infections that can occur in a given year depends
   heavily on the size of the existing pool of already-infected persons.   If
   you live in a city where only a handful of persons are infected, your odds
   of coming in contact with those few people is very small.  If many
   thousands of people are infected, then your odds of encountering an
   infected person, and thus becoming infected yourself, are much greater.

   This is why an epidemic "gathers steam" as it progresses, producing new
   cases per year at a faster pace, as more people become infected.

   Similar to the adage that it "takes money to make money", you could also
   say that it "takes infections to make infections".  The notion that small
   handfuls of infected people could spark a sudden explosion on the scale
   that was seen in the early 1980s, is profoundly inconsistent with existing
   knowledge about HIV infectivity rates.   No matter how promiscuous those
   few individuals or their partners might have been, they could not produce
   thousands of new cases within a span of a few years, even if they had
   tried to do so.

Section 6  Unreasonable Approval of the Vaccine ------------------- Page 26

   6  Unreasonable Approval of the Vaccine

   The commercially-made hepatitis B vaccine was considered as "safe,
   immunogenic, and efficacious" in a Sept. 1982 report by Dr. Don Francis of
   the CDC [15], without regard to the fact that gay men in the trials had
   started to become infected with an unknown, new disease, starting in 1981.   
   Quick approval of the vaccine for general use was not prudent.  Even if  
   the vaccine had no connection to AIDS, the researchers would have had no
   way to be certain of that fact, until 1984, when HIV was discovered and
   could be detected in blood products.

   Researchers were already well aware by the early 1970s of the existence of
   "slow" viruses [23].

   For example, the visna virus, a sheep retrovirus with a long incubation
   period, had been discovered as early as 1949 (Fields, Virology, Chapter 55).

   Scientists were also well aware of the dangers for vaccine contamination.

   In an earlier vaccine fiasco, a potentially cancer-causing monkey virus
   (SV-40) had contaminated vaccines in the 1960s and was injected into
   millions of people [22].

   Alaskan Native-Americans also claim to have been harmed by hepatitis
   vaccines[24].  The Yupik Eskimos and other Alaskan tribes were also used
   in hepatitis vaccine experiments.  In a 1990 Council meeting, the Dena'
   Ina Tribe's Health Committee declared an "almost total loss of confidence"
   in the U.S. vaccine programs because of a wide variety of health problems,
   including AIDS-like symptoms, following hepatitis vaccinations.

   In 1985, Dr. Don Francis also painted a rosy picture of the outcome of the
   hepatitis experiments on Alaskan natives, again calling them "safe,
   immunogenic, and efficacious" [25].

   If the approval of the vaccine was not simply motivated by profit, at the
   expense of safety, then there is at least one other possible explanation.  
   Perhaps some scientists were not concerned that test subjects were
   becoming infected, because they knew that the gay men had been infected
   intentionally with HIV during the experiment.  This may be harsh
   speculation, but it is justified by the other wise inexplicable
   irresponsibility entailed in the premature approval of the vaccine, in the
   context of the circumstances and the state of knowledge that existed at
   that time.

Section 7  Historical Context ------------------------------------- Page 27

   7  Historical Context

   The statistical analysis cannot be complete without reference to the
   political atmosphere in the late 1960s and the 1970s.

   In a 1969 Congressional appropriations hearing for the Department of
   Defense, a Pentagon official named Donald MacArthur, a biological warfare
   expert,  stated, "Within the next 5 or 10 years, it would probably be
   possible to make a new infective micro-organism which could differ in
   certain important aspects from any known disease-causing organisms.   Most
   important of these is that it might be refractory to the immunological and
   therapeutic processes upon which we depend to maintain our relative
   freedom from infectious disease." [9]   The proposed budget implied that
   this feat could be accomplished for the relatively modest sum of $10

   In the late 1960s, President Richard Nixon publicly renounced germ warfare,
   except for "defensive research."  In 1971 he ordered a large part of the
   army's biological warfare unit at Fort Detrick, Maryland, transferred over
   to the nearby National Cancer Institute (NCI), where Dr.  Robert Gallo
   would later discover the AIDS virus (HIV) in 1984.  With the transfer of
   the biological war unit to the NCI, the army's DNA and genetic engineering
   programs were coordinated into anti-cancer and molecular biology programs.  
   It is quite possible that military biowarfare research could have
   continued under the guise of legitimate cancer research [9]

   The Russians also signed the Biological and Toxin Weapons Convention in
   1973, but immediately set up Biopreparat, a huge program for biowarfare
   research.  Only in the late 1990s did the Yeltsin government admit to the
   existence of this secret program, which astonished American scientists
   with its scope, involving some 40 facilities.  The Soviets embarked on
   this program in large part because they had believed that the United
   States had not ended its bioweapons program, but had simply hidden it away.  
   A British intelligence officer recalled, "The notion that the Americans had
   given up their biological weapons program was thought of as the Great
   American Lie." [26].

   As far back as the 1950s, the United States maintained the ability to kill
   and incapacitate targeted people with biological weapons (see "In Search
   for the Manchurian Candidate", John Marks [27]).  The Technical Services
   Staff (TSS) of the CIA paid the Army Chemical Corp's Special Operations
   $200,000 per year in return for operations systems to infect enemies with
   disease (Chapter 5, [27]).   Dozens of germs and toxins were maintained
   for killing purposes.

   Specific instances of assassination efforts using poisons and diseases are
   documented, targeting various  figures including Fidel Castro and Patrice
   Lumumba of the Congo.   A Newsday article reprinted in the Boston Globe(1/
   9/77) reports that CIA operatives received swine flu virus at a CIA
   biological warfare training station, and then attempted to spread the
   virus to Cuban pigs.  Numerous other allegations of biological assaults
   against Cuban crops, livestock, and civilian populations are recorded.

Section 7  Historical Context ------------------------------------- Page 28

   Did the U.S. every really stop biowar activities, as Nixon claimed?    The
   Congressional Church Committee hearings in the mid 1970s explored abuses
   in the CIA, and revealed that millions had continued to be spent on
   unauthorized biowar research.

   Former CIA Director William Colby testified concerning a device called a
   "non-discernible microbioinoculator",  which was designed to deliver fatal
   injections of toxins, in such a way that could not easily be detected in
   an autopsy [27].  The CIA's apparent goal was to develop various ways of
   killing that would leave no trace.    It was also revealed that the CIA
   had stored enough shellfish toxin to kill a half-million people, an amount
   admitted to be far in excess of research needs.

   Colby wrote in his memoirs that his admission about the
   "microbioinoculator" had "blown off the roof", and led to his immediate
   dismissal.  President Ford replaced Colby with future President George
   Bush, who gained a reputation as a staunch defender of CIA secrecy.

   Undertones of violence existed in the Nixon administration, proven by his
   taped remarks urging the beating of anti-war protestors, and by G. Gordon
   Liddy's admission in his autobiography of discussions contemplating the
   murder of political columnist Jack Anderson [28].

   The Chicago Tribune published transcripts of Nixon's Oval Office remarks
   about gays and other minorities.   Nixon openly defamed Mexicans, blacks
   and Jews, saying Mexicans were prone to steal;  blacks lived like "a bunch
   of dogs"; and homosexuals "destroyed" strong societies.   Nixon spoke of
   the historical need for the Catholic Church to "clean out" its homosexuals.  
   He admired societies that tried to eliminate homosexuals, saying, "Let's
   look at the strong societies. The Russians. Goddamn, they root `em out.
   They don't let `em around at all."  Deploring the alleged takeover of San
   Francisco by "fags", Nixon proclaimed, "...I can't shake hands with anybody
   from San Francisco."

   The climate of hatred against gays intensified in the 1970s with thousands
   of homosexuals coming out of the closet with unprecedented political
   demands.  It is easy to imagine that they might have been vulnerable to
   covert government-sponsored medical experiments, similar to those secret
   radiation experiments that had been conducted on unsuspecting citizens
   during the Cold War years, up to the year 1974, when the government's
   investigation of the records documenting these crimes ended.

   During the 1970s, there was extensive animal retrovirus experimentation
   undertaken as part of Nixon's "War on Cancer".   Animal viruses were
   transferred between species and manipulated genetically.  As a result, new
   cancerous and immunosuppressive diseases were produced experimentally.   
   Could the AIDS virus have arisen from this dangerous and unprecedented
   experimentation?   Is it possible that anecdotal cases of pre-1960s AIDS
   in Africa were misdiagnosed, or might have represented false positives, or
   contaminated blood samples?  Could such cases have been contrived as a
   "cover-up" in order to discredit research pointing to a man-made origin
   of the AIDS epidemic, in vaccine programs of the 1960s and early 1970s in

Section 7  Historical Context ------------------------------------- Page 29

   In the 1970s Don Francis and Max Essex (later to become top scientists in
   AIDS) experimented extensively with feline leukemia virus (FELV), an HIV-
   like retrovirus that produced a disease in cats, similar to human AIDS.  
   Early in the AIDS epidemic, both scientists suspected that AIDS might be
   caused by a retrovirus, because the new disease was so reminiscent of
   these earlier studies. [30] .

   No human retrovirus was known until 1978 when Robert Gallo discovered a
   retrovirus that caused a rare type of human leukemia.  In the early 1980s,
   a second leukemia virus was discovered by Gallo.  These human "T-cell"
   leukemia viruses were termed HTLV-1 and HTLV-2.  When Gallo discovered HIV
   in 1984, he initially called it "human T-cell leukemia/lymphoma virus", or
   HTLV-3.   Later, the term "human immunodeficiency virus". (HIV) was
   substituted, which represented only a change in name.

   After Don Francis completed his work with the HIV-like cat retrovirus, he
   joined the Centers for Disease Control and headed the hepatitis B vaccine
   experiments, using gay men as guinea pigs in San Francisco and other
   cities (the very same experiments discussed throughout this document) [31].

   This document has demonstrated a statistical correlation between the
   government-sponsored hepatitis experiments and the outbreak of HIV in the
   gay male volunteers.  The historical context reinforces the concern that
   the correlation might have resulted from intentional infection of these
   men.  There is proof of extreme bigotry in high office.    There is a
   proven record of intrigue, deception, political corruption, unauthorized
   experimentation, and use of human guinea pigs.  There was a great deal of
   relevant scientific research which suggests that it would have been
   possible for our government to have either discovered or created HIV,
   before the hepatitis trials began.  This combination of circumstances
   makes the requirement to reinvestigate the hepatitis experiments all the
   more compelling.

