HIV as Surrogate Marker for Drug Use: A Re-Analysis of the San Francisco Men's Health Study

Bryan J. Ellison¹, Allen B. Downey ² & Peter H. Duesberg ¹
1: Dept. of Molecular and Cell Biology, Stanley Hall, University of California at Berkeley, Berkeley, CA 94720, USA
2: EECS Computer Science Division, 387 Soda Hall, University of California, Berkeley, CA 94720, USA

Original Publication
Genetica 95: 165-171, 1995
https://pubmed.ncbi.nlm.nih.gov/7744259/

Abstract

Our analysis of drug use and morbidity data from a cohort of 1034 men yields the following results: 1) HIV infection  is a strong indicator of drug use - HIV-positive respondents reported an average lifetime dose of recreational drugs  (excluding marijuana) 2.3 times higher than HIV-negative respondents. 2) Homosexuality is a strong indicator  of drug use - homosexual respondents reported an average lifetime dose 2.0 times higher than heterosexual  respondents. 3) The incidence of AIDS-defining diseases was not limited to respondents infected with HIV, but  was almost completely limited (98%) to respondents who reported using drugs.-We also address a previous report  (Ascher et al., 1993) that was based on the same database and purported to show that HIV alone correlates with  the development of AIDS. Specifically, we show that the relationship between HIV infection and CD4+ T Cell loss  is weaker than reported by Ascher et al., and provides little evidence for a causative relationship. These results  support the hypothesis that long-term, habitual drug use can cause the conditions known as AIDS (independent of  the presence of HIV), and refute the hypothesis that HIV alone causes these conditions independent of drug use.

Introduction  

A Nature Commentary by Ascher et al. purported to show that infection by the human immunodeficiency virus (HIV)alone correlates with the development of the acquired immune deficiency syndrome (AIDS)  (Ascher et al., 1993). According to their analysis of the San Francisco Men's Health Study (SFMHS), drug  use had no effect on T cell loss over time, while all cases of AIDS and all T cell depletion occurred among  HIV-positive men. Ascher et al. interpret this apparent correlation between HIV and AIDS as support for 'the well-established causal relationship between HIV and  AIDS' (Ascher et al., 1993).  

The power of the statistical tests Ascher et al. use is often insufficient to justify the conclusions they draw.  This paper will discuss the deficiency of these tests. In some cases, however, the absence of a detailed description of the methods they used made it impossible for us to verify their results.  

Ascher et al. also selectively cited existing lit erature on the debate over AIDS etiology, ignoring a wealth of evidence arguing against the HIV AIDS hypothesis and for a drug-AIDS hypothesis. The molecular and epidemiological inconsistencies of the  HIV hypothesis include the inability of the virus to  induce AIDS in chimpanzees and the existence of HIV free AIDS cases (Duesberg, 1992; Duesberg, 1993a).  There is ample evidence that alkyl nitrites, recreation al drugs primarily used by male homosexuals at risk for AIDS, and related compounds produce irreversible immune suppression and pneumonia in mice after long-term exposure, and that the nitrites follow a dose dependent relationship with the incidence of AIDS in humans (Holt et al., 1979; Haverkos, 1988; Newell,  Spitz & Wilson, 1988; Ortiz & Rivera, 1988). Cocaine and heroin, moreover, have long been associated with severe immune deficiencies in humans, independently of HIV infection (Duesberg, 1992).  

In addition to reviewing the data cited by Ascher et al., we were also able to perform our own analysis of the primary data. This paper presents some of our findings. Despite the extremely sparse reporting of AIDS  indicator diseases in the database, we found 45 cases of these diseases among HIV negative respondents. Also,  we quantified a higher level of drug use among HIV positive men than among their HIV negative counterparts. These results suggest a direct link between drug use and AIDS independent of HIV infection.  

The SFMH Study

The SFMHS is a population-based longitudinal survey of 1034 men, recruited in 1984 from the San Francis- co precincts most heavily populated with homosexual men. (For reasons unknown to us, Ascher et al. used only 1027 of these men in their analysis). Of these subjects, 816 identified themselves as homosexual or bisexual and 215 as heterosexual. At the end of the study's seventh year, 578 men remained HIV nega- tive, 46 had seroconverted, 400 had entered the study already seropositive, and 10 remained undetermined; 215 men had been officially recorded with diagnoses of AIDS. Data are collected by semi-annual interviews, which tabulate such self-reported data as medical con- ditions, use of pharmaceutical or recreational drugs, and a number of other responses that are difficult to verify.