Section 8  Conclusions  ------------------------------------------- Page 30

   8  Conclusions

   In the early 1980s, gay men who were given experimental hepatitis B
   vaccine showed significantly higher rates of HIV infection, compared to
   other gay men of equally high risk for HIV infection, in the SFMHS group.  
   The odds for these differences being due to random chance alone are
   extremely small, by some measures in a range of one-millionth of one-

   The justification for saying that the gay men in the SFMHS group were of
   equally high risk is based not merely on their characterization as such,
   in the Lemp study, or merely on the fact that the men were chosen from
   high-risk tracts within San Francisco.  It is based on comparison of HIV
   growth patterns within the two groups, which were highly similar for many
   years in the study period.  It is also based on the fact that the SFMHS
   group "caught up" and even surpassed the vaccine group in HIV infection,
   in spite of the "head start" that the vaccine group had in the early years. 
   This suggests that the non-vaccine SFMHS group may have been even higher
   risk.  Therefore, the vaccine group should by rights have shown noticeably
   lower prevalence of HIV, much less the higher levels that they actually
   showed, compared to the SFMHS.

   The SFCCC group showed even stronger evidence of being lower risk in
   comparison to the "control" group to which they were compared, which
   consisted of other early AIDS victims in the city.  The higher initial
   rate of AIDS in the SFCCC showed that mere participation in the hepatitis
   studies, in any capacity, had a statistically significant association with
   AIDS diagnosis, beyond what could be explained by risk alone.

   In the past, the high rate of HIV/AIDS among men in the hepatitis study
   has been attributed to their "high risk" status, or to random chance,  or
   else it has been denied that the rate of HIV/AIDS was in fact higher than
   that of the general population.  This analysis has shown those rather
   simplistic explanations to be inadequate.

   With these significant possibilities effectively ruled out, it is still
   conceivable that one might try to find alternative, benign explanations.   
   However, the search for benign explanations might begin to strain the
   imagination, no less than the thought of contaminated vaccines, or
   intentional infection by some covert, unknown means.

   Because the production of the vaccine involved use of pooled blood from
   high-risk gay men, the possibility of HIV-contaminated vaccines has
   sometimes been imagined as a tragic accident.

   However, vaccine (Heptavax-B produced by Merck Sharp & Dohme) was
   inactivated using three steps: pepsin, urea, and formaldehyde (formalin)
   (Francis et al., 1986) (see reference [6]):

      "In this study, we demonstrate that each of the three inactivation
       steps used in the manufacture of Heptavax-B independently will
       inactivate the infectivity of high-titered preparations of the
       AIDS virus"

Section 8  Conclusions -------------------------------------------- Page 31

   If this claim is correct, then not even the use of pooled blood from gay
   men should have caused vaccine contamination.

   Furthermore, samples of the vaccines were tested retroactively for HIV.  
   It was claimed that no HIV was detected in the vaccines.

   An important question is why there appears to be unusually high rates of
   HIV/AIDS associated with mere participation in the hepatitis studies, even
   for men who received no vaccine.  These differences clearly do not seem
   attributable merely to a higher risk status.

   The hepatitis trials in New York City would be worthy of similar analysis.   
   In 1980, the rates of HIV infection among vaccine trial participants were
   in the range of 20%, even higher than the rate in San Francisco [9].

   All possible explanations deserve consideration and investigation, even
   the most politically sensitive explanations.    Evidence suggests that
   accidental infection should have been unlikely, yet a nearly undeniable
   statistical correlation remains to the vaccines and the hepatitis studies.   
   To the degree that accident can be ruled out, the possibility for
   intentional infection is strengthened.

   Significant numbers of people do not approve of homosexuals, and would not
   object to their removal from society.   Historically, "undesirable"
   populations, such as prisoners, mentally retarded and others were used for
   unethical experimentation.    Even if no such criminal malice existed in
   the hepatitis studies, the best way to establish that fact would be to
   treat the possibility seriously, and investigate it thoroughly.

   If HIV is an old virus, present long before the 1970s, then this would
   only make it easier to suppose that the virus could have been discovered
   without public announcement, and then tested on an undesirable population.

   In the 1990s, horrific details of the government's Cold War
   experimentation during the 1940s up to the 1970s came to light.   Thousands
   of covert radiation experiments were performed on children, the mentally
   ill, hospitalized patients, pregnant women, Native Americans, and other U.
   S. citizens.  Thus, during the 1970s, it would not have been unprecedented
   for government scientists to experiment on gay men, the most hated
   minority in America.

   Our government should long ago have made full disclosure as to the fates
   of the men who volunteered for these experiments.  Exactly how many men
   who received the experimental vaccines died of AIDS, and in what years?

Section 8  Conclusions -------------------------------------------- Page 32

   Why has the scientific community failed to notice these profound
   statistical correlations?  It is understandable that these realizations
   might have escaped the notice of laymen, but they were well within the
   ability of trained scientists to discern easily.   The fact that they did
   not, even after allegations of a vaccine link, is evidence that these
   allegations have not been taken seriously enough.

   It is essential for the scientific community to explain the high rates of
   HIV/AIDS in the hepatitis study members.

   There is no suggestion being made here that starting a man-made epidemic
   would have been anything other than an act of madness.   It would be
   madness unprecedented in scale, but not historically unprecedented in its
   recklessness or cruelty.  It was little short of madness how the Reagan
   administration essentially ignored the new disease, tried to slash the CDC
   budget, and ordered the Surgeon General not to mention the word "AIDS" in
   public.  If a government was capable of ignoring a new disease to such a
   degree, it is only a short step further to infer that they might have been
   foolish enough to precipitate the disease.  It may have been an act of
   irrational religious fervor. Perhaps it was imagined, or foreseen, that
   the virus would be confined largely to risk groups.   Perhaps the
   perpetrators simply did not care about any citizens who did not live
   according to strict Christian sexual morality.  Perhaps the perpetrators,
   even if Americans, were infiltrated by foreign enemies who might have
   wished the entire country's destruction.

   It is not a necessity for this study to explain the exact nature of the
   madness, but merely to document why an act of madness has very probably

   The connection between the origin of the AIDS epidemic and the government
   experiments has been dismissed by the AIDS epidemic.  However, the
   statistical analysis presented here demonstrates a definite correlation.   
   A reopening of the entire matter is in order.

   The purpose of this document is not to cast final judgement concerning the
   origin of AIDS in the gay community.   It is to demonstrate that there is
   a strong and suspicious link, which has no obvious explanation, between
   the outbreak of AIDS and the government-sponsored hepatitis studies. The
   purpose is to call for investigation of this important question.

Section 9  Refuting Counter-Arguments  ---------------------------- Page 33

   9  Refuting Counter-Arguments

   This section will review and refute various attempted criticisms
   concerning the link of AIDS and the vaccine studies.

   * Perhaps the men in the vaccine trial had sex with each other, and
     infected each other

     It is true that the men signing up for a trial might be friends who also
     have sex with each other.  However, this factor could be equally true
     for the SFMHS (non-vaccine group), as for the SFHBVCS (vaccine group).

     The sponsors of a vaccine trial would not publish lists of everyone
     involved in the trial.  Fraternizing among participants would likely be
     limited to small groups of friends who might have known each other prior
     to the trial.  It would not likely be intermingling among the entire

     The men were also chosen because of a high-risk profile, meaning that
     they were regular clients of institutions such as bars and baths.   They
     would have been prone to sexual encounters with large number of the
     general gay male population, not merely with a small set of friends.

   * Perhaps the men in the vaccine group became complacent because of feeling
     "protected" by the vaccine, and started practicing more risky behavior

     The analysis already demonstrated that risk behavior was if anything,
     even higher in the "control" (non-vaccine) group that was used for

     The vaccine would have protected only against hepatitis, even if it
     worked.   It would not have protected against a host of other venereal
     diseases, including herpes, syphilis, and gonorrhea.    The sponsors of
     the trial would have been remiss if they had not explained this to the
     men involved.

   * AIDS has a 10-year incubation period.  Therefore, the men must have
     been infected prior to the start of the trial, because most became
     infected in less than 10 years.

     The Lemp study was specifically measuring the dates of initial HIV
     infection, and not the development of full-blown AIDS.    The incubation
     period is not relevant, in this context.

     The fact that we are looking at HIV seroconversion rather than "AIDS"
     also eliminates another argument used in the past against efforts to
     suggest a vaccine connection: that a reaction to the vaccine
     challenged the immune system, and hastened the development of AIDS. 
     Here, we are looking only at virus exposure, not disease symptom

     In the case of the SFCCC, where the analysis was examining initial rates
     of AIDS diagnosis, estimates of the HIV seroconversion dates were

Section 9  Refuting Counter-Arguments  ---------------------------- Page 34

     provided by the Rutherford study.  The dates were heavily clustered
     around the dates of recruitment into the hepatitis study.

   * Perhaps the men receiving the vaccine were simply monitored more
     closely, and their HIV status was detected more quickly than for men who
     did not receive the vaccine

     For the Lemp study, this line of argument clearly does not apply.   Both
     groups, vaccine and non-vaccine, had stored blood samples that were
     taken annually and later examined for HIV.

   * Perhaps the men in the vaccine group were actually higher risk than
     the non-vaccine group

     There are degrees of "high-risk".  Simply because the non-vaccine group
     was characterized in the Lemp study as "high-risk", it does not
     necessarily mean that they were equally as high risk as the men in the
     vaccine group.  However, this analysis is not relying merely on
     subjective descriptions.  It assesses their risk level by comparing
     their actual patterns of HIV growth.  The non-vaccine control group, the
     SMHS appeared, if anything, to be even higher risk for HIV infection.

     Furthermore, the SFHBVCS vaccine group was chosen only from men who had
     no previous exposure to HBV (hepatitis B virus), and were chosen from
     men who seemed in good health at the outset of the trial (else, it would
     have invalidated the vaccine trial results).  HBV was sexually spread
     and epidemic among gay men, which was part of the reason that gay men
     were chosen for the vaccine trial, in the first place.    One would
     expect that there would be a significant correlation between HBV
     exposure and HIV exposure.   The exact extent of this factor is
     difficult to measure, but it is nonetheless an additional, qualitative
     reason to believe that the men in the vaccine trial might have been
     actually lower in overall risk for previous HIV exposure, at the start
     of the vaccine trial.

     Dr. Lemp's study itself states, "Although these vaccine cohort members
     were recruited from sexually transmitted disease clinics, they represent
     lower-risk since none of the cohort members were seropositive for
     hepatitis B virus at time of recruitment."   Dr. Paul M O'Malley,
     Project Director of the SF Dept of Health AIDS research study, also
     concurred, "Their blood had not in 1980 shown signs of infection with
     hepatitis B, which can be spread through sexual activity.   The subjects
     were therefore assumed to be less sexually active than other SF clinic
     visitors."  If it is necessarily true that the men were less active,
     perhaps it would be more accurate to say that they had at least  
     beaten the odds, in terms of encountering infected partners.

   * HIV growth is an exponential function.  This can magnify differences
     over time between groups, compared to what you would see with a flat
     rate of infection, such as by exposure to carcinogens.

     It has been demonstrated that the hepatitis/vaccine study groups had, if
     anything, even less risk for HIV than the "control" groups used for this

Section 9  Refuting Counter-Arguments  ---------------------------- Page 35

     comparative analysis. All of the men in both groups are from the same
     geographical area, visiting the same limited number of bars/ baths in
     the city, showing similar behaviors.  It is therefore justified to treat
     the hepatitis/vaccine study groups as being at least equal in risk,
     with any error in that assessment being weighted against a conclusion
     that implicates a correlation to the hepatitis study.

     As long as the characterization of equal risk levels is accurate, then
     the earliest AIDS cases, which represent an essentially random sample
     taken from the total pool of men, should not disproportionately reflect
     either of the subgroups.

   * Perhaps the statistical sample size is too small in order to draw

     This objection is not applicable to the probability calculations of "n
     or more" in this document.  Similarly. if a coin is flipped with 30
     heads in a row, it is not "too small" of a number of trials in order to
     draw a confident conclusion.   This point is discussed further in the
     general statistical primer, Appendix F.

     Where "sample size" becomes an issue is in the question of whether the
     SFMHS group is sufficiently large in order to use as an estimate of
     the HIV rate for all high-risk men.

     The Lemp study in fact uses SFMHS for the whole gay male population of
     the city, not distinguishing between "higher/lower" risk.  The study
     states, "Since the SFMHS is a population-based probability sample, its
     seroprevalence estimates are likely to be representative of HIV
     seroprevalence for homosexual and bisexual men in San Francisco."

     As a double-check, we can take the year of 1982, and do another
     calculation based on a reversed hypothesis:  suppose that the 40% rate
     of HIV infection in the SFHBVCS group, rather than the 23% HIV infection
     rate of the SFMHS group, was the "real" rate of HIV for high-risk men. 
     What would the probability then be that the 799 men in the SFMHS might
     show their 23% rate of HIV infection, by random chance alone?   Might
     this be within the bounds of normal possibility?

     In this case, our subgroup size is 799 and the total group size is still
     35000.  The sample size (the number of men that we expect to have HIV in
     1982) is 40% of 35000 =14000 .   Of these 14000, the SFMHS group
     represents (23% x 799) = 184 men.   What we want is the probability
     that 184 or fewer would be in the SFMHS group.  Our program calculates
     "or greater", so we can simply calculate the odds for "185 or greater",
     and subtract this from one.

     Using the program, this produces an answer of zero, or essentially no
     chance (too small of a value to represent).   Thus, it does not matter
     which end of the spectrum that we choose as representing HIV
     prevalence in the overall gay population.  The difference between the
     vaccine and control groups in either case is more than what we could
     expect by random chance.

Section 9  Refuting Counter-Arguments  ---------------------------- Page 36

   * Perhaps the data concerning HIV infection rates are not reliable

     HIV antibody tests are more than 99% accurate [7].    The false-positive
     rate for the Elisa HIV antibody test is only 1 to 5 per 100,000 assays.

     The false negative rate is only  1 in 450,000 to 1 in 660,000 [8] . It
     can be assured that these tests must be reliable, because they have had
     long use in protecting our nation's blood supply.

     These error rates have no significant impact on the computed

     Furthermore, any false-positives or false-negatives would have tended to
     affect both the vaccine and non-vaccine cohorts in equal proportions,
     thereby tending to cancel out, in any case.

   In the absence of better data for the period prior to 1982, it is
   justified to use the Lemp data as representing the best of what is
   currently available.   The conclusion derived from this data would also
   represent the best conclusion that could be drawn at this time.

Section 10  About the Author -------------------------------------- Page 37

   10  About the Author

   I (Tom Keske) am a gay/AIDS activist in the Boston area, originally from
   Ohio.  I have been a data communications software engineer for 25 years, B.
   S.(with honors) in Computer Engineering, Case Western Reserve University,
   `74.  Education includes modest background in statistics, strong in math
   and programming.  I am an internet activist, writing frequently about AIDS
   and its origins.  I am age 48, and in a committed relationship with a
   lifetime partner of 28 years, Daniel.

   Academic honors include National Merit Scholar and National Honor Society.

   I was "survey statistician / computer programmer" responsible for
   telecommunication software at the Bureau of Census, in Washington, D.C.,
   from 1974 - 1979, where I received an "outstanding employee" award.   I was
   President of the Board of Directors,  Bradbury Park Condominiums, in the
   Maryland suburbs of Washington.  In 1979, I relocated to the Boston area,
   where I am currently employed as a senior staff engineer for a Fortune 500
   vendor of communications equipment.  Between 1980 and the present, I have
   worked on development of statistical multiplexors, intelligent matrix
   switches, a data PABX, Ethernet bridges,  multiprotocol routers and cable

   I have also worked in the area of data encryption, and have performed
   statistical analysis of encryption methods.  In the late 1990's, I
   successfully defended a self-designed encryption program in the face of a
   $1000 reward posted to break the code.  During college, I had faced
   discrimination at the National Security Agency during a job interview at
   Fort Meade, with polygraph questions about homosexuality.

   I am a supporter of progressive causes, such as in past participation as
   an Amnesty International "Freedom Writer".

   My activities instilled a keen awareness of political corruption and abuse,
   which led to an interest in research and criticism of the CIA and
   intelligence establishment.

   As a gay activist, I engaged in public vigil/hunger strike in front of the
   Massachusetts State House in support of the gay civil rights bill, which
   passed immediately afterward, after more than 15 years of effort.  There
   was modest coverage in smaller, local papers.  A state legislator said that
   the protest was lending moral weight to the cause.  I have also engaged in
   civil disobedience protests, outside the Supreme Court, during the first
   March on Washington, and in various cities including Raleigh, North
   Carolina and Atlanta, Georgia.

   I am health-conscious, and for hobbies enjoy swimming, hiking, bicycling,
   and chess.

   Thomas R. Keske,  205 Warren St.,   Randolph, Mass.   02368
   email:      (781) 961-1571

Section 11  References -------------------------------------------- Page 38

   11  References

   [1] The River: a Journey to the Source of HIV and AIDS, by Edward
       Hooper (former  BBC Africa correspondent)      (Penguin; Little, Brown,
       New York, 1999)


   [3] Lemp GF, Payne SF, Rutherford GW, Hessol NA, et al:
       Projections of AIDS morbidity and mortality in San Francisco,
      JAMA 1990 Mar 16;263(11):1497-501        PMID: 2407871; UI: 90172481

      The abstract of the Lemp 1990 study reads as follows:
      Abstract: To develop a model for predicting acquired immunodeficiency
      syndrome (AIDS) morbidity in San Francisco, Calif, through June 1993, we
      combined annual human immunodeficiency virus seroconversion rates for
      homosexual and bisexual men and for heterosexual intravenous drug users
      with estimates of the cumulative proportion of the population with AIDS
      by duration of human immunodeficiency virus infection and with estimates
      of the size of the at-risk populations. We projected AIDS mortality by
      applying Kaplan-Meier estimates of survival time following diagnosis to
      the projected number of cases. The median incubation period for AIDS
      among homosexual and bisexual men infected with the human
      immunodeficiency virus was estimated to be 11.0 years (mean, 11.8 years;
      95% confidence interval, 10.6 to 13.0 years). The model projects 12,349
      to 17,022 cumulative cases of AIDS in San Francisco through June 1993,
      with 9,966 to 12,767 cumulative deaths.

   [4] University of San Francisco (home page, includes links to faculty):

   [5] San Francisco HIV/AIDS Statistics as of Nov. 28, 1999,                                                          

   [6] Francis DP, Feorino PM, McDougal S, et al. The safety of the hepatitis
       B vaccine. Inactivation  of the AIDS virus during routine vaccine
       manufacture,  JAMA 1986 Aug 15;256(7):869-72.

   [7] Centers of Disease Control  NAC, from Guide to Information and
       Resources on HIV Testing, 1997

   [8] CDC FAQ:

Section 11  References -------------------------------------------- Page 39

   [9] AIDS and the Doctors of Death: An Inquiry into the Origin of the
       AIDS Epidemic (1988), and Queer Blood: The Secret AIDS Genocide Plot
       (1990),  by Dr. Alan Cantwell, Jr., Aries Rising Press, PO Box 29532,
       Los Angeles, Calif. 90029   (

   [10] Jaffe HW, Darrow WW, Echenberg DF, et al.: The Acquired
        Immunodeficiency Syndrome in a Cohort of Homosexual Men, A Six-Year
        Follow-up Study, Annals of Internal Medicine.   1985;103:210-214

   [11] Centers For Disease Control, Morbidity and Mortality Weekly Report,
        Sept. 27, 1985 / Vol 34  / No. 38

   [12] Moss AR,  Bacchetti P, Osmond D et al: Incidence of the Acquired
        Immunodeficiency Syndrome in San Francisco, 1980-1983, Journal of
        Infectious Diseases, Vol 152, No. 1, July 1985

   [13] Koblin BA,  Morrison JM, Taylor PE, et al.: Mortality Trends in a
        Cohort of Homosexual Men in New York City, 1978-1988, American Journal
        of Epidemiology, Vol 136, No. 6

   [14] Rutherford GW, Lifson AR, Hessol NA et al:  Course of HIV-1 infection
        in cohort study of homosexual and bisexual men: an 11-year follow up
        study, Br Med, Vol 301, Nov 24, 1990

   [15] Francis DP, Hadler SC, Thomppson SE, et al:  The prevention of hepatitis
        B with vaccine. Report of the Centers for Disease Control multi-center
        trial among homosexual men,, Ann Intern Med 1982 Sep;97(3):362-6 
        PMID: 6810736, UI: 82282328

   [16] Kaplan EH, Heimer R, A model-based estimate of HIV infectivity via
        needle sharing, J Acquir Immune Defic Syndr 1992;5(11):1116-8, Yale,
        PMID: 1403641, UI: 93020182

   [17] Vittinghoff E, Douglas J, Judson F, et al: Per-contact risk of human
        immunodeficiency virus transmission between male sexual partners,
        Am J Epidemiol 1999 Aug 1;150(3):306-11,   PMID: 10430236, UI: 99357305

   [18] Padian NS, Shiboski SC, Glass SO, Vittinghoff E:   
        Heterosexual transmission of human immunodeficiency virus
        (HIV) in northern California: results from a ten-year study,
        Am J Epidemiol 1997 Aug 15;146(4):350-7    PMID: 9270414, UI: 97416464

   [19] Godfried JP, Hessol NA, Koblin BA, et al: Epidemiology of Human
        Immunodeficiency Virus Type 1, Infection among Homosexual Men
        Participating in Hepatitis B Vaccine Trials in Amsterdam, New York
        City, and San Francisco, 1978 - 1990, Amer Journal of Epidemiology,
        Vol 137, No .8, 1993.