The SFMH Study, however, was not designed to test for the cause of AIDS. HIV was presumed from the beginning to be the etiological agent, so that no attempt was made to account for the inherent prob- lems of self-reported answers, nor for the preceding years of unrecorded drug use (especially among HIV- positives). Inadequate records were also kept of spe- cific AIDS diseases and of drug use patterns occurring during the course of the study.

We shall document these and other defects of the SFMHStudy as well as of the Commentary by Asch- er et al. based on that study. Our analysis will show that Ascher et al.'s claim that 'the population-based SFMHS provides a rigorously controlled epidemiolog- ical model for the evaluation of aetiological hypothe- ses' is unjustified (Ascher et al., 1993).

The Ascher et aL Commentary

The Commentary reported a perfect association between HIV and AIDS, as well as between HIV and a progressive decline in CD4+ T cells. It also found no relationship between drug use and T cell loss. But the analysis suffers from several logical flaws:

Table1. HIV-negativeAIDSdiseases(578totalHIV-negative men).

1) The HIV-based clinical definition of AIDS is circular, because it fails to report AIDS-defining con- ditions in the absence of HIV infection. Conventional diseases such as pneumonia or tuberculosis are only reclassified as AIDS in the presence of antibodies against HIV, or on the presumption of infection (Insti- tute of Medicine and National Academy of Sciences, 1986; Centers for Disease Control, 1987; Centers for Disease Control and Prevention, 1992). Ascher et al. imply they have taken all such diseases into account. The Commentary notes "Duesberg has also argued that the case definition of AIDS is 'circular' and that HIV- seronegative AIDS cases are excluded by definition. In the SFMHS, AIDS cases were diagnosed by the clini- cal criteria defined by the Centers for Disease Control (CDC), which are irrespective of HIV infection sta- tus." (Ascher et al., 1993). However, our review of the SFMHS database indicates that for the first five years of the study, respondents were only asked whether they had been diagnosed with AIDS or Kaposi's Sarcoma, not whether they had been diagnosed with any of the AIDS indicator diseases. In the latter part of the study, the surveys included questions about some AIDS indi- cator diseases, but in total more than two-thirds of the data about these diseases is missing.

2) HIV infection and drug use have been shown to correlate with one another. Risk behaviors for virus transmission, including extreme sexual activity and sharing of injection equipment, generally involve use of such recreational drugs as heroin, cocaine, alkyl nitrites, and others (Jaffe et al., 1983; Newell, Spitz & Wilson, 1988; Schechter et al., 1993). For this reason any epidemiological study of HIV infection must care-fully control for drug use. The authors compare drug use among the homosexual or bisexual respondents and the heterosexual respondents, using sexual preference as a proxy for serostatus. They thereby imply that there is no difference in drug use between HIV positive and HIV negative respondents. According to Ascher et al., "Except for amyl nitrite.., the percentage of subjects reporting heavy use of each drug was similar in both sexual preference groups." (Ascher et al., 1993). While heterosexuality is an effective proxy for HIV-negative serostatus (all but one of the heterosexual respondents was HIV-negative), homosexuality is not a reasonable proxy for HIV-positive serostatus (almost half of the 816 homosexual respondents were HIV-negative). Our analysis (below) shows that there is a significant cor- relation between drug use and serostatus.

3) The hypothesis of drug use as the cause of AIDS is analogous to the cigarette-smoking hypothesis of lung cancer and the alcohol hypothesis of liver cirrho- sis (Lauritsen & Wilson, 1986; Duesberg, 1992). That is, disease symptoms depend primarily on the lifetime dose of the relevant drug, a function of both quantity and number of years of consumption. Thus a proper longitudinal study would quantify each respondent's total lifetime drug use. Ascher et al. imply that they have done this: "We examined the cohort at 6-monthly [sic] intervals for 96 months, and obtained drug-use data and determined HIV serostatus at each examina- tion" (Ascher et al., 1993). But in the Commentary, all study subjects were stratified according to a single drug use report, taken during the first wave of the study. Fur- thermore, although the respondents were asked about use of nine to ten different psychoactive drugs (see below), only four of those drugs were incorporated into the paper's analysis. Moreover, no mention was made of AZT, the toxic and immune-suppressive DNA nucleoside analogue prescribed as AIDS therapy.