Section 11  References -------------------------------------------- Page 40

   [20] University of Southern California,

   [21] Jacques JA, Koopman JS, Simon CP, Longini IM: Role of primary
        infection in epidemics of HIV infection in gay cohorts, J Acquir
        Immune Defic Syndr 1994 Nov, PMID: 7932084, UI: 95017548

   [22] "The Polio Vaccine and Simian Virus 40", by By T.J. Moriarty,

   [23] "Retroviruses- An Introduction", JAMA HIV/AIDS Information Center,

   [24] "Alaska Health Issues and Indigenous Peoples" (video), Mary Ann Mills,
        Bernadine Atchison,  Delice Calcote, July 1991 Arctic Village Health
        Conference.  These women are Activists against medical experimentation
        on Alaska Native communities. 

   [25] Heyward WL, Bender TR, Francis, DP et al: The control of hepatitis B
        virus infection with vaccine in Yupik Eskimos Demonstration of safety,
        immunogenicity, and efficacy under field conditions,  Am J Epidemiol 1985
        PMID: 3160233, UI: 85248405

   [26] "The Bioweaponeers", the New Yorker, March 1998, by Richard Preston

   [27] In Search of the Manchurian Candidate, by John Marks, 1988,
        Times Books, ISBN: 0-440-20137-3 Senator Edward Kennedy said of this
        expose,"John Marks has accomplished   what two U.S. Senate committees
        could not".

   [28] Secret Agenda, by Jim Hougan, Random House, 1984, ISBN: 0-394-51428-9.
        The Los Angeles Times called this book "a monument of research and
        fact-finding". Hougan was Washington Editor of Harper's magazine, and
        helped produce the Emmy Award winning documentary, "Confessions of a
        Dangerous Man"

   [29] "The Role of Robert Gallo in the Origin of AIDS", Kwame Ingemar
        Mr. Ljungqvist is editor the Swedish scientific journal, "Science of
        the 21st Century"

   [30] 11/96 "1 in 10 Talk Show" interview with Max Essex

   [31] Harvard Public Health Review, "The Gathering Storm", by Sarah Abrams
        (describes the career of researcher Don Francis)

Section 11  References -------------------------------------------- Page 41

  12  Acknowledgments

  Much thanks to Dr. Alan Cantwell, Jr., author of "Queer Blood (1993)" and
  "AIDS and the Doctors of Death" (1988) [9] for help in editing this
  document and reviewing the facts.  Dr.  Cantwell has spent some 14 years
  investigating the hepatitis vaccine experiments.

  Thanks also to Billi Goldberg, San Francisco AIDS researcher/activist,
  whose discussions of the 1990 Lemp study inspired this further statistical

  13  Document Reproduction

   This document may be freely reproduced and distributed, with attribution.

Appendix A  Demonstrating the Validity of the Statistical Approach- Page 42

   Appendix A  Demonstrating the Validity of the Statistical Approach

   The logical approach used in this document is to compare groups of gay men
   who received vaccines, or who otherwise participated in hepatitis studies,
   with other groups of similar gay men who did not engage in these
   activities.  This type of statistical analysis is virtually identical to
   how researchers compare their own test vaccine group to a control/placebo

   Researchers are typically trying to prove that their vaccine group showed
   statistically lower incidence of the targeted disease, compared to the
   control group.  Or, perhaps, they might try to show that the vaccine group
   showed statistically higher levels of protective antibody response,
   compared to the control group.

   The only difference in this document's use of the same technique is that
   the aim is to show the statistical presence of another disease, instead of
   the absence of the disease that the vaccine tries to prevent.   The
   "control" groups are defined retrospectively, by identifying groups of men
   who were of demonstrably equal or higher risk for HIV/AIDS.

   The following is an abstract of a Swiss study involving hepatitis B
   vaccine.  This will demonstrate how the vaccine researchers are using very
   similar statistical calculations:

     "Evaluation of tolerability and antibody response after recombinant human
      granulocyte-macrophage colony-stimulating factor (rhGM-CSF) and a single
      dose of recombinant hepatitis B vaccine.

      Tarr PE, Lin R, Mueller EA, Kovarik JM, Guillaume M, Jones TC

      Sandoz Pharma Ltd, Basel, Switzerland.

      Recombinant human granulocyte-macrophage colony stimulating factor
      (rhGM- CSF) has been shown to augment antigen presentation by
      macrophages and dendritic cells in vitro, and to increase antibody
      responses to injected antigens in experimental animals.   To evaluate
      the usefulness of rhGM- CSF as a vaccine adjuvant, 108 healthy
      volunteers were randomly assigned to receive an injection of rhGM-CSF
      (n = 81) or placebo (control group; n = 27), followed by an injection
      with recombinant hepatitis B vaccine into the same site.   During the
      study period of 28 days, protective antibody titers to hepatitis
      surface antigen (anti-HBs10 mIU ml-1) were observed in 11 of 81
      subjects receiving rhGM- CSF, but in none of the controls (P = 0.035). 
      Injections were well tolerated.  A single i.m. or s.c. injection of 20-
      40 micrograms of rhGM- CSF significantly enhances antibody responses
      when given at the same site as recombinant hepatitis B vaccination.

      Publication Types: Clinical trial Randomized controlled trial

Appendix A  Demonstrating the Validity of the Statistical Approach- Page 43

      PMID: 8961505, UI: 97120835"

   The study involved 81 people in a vaccine group, and 27 in a placebo
   (control) group, for a total of 108.  In the vaccine group, 11 showed
   protective antibodies, but none did in the placebo group.  When the
   researchers say (P = 0.035), they mean that the probability of this
   outcome is 3.5%.

   Using the same program in [27], we get the same result:

       Subgroup size = 81
       Total group size = 108
       Sample size = 11
       n = 11
       PROBABILITY IS: 0.035144

   Note that the researchers must make assumptions that are virtually
   identical to what we must make in evaluating whether the vaccine was
   causing HIV infection.   When they try to prove that the vaccine prevented
   hepatitis B infection, they must assure that the vaccine group is at equal
   risk for hepatitis, compared to their placebo group.   There cannot be
   differences in age, general health, risk of exposure, etc, that might
   account for the different outcomes between the two groups.  The
   researchers  must compute statistically that the differences in hepatitis
   rates between the two groups are not merely a result of "random chance".

   There is no such thing two absolutely identical groups, but the process of
   "averaging out" can mitigate the effects of small differences.

   The Francis, et al study evaluated the gay hepatitis B vaccine using the
   same type of analysis [15].   It estimated the effect of the vaccine based
   on about 907 men in a vaccine group , and 495 in a placebo group (this is
   including other cities as well, not just San Francisco).   It had to assume
   that these men were roughly equal in risk for acquiring hepatitis, just as
   we have had to demonstrate equal risk of men for acquiring HIV.

   The Francis study concluded that the vaccine was beneficial in preventing
   hepatitis B on the strength of 56 new hepatitis infections in the control
   group, versus only 11 in the vaccine group (probability = .0004, or 1 in
   25000).   These results are not nearly as compelling as the figures cited
   in this document  linking HIV infection to the hepatitis studies and

   Yet, the Francis figures were used to justify dispensing the vaccine to
   millions of people, whose health and life  would be in the balance.     It
   is therefore difficult to argue that the figures in this document are not
   on as solid of a scientific basis, and adequate justification to draw a

Appendix B  Letter From Dr. George Lemp --------------------------- Page 44

   Appendix B   Letter From Dr. George Lemp

   >Dear Mr. Keske:

   Thank you for your interest in my research. The data cited appear
   reasonably accurate. I assume the author looked at the published graphs
   and guessed the approximate data points. My time doesn't allow me to try
   to dig up the original data points, but I took another look at the graph
   and the author's guesses seem reasonable (perhaps off by 1% at a few
   points). Who was the author and where were these data cited? The JAMA
   article would be on file at any University or major hospital medical
   library in your area. JAMA is widely held by libraries and should be
   available. If you have trouble finding it, please email your address and
   we'll mail you a reprint.


   George Lemp

   >X-Mailer: QUALCOMM Windows Eudora Pro Version
   >Date: Mon, 31 Jan 2000 09:38:08 -0800
   >From: Universitywide AIDS Research Program <>
   >Subject: Fwd: 1990 Study Data
   >>From: "Thomas Keske" <>
   >>To: <>
   >>Subject: 1990 Study Data
   >>Date: Fri, 28 Jan 2000 23:24:48 -0500
   >>X-Mailer: Microsoft Outlook Express 5.00.2919.6600
   >> Jan. 28, 2000
   >>Dr. George Lemp
   >>University of California

Appendix B  Letter From Dr. George Lemp --------------------------- Page 45

   >>Dear Dr. Lemp,
   >>I hope that my emailing will not impose on your time. I much
   >>appreciate all the work that you have done for AIDS research.
   >>I have a very brief question, and would much appreciate if you
   >>could reply.
   >>I am trying to follow a thread on, which quoted data
   >>from your 1990 study, showing rates of HIV in the early 1980's:
   >> SFHBVCS: 1978 - 0.3%, 1979 - 4%, 1980 - 15%, 1981- 28%, 1982 - 40%,
   >> 1983 - 46%, 1984 - 47%, 1985 - 48%, 1986 - 48%, 1987 - 49.3%
   >> SFMHS: 1978 - 0%, 1979 - 2%, 1980 - 4%, 1981- 10%, 1982 - 23%,
   >> 1983 - 42%, 1984 - 48%, 1985 - 49%, 1986 - 49.3%, 1987 - 49.3%
   >> SFHBVCS = San Francisco City Clinic Cohort Study
   >> SFMHS = San Francisco Men's Health Study
   >>The author said that this data was extracted from charts in the 1990
   >>study, but I have been unable to locate the full text of the study:
   >> Lemp GF, Payne SF, Rutherford GW, Hessol NA, Winkelstein W Jr, Wiley JA,
   >> Moss AR, Chaisson RE, Chen RT, Feigal DW Jr, Thomas PA, Werdegar D.
   >> Projections of AIDS morbidity and mortality in San Francisco. JAMA 1990
   >> Mar 16;263(11):1497-1501
   >>Could you please tell me if the data above appears to be
   >>reasonably accurate, or how I could obtain the full study?
   >>I am asking only as an interested layman.
   >>Thanks very much for your time.
   >>Regards, Tom Keske

Appendix C  Error Analysis for Lemp Data Calculations ------------- Page 46

   Appendix C  Error Analysis for Lemp Data Calculations

   C.1  Effect of Variation in High Risk Population Estimate

   It might seem at first glance that the calculation could be in error if
   the estimate of the size of the "high risk" gay population is too high.  
   In actuality, it turns out that lowering the estimate of the total high
   risk population size will work to decrease the probability that the result
   could be attributed to random chance.   The is because reducing the
   estimate of the high risk population size also lowers the "sample size" of
   HIV+ men that we expect to draw in any one year (23 percent of the total
   high-risk men, for the example year of 1982).

   Below is a listing of the computed probabilities for different values of
   the estimated number of high-risk gay men in San Francisco, ranging from
   as many as 100000, to as few as 10000.

   As it can be seen, these variations matter little in the resulting

    Subgroup size = 359, Total group size = 100000, Sample size = 23000, n=144
    PROBABILITY IS: 3.3273e-13

    Subgroup size = 359, Total group size = 50000, Sample size = 11500, n=144
    PROBABILITY IS: 2.98706e-13

    Subgroup size = 359, Total group size = 30000. Sample size = 6900, n=144
    PROBABILITY IS: 2.58323e-13

    Subgroup size = 359. Total group size = 25000, Sample size = 5750, n=144
    PROBABILITY IS: 2.40079e-13

    Subgroup size = 359, Total group size = 20000, Sample size = 4600, n=144
    PROBABILITY IS: 2.14932e-13

    Subgroup size = 359, Total group size = 15000, Sample size = 3450, n=144
    PROBABILITY IS: 1.78359e-13

    Subgroup size = 359, Total group size = 10000, Sample size = 2300, n=144
    PROBABILITY IS: 1.21827e-13

   The largest of these probabilities is roughly 1 in 3,000,000,000,000.

Appendix C  Error Analysis for Lemp Data Calculations ------------- Page 47

   C.2  Effect of Errors in HIV Infection Rate Figures

   There could have been errors in reading the chart data from the Lemp study.  
   Dr. Lemp had suggested that this might have amounted to a percent or so
   (Appendix B).

   The following computations test the effect of errors in the probability
   calculation for 1982,  by reducing the number of HIV+ men in the vaccine
   group, and increasing the number of HIV+ in the non-vaccine group, in
   increments of 1%.  Both of these adjustments work to increase the
   probability that the outcome might be attributable to random chance.    The
   computations range from 1% adjustments, to 3% adjustments.  This was only
   an informally suggested error rate, so it is being tripled for safety,
   with worst case assumed jointly for each affected variable:

   * Adjusting sample size +1% and vaccine group size -1%

     non-vaccine = 8400    (24% of 35000)
     vaccine     =  140     (39% of 359)

     Subgroup size = 359, Total group size = 35000, Sample size = 8400, n = 140
     PROBABILITY IS: 1.63957e-10

   * Adjusting sample size +2% and vaccine group size -2%

     non-vaccine = 8750    (25% of 35000)
     vaccine     =  136     (38% of 359)

     Subgroup size = 359, Total group size = 35000, Sample size = 8750, n = 136
     PROBABILITY IS: 4.00199e-08

   * Adjusting sample size +3% and vaccine group size -3%

      non-vaccine =  9100   (26% of 35000)
      vaccine     =   133    (37% of 359)

     Subgroup size = 359, Total group size = 35000, Sample size = 9100, n = 133
     PROBABILITY IS: 2.41677e-06

   Allowing for errors of  +3% in the HIV+ sample size and -3% in the number
   of HIV+ vaccine group men  (jointly) has the effect of improving the odds
   that the correlation could be a product of random chance, but not to a
   significant degree (worst case of roughly 1 in 400,000)

Appendix D  Letter from Case Western Reserve Statistics Department  Page 48

   Appendix D   Letter from Case Western Reserve Statistics Department

   The following note from the CWRU Statistics Dept was in response to a
   query that I made (appended), trying to validate the general reasoning
   behind the analysis.  In this query I had rephrased the question, to avoid
   biasing,  as one of evaluating the safety of a "food additive" (instead of
   a vaccine) that was being tested as a possible "carcinogen" (cancer-
   causing, instead of AIDS-causing).   I posed the question using the
   identical numbers from the vaccine analysis, for the year 1980 in the Lemp
   study, for group size, sample, size, etc.   Below is the email exchange:

   From: Joe Sedransk
   Department of Statistics, CWRU
   Cleveland, OH  44106-7054

   Mr. Keske: Your reasoning is mostly correct. The main assumption is that in
   the absence of the food additive the mortality rate would be 4%; that is,
   that the lab animals are "similar" to the general population (with a mortality
   rate of 4%). [I'd ask how the mortality rate of 4% was determined.] Then you
   would find the probability of 54 or more deaths out of the 359 lab animals,
   assuming a mortality rate for each of 4%. (There is also an assumption that
   the events (life/death) are independent among the 359 animals. This would
   usually be true, but should be verified.) I did a crude calculation using a
   normal distribution approximation (approximating what you did) and found that
   the probability of 54 or more deaths is extraordinarily small. The only
   problem with your formulation is that it is 54 or more out of 359 rather than
   out of 1400. I hope that this helps.

   Joe Sedransk

   At 11:29 PM 2/4/00 -0500, you wrote:
   > Jan. 31, 2000
   >Dear Mr. Sedransk,

   > I am an alumnus of the CWRU class of `74, in Computer
   > Engineering.

   > I was wondering if it would be too much trouble if you could help
   > to clarify my understanding of a simple type of statistics problem,

Appendix D  Letter from Case Western Reserve Statistics Department  Page 49

   > or if you could direct me to  another resource. I am trying to
   > understand how to evaluate the following type of

   >      A group of 359 lab animals using a food additive showed
   >      a 15% rate of cancer in a year (=54 animals). The normal rate
   >      of cancer, measured in a total population of 35000 such animals,
   >      was 4% (=1400 animals)
   >      QUESTION: Is this a normal statistical variation, or should
   >      it be judged that the food additive is unsafe?

   >  It seems to me that this is similar to a problem where you
   >  have 35000 marbles in a bag, 359 are black and the rest white.
   > If you draw a random sample of 1400 marbles, what is the probability
   > of getting 54 or more black marbles, by random chance alone?
   > This can be computed from the binomial distribution curve. I've computed
   > the probability as 5.6 x (10 to the -17). Therefore, my conclusion is
   > that the food additive should almost certainly be suspected as
   > carcinogenic, and should not be approved for mass consumption. This
   > question came up only as part of a newsgroup discussion (nothing related
   > to business and school). We were (embarrassingly) unable to agree on the
   > answer. Could you please help us to settle it?
   > Regards,
   >Tom Keske (class of `74)

   NOTE: the issue of how the 4% figure was derived is explained earlier in
   the document (basically, taking the HIV rate of the high-risk SFMHS group
   of gay men as being representative for other high-risk gay men in San

   The 54 men of the 359 in the SFHBVCS (vaccine) group of gay men is the
   number that actually acquired HIV by 1980 (15% of the group, per the Lemp
   study, as opposed to the roughly expected value of 4%).    The "random
   sample" of all HIV positive men for 1980 is 1400 (4% of 3500).  In this
   1400 is included the 54 from the group of 359 SFHBVCS men.  Later, I chose
   1982 rather than 1980 as the main focus, because the data was measured
   rather than extrapolated.

Appendix E  Software Epidemic Modeling Analysis ------------------- Page 50

   Appendix E  Software Epidemic Modeling Analysis

   Various epidemiological anomalies concerning the origin and spread of HIV
   can be demonstrated through the use of computer modeling software.

   The modeling software referenced in this section was developed by the
   author.  The sources are not listed here because of length (more than 1500
   lines, for two programs), but are available on request from the author.   
   The programs can run on a PC with Microsoft Visual C++, or any ANSI
   standard C compiler.

   The software is general-purpose and flexible, capable of using any
   modeling assumptions that the user might to make, and allowing various
   different models to be tested.

   The programs do not contain built-in assumptions about parameters such as  
   infectivity rates or degrees of risk behavior.   These parameters are
   defined by means of an interactive dialogue, when the program is run.

   The vepid.c software is capable of specifying any number of risk subgroups
   that engage in particular mixes of activities, at different frequencies.  
   Infectivity rates may be time-varying, to account for factors such as the
   stages of HIV infection, where the first few weeks might involve higher

   The program also ask for initial rates of HIV prevalence, when the
   modeling period begins.

   The program keeps track of each individual member of the modeled
   population, and whether they are currently infected, or not.   The members
   of the population are randomly paired for sexual contacts, according to
   their risk group's quota for the year, evenly spread throughout the months
   of each year.    Partners are chosen based on their willingness for a
   compatible activity.

   When an uninfected person is paired with an infected person, the program
   decides whether the uninfected person will become infected or not.    This
   is random, but is kept strictly within the bounds of the average
   probability for infection, based that person's role in the current contact.

   The program prints totals of new and cumulative infections, for each year.

   The epid.c program is a simpler and faster epidemic modeling package that
   models a single population group, with up to two specified active/passive

Appendix E  Software Epidemic Modeling Analysis ------------------- Page 51

   The modeling examples that follow will all  take a number of measures to
   produce "worst case" projections for HIV growth, when looking at the
   general gay population:

   * Both receptive and insertive sexual roles are treated as having the same
     infectivity as the more risky "receptive" role.

   * No account is made for monogamous or partially monogamous partners.   All
     members of the target population are treated as if being promiscuous.

   * No account is made for safe practices, such mutual masturbation, dildos,
     condoms, etc.