Our analysis of the data and our results

Statistical tests are based on Cochran and Cox t-tests (with unequal variances), or on univariate linear regression.

AIDS without HIV

Clinically, AIDS is defined entirely in terms of about 30 conventional diseases presumed to have been comparatively rare in 20- to 55-year-old men before 1980. If a patient with one or more of these indicator diseases tests positive for HIV antibodies, he is usually diagnosed with AIDS. The AIDS case definition has been expanded several times by the CDC, each time adding new diseases to the diagnostic list. The 1987 CDC definition incorporated some two dozen diseases, and was the universal standard for diagnosis until January 1, 1993, when five new conditions were added (Centers for Disease Control, 1987; Centers for Disease Control and Prevention, 1992). As of 1992, 39% of the AIDS-defining conditions reported in the United States were not diseases of immunodeficiency, including Kaposi's sarcoma, several lymphomas, dementia, and wasting disease (Centers for Disease Control, 1993).

Despite the absence of most of the relevant disease data, we found 45 HIV-negative men with AIDS- defining conditions (according to the CDC), as listed in Table 1. Of those, 34 qualified under the older (1987) definition; 11 qualified only under the 1993 definition, having either <200 CD4+ T cells or thrombocytopenia. Fifteen of the 45 were reported in the first round of interviews in 1984. Many HIV-negative men were followed up only inconsistently thereafter or were lost altogether, and were therefore inadequately monitored for AIDS diseases. Given the missing data on many of the study subjects and on other potential AIDS-defining conditions, this must be taken as a minimum estimate.

At least eight of these cases were published a few years earlier by Winkelstein and other researchers with the SFMHS (Lang et al., 1987). The earli- er study also included as many as 133 other cases of AIDS- or early-AIDS-related symptoms, such as lymphadenopathies, persistent rashes, wasting conditions, or histories of worsening herpes infections. This earlier paper, and our finding of HIV-free AIDS indicator diseases, undermine the claim of Asch- er et al. that 'No cases of AIDS... have been recorded among these [seronegative] men' (Ascher et al.,1993).

The relatively high incidence of opportunistic dis-eases listed in Table 1, found in men recruited between the ages of 25 and 54, is particularly revealing, for such conditions would normally not be expected to occur among the healthiest age group in the population. The presence of these diseases indicates some non-HIV factor at work. It is important to emphasize that had any of these 45 men been positive for antibodies against HIV, they would likely have been recorded as AIDS cases.

Table2. Average age, AIDS vs. non-AIDS.

An analysis of the 'no drug use' subgroup

A second question arose over the figure displayed in the Commentary. One of the six longitudinal T cell curves purported to describe HIV-positive men reporting 'no drug use'. The authors' criteria for this classification were based on self-reported use of four drugs - marijuana, cocaine, nitrites, and amphetamines - during the two years prior to the beginning of the study. We examined the database to find the subjects who met these criteria, and inspected these men's recorded answers to check whether they remained a truly drug free control group, looking at all other drugs and at AZT.

That 'no drug use' curve was apparently based on twenty seropositive men in the database who reported no use of the four drugs upon entry. Of these men, one reported using a different drug (downers) immedi- ately prior to entry, five others admitted using various drugs (marijuana, nitrites, cocaine, and hallucinogens) at later times during the study; thirteen reported taking AZT. After accounting for overlaps, this left only three of the original twenty who claimed to be completely drug-free, at least during the study. One of these men showed a sharp decline in CD4+ T cell counts during the first two years of the study, was diagnosed with AIDS, developed a Pneumocystis carinii pneumonia, and died months later. The remaining two individuals appear not to have declining T cell counts, and neither was diagnosed with AIDS or died.

Even if the "no drug use" group were well-defined, it would still be inappropriate to present a simple aver- age of T-cell counts, especially, as in the Commentary, without providing error bars to indicate the variation within the group. First, an individual's T-cell count varies widely over time, making time-series data of this kind dubious. Secondly, not all subjects participat- ed in all waves; thus the "group" represented at each wave is made up of different respondents. Finally, the use of unnormalized T-cell counts weights individuals with naturally high T-cell count more than individuals with naturally lower counts. The unreliability of T-cell counts as a measure of the health of the immune sys- tem is well-known; averaging this data over ill-defined and varying groups does not help - it only obscures whatever conclusions might be drawn from individu- al trends. Figure 1 shows the T-cell counts for each of the 20 "drug-free" subjects over time. Any decline in the average T-cell count in this group is primarily due to a single individual with an unusually high and sharply-declining T-cell count. Generalizations about the health of the members of this group on the basis of this chart would be entirely unjustified.