   * The rate of sexual activity is assumed for the entire population is
     assumed to be as high as  for the "high risk" men in the vaccine trials.

   E.1  Per-Contact Infection Rates

   Published figures exist for per-contact probabilities of infection for
   various sexual acts/roles  with infected partners[16][17][18]. This makes
   it possible to do computer modeling to examine the spread of HIV.   
   Average rates of infectivity, per-contact with an HIV+ partner, are:

   *       Anal receptive:  .0082
   *       Anal insertive:   .0067
   *       Oral receptive:   .0006
   *       Oral insertive:    0 (slight, theoretical only)
   *       Vaginal, male-to-female:  .0009
   *       Vaginal, female-to-male:  .0001125

   In order to model, you also need data as to the sexual practices and
   frequencies in the target populations.   For the hepatitis B vaccine trial
   participants in San Francisco, this is listed as 67 contacts with
   different partners per year [18].  For New York gay men in the vaccine
   trials, the rate was lower, at about 40 partners per year.    Virtually all
   men reported a mix of both anal and oral sex, with lower-risk oral sex
   being somewhat more prevalent.

   Since these figures are for high-risk men, it would be a generous over-
   estimate to apply the same rate to all gay men in the city.

Appendix E  Evidence of Program Accuracy -------------------------- Page 52

   E.2  Evidence of Program Accuracy

   E.2.1  Consistency with Independent Mathematical Test

   As a check whether the program is working correctly,  we can try a simple
   case that approximates a coin-flipping problem, where we can compute the
   expected answer by another means.  Say that you have a population of
   100000 where half are infected. Say that the probability of "infection" is
   50% (= .5), and that these people pair up for a single sex act (50000
   pairings).  How many should be infected?

   The program says:

   % epid

   Enter Population Size ( <= 100000): 100000
   Enter number initially infected: 50000
   Probability, infection per ACTIVE  contact, type #1: .5
   Probability, infection per PASSIVE contact, type #1: .5
   Probability, infection per ACTIVE  contact, type #2: .5
   Probability, infection per PASSIVE contact, type #2: .5
   Enter average number of contacts per year: 1
   Enter number of years: 1
   Enter random seed (any number between 1 and 4294967295): 987987347
   New infections in year #1 = 12442, GRAND TOTAL = 62442
   You might suppose that the expected value of new infections is (50000 * .5) =
   2500.  However, you must take into account that the pairings are random,
   not simply pairings of infected persons with uninfected persons.  When an
   infected person is paired with another infected person, or an uninfected
   person is paired with another uninfected person, nothing changes.  The
   question is, how many pairings of uninfected and infected persons should
   there be?   The answer that we really expect is about half of the number
   of pairings of infected + uninfected partners.

   This is computed using the "combinations" function, described earlier. 
   The total pairings are

   (100000 C 2) = 5e+09.   Pairings of two infected or two uninfected
   partners  would  each amount to  ((50000 C 2) / 5e+09) = 25%.   The mixed
   pairings would constitute the remaining 50%.

   Thus, we should expect roughly (50000 * 0.5 * 0.5) = 12500 new infections,
   versus the program's projected 12442, which is very close to expected
   (within bounds of expected, random variation).

   E.2.2  Consistency with Real-Life Experimental Results

   A California study of heterosexuals  [18] followed 360 heterosexual woman
   who were HIV negative, but had regular male partners who were HIV+.   These
   women continued to have unprotected sex with their male partners.  In a
   ten-year period, the study reported 68 new HIV infections among the 360

Appendix E  Evidence of Program Accuracy -------------------------- Page 53

   The vepid.c modeling software comes reasonably close in attempting to
   duplicate the results of the California study.  Starting with 360 infected
   men, the program reported 73 new infections among the women- very close to
   the reported value of 68.

   Below is the output of the "vepid.c" epidemic modeling software, for this

   Subgroup #1 represents the 360 males (all initially infected). Subgroup #2
   represents the infected men's 360 female partners (0 initial infections).

   The abstract of the study did not list a number of sexual contacts per
   year, so I made a conservative estimate of one intercourse every other
   week (26 contacts per year).

   % vepid
     Enter Total Population Size ( <= 100000): 720
     Enter no. of activities to model:  1
     Does activity #1 involve exactly 2 partners (y or n)? y

     Enter av probability of infection,
            activity #1, ACTIVE   role: .0001125

     Enter av probability of infection,
            activity #1, PASSIVE role: .0009

     Enter size of subgroup #1: 360

     Enter activity #1, ACTIVE , average no. contacts per year
            for subgroup #1:   26

     Enter activity #1, PASSIVE, average no. contacts per year
            for subgroup #1:   0

     Enter number initially infected for subgroup #1:  360
     Enter size of subgroup #2: 360

     Enter activity #1, ACTIVE , average no. contacts per year
            for subgroup #2:   0

     Enter activity #1, PASSIVE, average no. contacts per year
            for subgroup #2:   26

     Enter number initially infected for subgroup #2:  0

Appendix E  Evidence of Program Accuracy -------------------------- Page 54

     Enter number of years to model: 10

     Enter random seed (any number between 1 and 4294967295): 987987987

           AV prob infection, ACTIVE , 0.0001125    num_adjust,  = 0

           AV prob infection, PASSIVE, 0.0009       num_adjust,  = 0

     POP SIZE: 720

        TOTAL FOR SUBGROUP 0 = 360

        Contacts/yr for activity #1, ACTIVE : 26

        TOTAL FOR SUBGROUP 1 = 360

        Contacts/yr for activity #1, PASSIVE: 26

     New infections in year #1 = 13, GRAND TOTAL = 373
     New infections in year #2 = 9, GRAND TOTAL = 382
     New infections in year #3 = 8, GRAND TOTAL = 390
     New infections in year #4 = 9, GRAND TOTAL = 399
     New infections in year #5 = 7, GRAND TOTAL = 406
     New infections in year #6 = 6, GRAND TOTAL = 412
     New infections in year #7 = 6, GRAND TOTAL = 418
     New infections in year #8 = 3, GRAND TOTAL = 421
     New infections in year #9 = 7, GRAND TOTAL = 428
     New infections in year #10 = 5, GRAND TOTAL = 433
     TOTAL CONTACTS: 187200
     Subgroup #1 infections: initial = 360, new = 0, total = 360
     Subgroup #2 infections: initial = 0, new = 73, total = 73

   E.3  First Year, SFHBVCS

   The first modeling is of the 359 men in the vaccine trial, who went from .
   3% infection (1 man) in 1978 to 4% infection (14 men) in a single year.  
   Is this rate suspiciously high?  The following treats the 360 men as a
   "closed" population, having sex with each other (which should make cases
   rise even faster).   The average number of partners was increased from 67
   to 104, to make it even more conservative.   The program output follows:

Appendix E  Evidence of Program Accuracy -------------------------- Page 55

   % epid
   Enter Population Size ( <= 100000): 360
   Enter number initially infected: 1
   Probability, infection per ACTIVE  contact, type #1: .0082
   Probability, infection per PASSIVE contact, type #1: .0082
   Probability, infection per ACTIVE  contact, type #2: .0006
   Probability, infection per PASSIVE contact, type #2: .0006
   Enter average number of contacts per year: 104
   Enter number of years: 1
   Enter random seed (any number between 1 and 4294967295): 24525
   New infections in year #1 = 2, GRAND TOTAL = 3

   The observed number of infections was nearly 5 times higher than expected
   by a generous modeling estimate.

   To double check that the program's modeling is not simply too low, we can
   try another set of years, with a larger initial pool of infected men.   In
   1982, 40% of the 359 were infected (144 men).  By 1983, 46% were infected
   (165 men).  The percent of the total population becoming newly infected is
   higher (6% versus 3.7%) and the absolute numbers of men newly infected is
   higher (21 versus 14).   The only difference is the pool of men initially

   For this, the program shows:

   % epid

   Enter Population Size ( <= 100000): 360
   Enter number initially infected: 144
   Probability, infection per ACTIVE  contact, type #1: .0082
   Probability, infection per PASSIVE contact, type #1: .0082
   Probability, infection per ACTIVE  contact, type #2: .0006
   Probability, infection per PASSIVE contact, type #2: .0006
   Enter average number of contacts per year: 104
   Enter number of years: 1
   Enter random seed (any number between 1 and 4294967295): 24525
   New infections in year #1 = 42, GRAND TOTAL = 186

   The program in this case shows more men being infected that actually
   observed, demonstrating that it is not simply a matter of the infectivity/
   frequency estimates that causes our previous low value.  What makes the
   difference is the number  initially infected.

   The computer model is saying that in order to have extremely high rates of
   new HIV growth, it is necessary to have a significantly large initial pool
   of infected men.  High rates of HIV growth are not feasible in a scenario
   where only a small handful of men are initially infected.  When such an
   unreasonably high rate of HIV growth is observed, it suggests that some
   mechanism exists to spread the virus that is above and beyond simply a
   high rate of sexual contact.

Appendix E  Evidence of Program Accuracy -------------------------- Page 56

   E.4  Patient Zero Scenario

   For this test, the program estimated the course of HIV growth over a 20
   year period, starting with a single, infected person (a "Patient Zero"
   type of scenario), for a gay population of 100,000, having behaviors
   similar to high-risk San Francisco men.

   % epid
   Enter Population Size ( <= 100000): 100000
   Enter number initially infected: 1
   Probability, infection per ACTIVE  contact, type #1: .0082
   Probability, infection per PASSIVE contact, type #1: .0082
   Probability, infection per ACTIVE  contact, type #2: .0006
   Probability, infection per PASSIVE contact, type #2: .0006
   Enter average number of contacts per year: 67
   Enter number of years: 20
   Enter random seed (any number between 1 and 4294967295): 4536356356
   New infections in year #1 = 2, GRAND TOTAL = 3
   New infections in year #2 = 5, GRAND TOTAL = 8
   New infections in year #3 = 4, GRAND TOTAL = 12
   New infections in year #4 = 5, GRAND TOTAL = 17
   New infections in year #5 = 4, GRAND TOTAL = 21
   New infections in year #6 = 7, GRAND TOTAL = 28
   New infections in year #7 = 12, GRAND TOTAL = 40
   New infections in year #8 = 10, GRAND TOTAL = 50
   New infections in year #9 = 12, GRAND TOTAL = 62
   New infections in year #10 = 19, GRAND TOTAL = 81
   New infections in year #11 = 37, GRAND TOTAL = 118
   New infections in year #12 = 39, GRAND TOTAL = 157
   New infections in year #13 = 55, GRAND TOTAL = 212
   New infections in year #14 = 64, GRAND TOTAL = 276
   New infections in year #15 = 81, GRAND TOTAL = 357
   New infections in year #16 = 144, GRAND TOTAL = 501
   New infections in year #17 = 168, GRAND TOTAL = 669
   New infections in year #18 = 242, GRAND TOTAL = 911
   New infections in year #19 = 351, GRAND TOTAL = 1262
   New infections in year #20 = 433, GRAND TOTAL = 1695

   By the end of the 4th year of the 20-year period, there would have been
   about 50 infections.  In 10 more years, most of these cases would have
   progressed to full-blown AIDS.  At that time, there were still only about
   250-300 infections, total.

   When AIDS broke out, it took only a few dozen usual cases of Kaposi's
   Sarcoma before it was apparent to the medical establishment that there was
   an unusual problem.   Thus, a realistic model of HIV growth says that the
   AIDS epidemic should have become apparent, at a time when HIV prevalence
   was still quite low.   The rapid saturation of HIV in the gay community,
   subsequent to the initial outbreak of AIDS, points to the fact that there

Appendix E  Evidence of Program Accuracy -------------------------- Page 57

   was a mass, simultaneous infection of a larger number of men.

   E.5  Estimated Seed Size in SF

   How many initial, simultaneous, mass infections would have to suddenly
   appear in the late 1970s, in order to account for the rates of explosive
   growth that followed?  For this discussion, this is what is meant by the
   "seed size".

   Lemp's data shows near zero infection in the San Francisco gay population  
   in 1978, rising to 49.3% for the entire gay male population of the city,
   by 1987.   This would be approximately (56000 * .493) = 26708 HIV
   infections in 9 years.

   The first reported case of transfusion AIDS in San Francisco was in 1982,
   4 years after the start of the hepatitis study recruitment.  Blood supply
   screening began in 1985.  The first retroactively estimated case of
   transfusion related HIV infection was 7 years earlier, in 1978, also
   coinciding with the start of the hepatitis study [20].

   To be generous, we can push the date of essentially-zero HIV prevalence to
   1976, the year before the first back-dated projections of gay HIV
   seroconversions listed for high-risk men, cited by Rutherford [14].

   Approximately how many men would it take for a seed size, in order to get
   26708 HIV infections by 1987 (11 years)?

   This can be estimated by running the modeling program repeatedly, taking
   an initial guess, and then working up or down, in iterative attempts.

   As it turns out, the necessary seed size in 1976, as a conservative
   estimate, would need to be between 1900 and 2000 men:

   % epid
   Enter Population Size ( <= 100000): 56000
   Enter number initially infected: 1900
   Probability, infection per ACTIVE  contact, type #1: .0082
   Probability, infection per PASSIVE contact, type #1: .0082
   Probability, infection per ACTIVE  contact, type #2: .0006
   Probability, infection per PASSIVE contact, type #2: .0006
   Enter average number of contacts per year: 67
   Enter number of years: 11
   Enter random seed (any number between 1 and 4294967295): 11324234
   New infections in year #1 = 625, GRAND TOTAL = 2525

Appendix E  Evidence of Program Accuracy -------------------------- Page 58

   New infections in year #2 = 845, GRAND TOTAL = 3370
   New infections in year #3 = 1070, GRAND TOTAL = 4440
   New infections in year #4 = 1351, GRAND TOTAL = 5791
   New infections in year #5 = 1732, GRAND TOTAL = 7523
   New infections in year #6 = 2097, GRAND TOTAL = 9620
   New infections in year #7 = 2616, GRAND TOTAL = 12236
   New infections in year #8 = 3046, GRAND TOTAL = 15282
   New infections in year #9 = 3487, GRAND TOTAL = 18769
   New infections in year #10 = 3906, GRAND TOTAL = 22675
   New infections in year #11 = 3949, GRAND TOTAL = 26624

   E.6  From Where Comes the Seed?

   There is no recorded evidence of extensive HIV infection in the gay
   community, anywhere in America, in the mid-1970s.   Traveling and vacation
   within the U.S. borders could not be a sufficient factor to account for
   simultaneous mass infection of 2000 men in San Francisco, merely in the
   space of a few years.

   Immigration from other U.S. cities in time period also could not explain
   the number of men simultaneously mass-infected,  given the lack of
   evidence for any appreciable degree of HIV elsewhere in the country.

   The fact of a few anecdotal cases of supposed HIV infection from earlier
   years, such as a case claimed in 1959 in St. Louis, do not alter this fact.  
   Having a few stray cases, even if these are not simply myths, does not
   create a scenario to allow rapid infection of a large number of men within
   a few years.

   The same is true even of travel and vacation to other foreign locations,
   such as Africa or Haiti.  As an example, following is an estimate of what
   would happen if nearly the entire gay population of San Francisco
   vacationed in Haiti for a couple weeks, mingling with 100% infected men,
   and having an average of 3 sexual contacts during that vacation:

   % epid
   Enter Population Size ( <= 100000): 100000
   Enter number initially infected: 50000
   Probability, infection per ACTIVE  contact, type #1: .0082
   Probability, infection per PASSIVE contact, type #1: .0082
   Probability, infection per ACTIVE  contact, type #2: .0006
   Probability, infection per PASSIVE contact, type #2: .0006
   Enter average number of contacts per year: 3
   Enter number of years: 1
   Enter random seed (any number between 1 and 4294967295): 222345
   New infections in year #1 = 352, GRAND TOTAL = 50352

Appendix E  Evidence of Program Accuracy -------------------------- Page 59

   In other words, a grand total of only 350+ infections.  Of course, nowhere
   near the whole gay male population of SF men is going to vacation in Haiti
   in the space of a couple years, nor will this number immigrate.

   It is a "Catch-22" which forbids large, sudden simultaneous mass
   infections:  it requires a large pool of existing infections.  To make a
   large pool of infections takes time, when you are starting from only a few
   infections.  In this necessary time, AIDS would reveal itself much earlier. 
   The degree of apparent seeding shows an artificial nature, more consistent
   with a hypothesis of unnaturally produced mass infection, such as in the
   hepatitis experiments.

   E.7  Variable Infectivity Per Stage

   A 1994 study at University of Michigan (Jacquez, et al) [21] attempted to
   explain the sharp rise and rapid fall-off of HIV infection, stating that
   "Thousandfold differences in transmission probabilities by stage of
   infection are needed to fit the epidemic curves".   Their hypothesis was
   that the initial infection stage, characterized by flu-like symptoms,
   would cause an infectivity rate 1000 to 3000 times higher than the rate of
   infectivity in the "long, asymptomatic phase", lasting 10 years or more.

   The per-contact risk for anal sex is broken out in this study as "0.1-0.3
   per anal intercourse in the period of initial infection , 10(-4) to 10(-3)
   in the long asymptomatic period, and 10(-3) to 10(-2) in the period
   leading to AIDS."

   When fed into the program, allowing a 1-month duration for an initial
   infection period, and using the higher of the Jacquez figures in each case (=
   0.3 for initial infection, .001 for asymptomatic) this still did not
   appear to explain the pattern of HIV growth in the early epidemic years.  
   In fact, when starting with a single infection, over a span of 11 years,
   the variable-rate infectivity figures actually came out to be  lower than
   flat rate figures used in the previous examples (trying 56000 men, 200
   initially infected, for 11 years).  Only in the first year did the Jacquez
   figures produce a higher rate of infections (113 versus 59).  In later
   years, it fell off sharply (after 5 years, 699 total for variable
   infectivity, 795 for flat rate; after 11 years, 1988 for variable rate and  
   3992 for flat rate).

   This is probably because the high rate of .1-.3 only applies to an initial
   infectivity period that is very short (2-8 weeks).  After that, the listed
   infectivity figures for Jacquez are actually lower (.001 versus .0082),
   for a much longer period of time (10 years).

   There are additional, possible objections to the notion that a high rate
   of infectivity in the "initial infection" stage could account for the
   early explosion of AIDS.  This initial period is characterized by symptoms

Appendix E  Evidence of Program Accuracy -------------------------- Page 60

   such as headache,  vomiting and diarrhea  This is not a scenario where
   even promiscuous gay men are likely to seek, or succeed in finding, a lot
   of partners.

   If the body were that overcome with huge amounts of virus, the symptoms
   might be more severe than merely flu-like (one might imagine that the
   person would be dying).  The authors of the Michigan study acknowledged
   that the questions of viral load during the different stages were a matter
   of controversy.

   If the body is that vulnerable to massive proliferation of virus, when
   initially exposed, then it is more difficult to understand why the
   asymptomatic phase, with lower viral load, should be all that less
   infectious.  It would seem only to require a small amount of virus to
   cause infection, if the virus can duplicate that quickly and freely in an
   unprepared host.

   In the case of IV drug injection, the amount of virus is far less than   in
   a typical amount of semen, yet the infectivity is very efficient.    This
   also suggests that the amount of virus required for infection would not
   necessarily need to be great.

   It is possible also that the infectivity of HIV has changed over time.   It
   is not in the best evolutionary interest of a virus to kill its only
   natural host.  It is a commonplace phenomenon for viruses to becomes less
   virulent over time.  However, if such drastic changes have occurred merely
   within the last 20 years, then it might suggest that HIV was a relatively
   "new" virus, and detract from the likelihood that it has been infecting
   humans since the 1950s or 1930s.

   More studies concerning viral loads at different stages of infection, and
   concerning the effect of viral load on infectivity of unprotected sex,
   would be useful.

   There is a risk of circular reasoning in the example of the University of
   Michigan study-  a "good" model must fit the observed curve.   Thus, the
   model cannot test whether the observed curve is a natural phenomenon- that
   is an implicit assumption .

   At this point, it appears more likely that researchers are stretching to
   find explanations for the high initial rates of HIV spread, and the sharp
   drop-off that followed.   Perhaps part of the reason for these contortions
   is the refusal to examine a more controversial hypothesis: an artificial,
   mass seeding of HIV infections into the gay population.

Appendix F  General Statistical Primer -------------------------- Page 61

   Appendix F  General Statistical Primer

   There is sometimes a naive tendency to assume the cynical attitude that
   "you can prove anything with statistics."   However, you do not live in a
   world of truths that possess absolute certainties.  You live in a world
   that is probabilistic in nature.  For better or worse, statistics is one
   of your best tools for assessing truth and falsehood in the world around
   you.  Your ability to be deceived by statistics, or to be enlightened by
   it, will depend in large part on how much effort you put into acquiring a
   good command of the subject, so that you can critically evaluate
   statistical claims.