Drug use as a correlate of HIV infection and AIDS

The SFMHS database suffers several shortcomings that limit its ability to test the drug- and HIV- hypotheses of AIDS. Because many of the drug-use questions were not answered by many of the subjects at various times throughout the study and all drug data were self-reported and unverified, it was difficult to construct a reliable measure of lifetime drug use. Nevertheless, we were able to convert data about frequency of drug use into an estimate of total doses, and to integrate this data over time into an estimate of total lifetime dose. For example, if an individual reporting annual use of a drug was assigned one dose unit, another individual reporting weekly use would be assigned 52 dose units.

Data were available from fourteen semi-annual surveys; we added a fifteenth data point representing the eighteen months preceding the beginning of the study. Drug use was condensed into two indices for each individual - the total combined drug use reported over the 8.5 years (marijuana, nitrites, cocaine, 'crack' cocaine, amphetamines, hallucinogens, uppers, downers, ethyl chloride, heroin, and 'other'), and a similar index without marijuana. The category 'crack cocaine' was added in wave 10, at which point the category 'other' was no longer used in our analysis, so that only ten drug categories were examined for any given wave, and AZT was not included in the analysis. Thus we used data from up to 150 drug responses gathered from each individual (15 waves times 10 drugs each).

Fig. 1. IndividualT-Cellcountsof20HIV-positive respondents reporting no drug use during the study. In the Commentary, Ascher et al. average these curves and present the resulting smooth line without error bars(Ascheret al., 1993). They attribute the downward trend in that line to HIV-infection, rather than to the sharp decline of a few individuals.

Open

Table 3 compares the mean lifetime dose of drugs among HIV-positive and negative respondents and among homosexual and heterosexual respondents. It confirms the role of HIV as a surrogate marker for drug use. HIV-positive men on average used 128% more drugs (ignoring marijuana) than did the HIV- negative subjects (the difference is statistically significant). Moreover, the drug use data from later in the study contradict the claim of Ascher et al., based only on drug data from the first survey, that drug use is similar among homosexuals and heterosexuals. Table 3 shows that homosexuals used twice as many drugs as the heterosexuals (again ignoring marijuana), a statistically significant difference. Use of nitrite inhalants was particularly heavy among homosexuals, and AZT (naturally) was used almost exclusively by HIV posi- tives respondents (Ascher et al., 1993).

Among HIV-positive subjects, we looked for a difference in total lifetime drug use between respondents who had been diagnosed with AIDS and those who had not. Contrary to our expectations, we did not find a statistically significant difference. The statistical power of  the test was limited by some of the problems we have  mentioned above, as well as the following:  

Conclusions

Table3. Average doses of drugs(other than marijuana) per year, HIV+ vs. HIV- and homosexuals vs. heterosexuals.

The Ascher et al. Commentary was a faulty analysis of the SFMHS database:

a) Because of the authors' dependence on the circular CDC definition of AIDS, they ignored the inci- dence of AIDS indicator diseases among the HIV- negative respondents. One consequence of their artificially perfect association between HIV and AIDS is that all other correlations between risk factors and AIDS are obscured.

b) By failing to quantify the enormous difference in drug use between HIV-positives and -negatives, they overlooked this potential explanation of the greater decline in CD4+ counts among HIV antibody-positive men.

c) By stratifying subjects according to a single drug use report, they included users of recreational drugs and AZT in their 'no drug use' group, and were able to ignore the problem of missing data later in study.

d) By failing to analyze the data with and without marijuana they allowed some drug associations to be obscured or minimized. Study subjects reported large amounts of marijuana use, which numerically tended to drown out reported use of more powerful drugs. The general willingness to admit marijuana use may reflect the lesser degree of social stigma attached to this particular drug. In any case, mari- juana is less cytotoxic than other drugs examined in the study (such as nitrites [Haverkos & Dougher- ty 1988]), and is likely to be insignificant in the etiology of AIDS.

e) T cell counts provide a questionable marker of AIDS progression, especially in the SFMHStudy. For one thing, several important AIDS-defining diseases do not result from immunodeficiency; Kaposi's sarcoma, for example, has been described in homosexual men without measurable immune deficiencies (Murray et al., 1988; Spornraft et al., 1988; Friedman-Kein et al., 1990). Furthermore, the individual T cell counts recorded in the database often fluctuated greatly with time, and the aver- age T cell curves (not reproduced here) showed major irregularities that Ascher et al. smoothed out with some undescribed adjustment (Fig. 1). Finally, many CD4+ counts were missing throughout the study, apparently for individuals who could not be contacted at various waves. The average T cell curves therefore reflect a constantly changing subset of men over time.

Furthermore, the SFMHS database may be inadequate to test drugs, HIV, or any other agent as a pro- posed cause of AIDS:

a) Of the 446 men who were seropositive after seven years, only 46 had seroconverted at known times during the study; the other 400 had been infected at unknown times preceding the study.

b) AIDS-defining diseases were poorly monitored. The SFMHStudy relied at times on self-reported answers rather than medical records and only inquired about a changing and severely incomplete list of those diseases. Many of the men were followed up inconsistently after the first six months, which may explain the observation that 15 of the 45 cases of AIDS indicator diseases among HIV- negative respondents were recorded during the first wave.

c) No data about drug use exist for most of the years preceding the study, so that it is impossible to quantify lifetime drug consumption. Without more reliable data on lifetime drug use, no controlled analysis of drug-AIDS associations is possible in this cohort.

Nevertheless, some conclusions can be drawn from the SFMHStudy. Perhaps most significant of these is our finding of 45 HIV-free cases of CDC-defined AIDS diseases. These can be added to the list of nearly 5000 cases of HIV-free AIDS-defining diseases and immunodeficiencies already uncovered in at least 75 references in the existing literature (compiled in a review paper) (Duesberg, 1993a). The seronegative AIDS conditions have been found primarily in the same risk groups as seropositive AIDS, including central Africans, male homosexuals, intravenous drug users, and hemophiliacs.

The association between drugs and AIDS-defining diseases in this database appears to be greater than the association between HIV and AIDS. AIDS conditions were reported among 260 men, 215 of whom were infected with HIV- a total of 82%. Of the same group, 93% admit using drugs other than marijuana, a total of 96% including marijuana, and a total of 98% including both marijuana and AZT.

The most compelling argument against the drug hypothesis of AIDS would be the existence of a group of drug-free, HIV-positive AIDS cases, but such could not be found in the SFMHS. On the other hand, the strongest argument against the virus-AIDS hypothesis would be a group of drug-using, HIV-negative AIDS cases, which may indeed exist in the SFMHS cohort (see above).

Further results might possibly still be extracted from the SFMHS database, including an analysis of the effects of AZT; of the 400 seropositive men in the study, 215 (54%) reported use of this DNA chain- terminating chemotherapy. But an epidemiological study of the causality of AIDS, controlling for both HIV and lifetime drug exposure, will require some other more adequate cohort. From the standpoint of biological plausibility, the toxicity of long-term drug use makes much more sense as a damaging agent to the immune system than do the 'enigmatic mechanisms' of the conventional retrovirus HIV (Ascher et al., 1993).

Acknowledgements

We are indebted to Warren Winkelstein, Jr., for providing access to the data from the SFMHS study, and to Eric Vittinghoff for technical assistance. We also thank Phillip Johnson, Richard Strohman, Malcolm Zaretsky, John Lauritsen, Charles A. Thomas, Steve Smale, and Harvey Bialy for critical reviews. Supported in part by the Council for Tobacco Research, USA, and private donations from Glenn Braswell (Los Angeles, Calif., USA), Dr. Richard Fischer (Annan- dale, Va., USA), Dr. Fabio Franchi (Trieste, Italy) and Dr. Friedrich Luft (Berlin, Germany).

References

• Ascher, M.S., H.W. Sheppard, J. Winkelstein, & E. Vittinghoff, 1993. Does drug use cause AIDS? Nature (London) 362:103-4.

• CentersforDiseaseControl, 1987.Revision of the CDC surveillance case definition for acquired immunodeficiency syndrome. J. Am. Med. Assoc. 258:1143-1154.