   The vaccine analysis is similar to a problem in drawing samples of marbles
   of different colors from a jar.   Suppose that you had a jar with 100
   white marbles and 100 black marbles, evenly mixed.   You close your eyes,
   and draw a sample of 20 marbles at random.

   Because there are equal numbers of white and black marbles in the jar, you
   would expect, on average, that the 20 that you draw would also be roughly,
   evenly mixed- about 10 white marbles and 10 black.   Of course, this is
   only an average, that you would expect to find over many similar trials.  
   In any one sample of 20 marbles, you might have a few more white, or a few
   more black.  This is a normal variation by random chance, or what you
   would call "luck of the draw".

   Take an extremely simple case- a jar has one white marble and one black
   marble.  You close your eyes and draw one marble.   What is the chance
   that it will be white?

   When this problem is fed into the computer program listed in the appendix,
   the answer is:

   * Chance of drawing a white marble, when pulling one marble at random from a
     jar containing 1 white marble and 1 black marble:

   Subgroup size = 1
   Total group size = 2
   Sample size = 1
   n = 1

   A probability of .5 means 50%, or an even 50-50 chance, the same as the
   chance of getting "heads" when flipping a coin.

   When you flip a coin, or draw from two colored marbles, there is no
   particular reason that one outcome would be favored over another, so you
   tend over many trials to get half (50%) of each result.   This same logic
   applies to any two events where there exist no logical reasons to favor
   one outcome over the other.    If you randomly painted an "X" on half of
   the people in a room, and "Y" on the other half, then blindly chose a
   random sample of people in the room, you would expect to get about half "X"
   and half "Y" on average, regardless of how the initial selection had
   happened to have been made.

Appendix F  General Statistical Primer -------------------------- Page 62

   The same is true of the "random" grouping of men who received a vaccine,
   versus those who didn't.  It should be simply irrelevant as a factor when
   drawing samples of men "chosen" at random to become HIV infected, as if
   you had merely painted "X" or "Y" on them.

   If you have a factor that does favor an outcome, such as having a coin
   weighted on one side, then it is no longer a 50-50 chance.  You have to be
   very careful not to have such hidden bias.

   You also have the opportunity, if you see someone flipping a coin and
   getting 30 heads in a row, to know that something is fishy.  It is
   probably a trick coin, because the probability would be only about 1 in a
   billion, by random chance.   It is similar with our vaccines- something is
   clearly fishy.  What exactly it might be, we have to investigate, but
   there is something that needs investigation.

   The problem of drawing marbles from a jar is a bit different than flipping
   a coin.  If you keep flipping a coin, the odds of getting "heads" is the
   same on every flip.   If we pull a second marble from our jar, we are sure
   to get the white, because it is the only marble left.   Our program should
   show this: if we draw two marbles, we should have 100% percent chance of
   getting a white marble:

   * Chance of drawing a white marble, when pulling two marbles from a jar
     containing 1 white marble and 1 black marble:

   Subgroup size = 1
   Total group size = 2
   Sample size = 2
   n = 1

   This is a case so obvious as to be nearly silly, but it demonstrates how
   marble-drawing problems (hypergeometric distributions) sometimes differ
   significantly from coin-flipping problems (binomial distributions).    It
   also helps to confirm that our probability-computing program is working
   correctly.   The program comes in handy, because the computations become
   extremely lengthy, when the mixtures of marbles and the tested conditions
   become more complicated.

   What would be the odds of finding 10 or more white marbles, when drawing
   20 marbles from a jar holding 100 white marbles and 100 black marbles? 
   This is the most-expected result.

   * Chance of 10 or more white marbles, drawn in a sample of 20 marbles,
     randomly pulled from a jar of 100 white marbles and 100 black marbles:

   Subgroup size = 100
   Total group size = 200
   Sample size = 20
   n = 10
   PROBABILITY IS: 0.592851

Appendix F  General Statistical Primer -------------------------- Page 63

   That is, there is about a .59, or a 59% chance of getting 10 or more white marbles.

   Since this is the most expected result, we should see a slightly smaller
   probability for drawing 11 or more white marbles, and even less for
   drawing 12 or more white marbles.   The computer program output below
   shows that this is true:

   * Chance of 11 or more white marbles, drawn in a sample of 20 marbles,
     randomly pulled from a jar of 100 white marbles and 100 black marbles:

   Subgroup size = 100
   Total group size = 200
   Sample size = 20
   n = 11
   PROBABILITY IS: 0.407149

   * Chance of 12 or more white marbles, drawn in a sample of 20 marbles,
     randomly pulled from a jar of 100 white marbles and 100 black marbles:

   Subgroup size = 100
   Total group size = 200
   Sample size = 20
   n = 12
   PROBABILITY IS: 0.240184

   We had a 59% chance of drawing 10 or more white marbles.  To draw 11 or
   more white marbles, the chances fall to 40%.   To draw 12 or more white
   marbles, the chances fall to 24%.

   What are the chances that the entire group of 20 marbles will be 100% all-

   * Chance of finding 20 all-white marbles, drawn in a sample of 20 marbles,
     randomly pulled from a jar of 100 white marbles and 100 black marbles:

   Subgroup size = 100
   Total group size = 200
   Sample size = 20
   n = 20
   PROBABILITY IS: 3.32169e-07

   This works out to about 1 chance in 3 million- an extremely unlikely
   outcome.  Even though we look at a small number of marbles- only 20 - the
   improbability becomes enormous.  This shows why these types of problems
   are sometimes not completely obvious, at first glance.

Appendix F  General Statistical Primer -------------------------- Page 64

   This is not to say that the outcome cannot ever happen.   It does indeed
   happen.  This would be almost exactly the rate at which you would expect
   to find it happening, if you repeated the experiment an infinite number of
   times- roughly one time out every 3 million attempts.

   What if you are trying to analyze a situation where there are multiple
   possible explanations for getting the result that you did, besides simply
   one of random chance?   Suppose, for example, that the experimenter forgot
   the importance of evenly mixing the marbles in the jar.  Instead, 100 black
   marbles were poured on the bottom, and 100 white marbles poured on top.   
   This might guarantee a near 100% chance that you could pull a handful of
   all-white marbles, instead of the rightful probability of near-zero.

   Most of the preceding problems have involved an equal number of marbles of
   one color an another.  You can also have imbalances in the numbers of
   marbles of each type in the jar.  If you have 9999 white marbles, and only
   1 black, then draw one marble, the chance of getting the black marble is
   very small (1 in 10000).

   Our program is geared for the general case: X objects of one type, Y
   objects of another type, drawing a sample of Z objects, then computing the
   probability of getting "n" or more of the "X" type in the sample of Z.  
   You fill in the values of X, Y, Z, and n, for any problem of this sort.

   Suppose that this situation were one where you were trying to evaluate the
   safety of a food additive, to make sure that it was not a carcinogen
   (cancer-causing).  Say, that you have studied a large population of X
   animals, and that in a given year, you normally see Y animals that get
   cancer (maybe this would be a flat percentage, such as 1% of X).     Now,
   you study a smaller group of Z animals, using the food additive.  You find
   that out of these, you have "n" that get cancer in a year.   You would
   expect for "n" to be roughly equal to Z * (Y/X).   You expect it to be
   slightly different than this, because of random chance variation, but you
   do not expect it to be greatly different.   You can compute the
   probability for the difference that you see, just as you can for a problem
   in drawing marbles from a jar.

   Say that the odds against a higher cancer rate being attributable to
   random chance is only 1 in a million.  Do you approve the food additive
   (or in our case, the vaccine)?   Of course not.   There is indeed still a
   chance that the higher rate is only random, but you would not take that
   kind of chance.

   What constitutes "statistical significance" is not something that has a
   completely hard-and-fast rule.    A common convention used in many
   evaluations is a level of "1 in 1000".   Other, less demanding problems
   might test for significance at, say, the 1%  level (1 in 100), or the   5%
   level (1 in 20).

Appendix F  General Statistical Primer -------------------------- Page 65

   The desired level of significance is often chosen based on the needs of
   the problem under consideration.    Since our problem is a critical one of
   vaccine safety, we are more than meeting a reasonable definition of
   "statistical significance".

   If you draw a graph of probabilities for different outcomes, for these
   types of problems, you tend to get a "bell-shaped" curve.    The center of
   the curve is the most expected result (such as 50% white marbles drawn in
   a sample from a jar with a 50-50 white and black mix).  The edges of the
   curve as the least likely results (such as all-white marbles).

   As your population size and sample size grow (i.e., more marbles in the
   jar and bigger handfuls taken out), the odds against getting any one
   particular outcome (such as getting exactly 1,203 white marbles and 2,459
   black marbles in a giant handful) becomes extremely small.   This type of
   "improbable" result is nothing unusual or suspicious, because there are
   many, many such uninteresting, particular mixes that are equally possible.

   Our program is not computing odds for single outcomes of this sort.    It
   is computing a whole range of combined possible outcomes ("n or more"). 
   This is in effect an "area under the curve" for the bell-shaped
   probability curve.   That is why the results are meaningful, regardless of
   the population and sample sizes, or the shape of the curve.

   As mentioned earlier, the extremely small probability values computed in
   our analysis point to the fact that our sample sizes are large enough to
   be significant.  Smaller populations and samples tend in general to yield
   higher probability values.

   For example, the odds of getting 5 heads out of 6 coin tosses is about 10%,
   which is perfectly feasible.  The odds of getting 500 heads out of 600
   coins tosses is about 3 times (10 to the -65th power).  A trillionth would
   be 10 to the -12th power, so this is an unimaginably small possibility.

   The ratio of "head" outcomes to the total coin-toss trials is the same in
   both cases, at 5:6.

   Yet the probabilities are vastly different.   Six coins tosses are clearly
   not enough to guarantee that the outcome of 5 "heads" was not by random
   chance.  Six hundred coins tosses are far more than enough.

   The same phenomena apply to the 799 men in the SFMS and the 359 in the

   The computed probability is the real probability, and is fully meaningful,
   so long as the data used to make the calculation are reliable.

   If the formula or program seem complicated, that is the only real illusion.  
   It is enough to realize that this is a common type of problem, found in
   nearly any statistics text.   It would be pointless to attempt to create a
   mirage, in a document that will subject to the criticism of experts.