• Centers for Disease Control, 1993. HIV/AIDS Surveillance; year-end edition. February Issue.

• Centers for Disease Control and Prevention, 1992. 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. Morb. Mort. Weekly Rep. 41:1-19. Duesberg, P., 1992. AIDS acquired by drug consumption and other noncontagious risk fac- tors. Pharmacology & Therapeutics 55:201-277.

• Duesberg, P.H., 1992. AIDS Acquired by Drug Consumption and Other Noncontagious Risk Factors, Pharmac. Ther. Vol. 55 pp. 201-277.

• Duesberg, P., 1993a. The HIV gap in national AIDS statistics. Biotechnology 11:955-956.

• Duesberg, P.H., 1993b. Can epidemiology determine whether drugs or HIV cause AIDS? AIDS-Forschung 12:627-635.

• Friedman-Kien, A.E., B.R. Saltzman, Y. Cao, M.S. Nestor, M. Mirabile, J.J. Li & T.A. Peterman, 1990. Kaposi's sarcoma in HIV-negative homosexual men. Lancet 335:168-169.

• Haverkos, H.W., 1988. Epidemiologic studies - Kaposi's sarcoma vs. opportunistic infections among homosexual men with AIDS. In: Health Hazards of Nitrite Inhalants, Research Monograph 83, pp. 96-105, H.W. Haverkos & J.A. Dougherty (eds.) National Institute on Drug Abuse, Rockville, MD.

• Haverkos, H.W. & J.A. Dougherty (eds) (1988) Health Hazards of Nitrite Inhalants. NIDA Research Monograph 83, National Institute on Drug Abuse, Washington DC.

• Holt, P.G., L.M. Finlay-Jones, D. Keast & J.M. Papadimitrou, 1979. Immunological function in mice chronically exposed to nitrogen dioxide (NO2). Environmental Research 19: 154-162.

• Institute of Medicine and National Academy of Sciences, 1986. Confronting AIDS. National Academy Press, Washington, DC.

• Jaffe, H.W., K. Choi, P.A. Thomas, H.W. Haverkos, D.M. Auerbach, M.E. Guinan, M.E Rogers et aL, 1983. National case-control study of Kaposi's sarcoma and Pneumocystic carinii pneumonia in homosexual men: Part 1, epidemiologic results. Ann. Intern. Med. 99:145-151.

• Lang, W., R.E. Anderson, H. Perkins, R.M. Grant, D. Lyman, W. Winkelstein, R. Royce & J.A. Levy, 1987. Clinical, immunologic, and serologic findings in men at risk for Acquired Immunodeficiency Syndrome. J. Am. Med. Assoc. 257:326--330.

• Lauritsen, J. & H. Wilson, 1986. Death Rush, Poppers and AIDS. Pagan Press, New York.

• Murray, H.W., D.A. Scavuzzo, C.D. Kelly, B.Y. Rubin & R.B. Roberts, 1988. T4+ cell production of interferon gamma and the clinical spectrum of patients at risk for and with acquired immunodeficiency syndrome. Arch. Intern. Med. 148:1613- 1616. Newell, G.R., M.R. Spitz & M.B. Wilson, 1988. Nitrite inhalants: Historical perspective. In: Health Hazards of Nitrite Inhalants, Research Monograph 83, pp. 1-14, H.W. Haverkos and J.A. Dougherty (eds.) National Institute of Drug Abuse, Rockville, MD.

• Ortiz, J.S.&V.L.Rivera, 1988. Altered T-cell helper/suppressor ratio in mice chronically exposed to amyl nitrite. In: Health Hazards of Nitrite Inhalants, Research Monograph 83, pp. 59-74, H.W. Haverkos and J.A. Dougherty (eds.) National Institute on Drug Abuse, Rockville, MD.

• Schechter, M.T., KJ.P. Craib, K.A. Gelmon, J.S.G. Montaner, T.N. Le & M.V. O'Shaughnessy, 1993. H1V-1 and the aetiology of AIDS. The Lancet 341:658-659. Spornraft, P., M. Froschl, J. Ring, M. Meurer, F.D. Goebel, H.W. Ziegler-Heitbrock et al., 1988. T4/T8 ratio and absolute T4 cell numbers in different clini- cal stagesof Kaposi's sarcoma in AIDS. Br. J. Dermatol. 119:1-9.