Causes, Cases, And Cohorts:
The Role Of Epidemiology In The Historical Construction Of AIDS
Gerald M. Oppenheimer
In his history of the HIV (human immunodeficiency virus) epidemic, Mirko Grmek reports that the term acquired immune deficiency syndrome , the first generally accepted name for this new disorder, was coined at a 1982 meeting held at the Centers for Disease Control (CDC) in Atlanta. Thereafter, the CDC epidemiologists spread and legitimated the neologism by using it extensively in official publications.[1]
Mirko D. Grmek, Histoire du SIDA (Paris: Payot, 1989), pp. 58-61. The descriptive term acquired immunodeficiency syndrome, or AIDS, became synonymous with the new disorder; it has recently been replaced by human immunodeficiency virus (HIV) infection. Though, in general, this essay uses the new acronym, in discussing specific studies it will employ whatever term was used by the investigators reporting.
By attributing to the CDC the power to control the name of the disease, Grmek indirectly demonstrates how prominent a part that agency and its epidemiologists played in defining this new "medical mystery."In this essay I examine the role of epidemiologists, in the CDC and elsewhere, in characterizing HIV infection. Faced with a new disease of unknown origin, epidemiologists and their collaborators constructed, over time, hypothetical models to explain the disorder in order to contain it. Prior to the isolation of a causal virus, epidemiologists played a central role in defining the new syndrome, developing first a "life-style" model and later a model based on hepatitis B. Though subsequently supplanted from their special position by virologists and other "bench" scientists working in laboratories (who named the virus and thereby redesignated the disease), epidemiologists have continued to define important dimensions of the disorder and to raise disquieting questions. Specifically, they were concerned with discovering risk factors for HIV infection, its modes of transmission, the natural history of the disease, the extent to which it had spread within population groups, and the projection of future prevalence and incidence rates.
This is a substantially revised and updated version of the chapter entitled "In the Eye of the Storm: The Epidemiological Construction of AIDS," in AIDS: The Burdens of History , ed. Elizabeth Fee and Daniel M. Fox (Berkeley: University of California Press, 1988), pp. 267–300. This essay has benefited from the generous comments of Ronald Bayer, Ben Brody, Elizabeth Fee, Daniel M. Fox, Robert Padgug, and Anne Stone.
Although epidemiologists have increasingly lost to biomedical scientists the power to construct the meaning of the HIV infection, epidemiologists in the CDC retain an important prerogative: they continue to frame the population-based definition of AIDS. Because the CDC has responsibility for monitoring infection in the United States, it has formulated, over time, the surveillance definitions of AIDS. Recently this population-based definition, as well as the reporting system itself, has been found deficient by demographers and quantitative social scientists. Their critiques raise the possibility that the power of epidemiologists to frame the disorder, already limited by biomedical scientists, may be further eroded by social scientists newly attracted to the study of AIDS. Nonetheless, since 1981 epidemiology has had a profound effect on the characterization of HIV infection in the United States. To a large degree, that characterization reflects something of the nature and concerns of American epidemiology itself.
Epidemiology, unlike virology, has a strong social dimension in that it explicitly incorporates perceptions of a population's social relations, behavioral patterns, and experiences into its explanations of disease processes. Given their training, epidemiologists fairly consistently defined HIV infection as a biological process occurring within a determinant social matrix. That the infection was first identified among young male homosexuals and intravenous drug users certainly reinforced that professional proclivity.
The results of this exercise in epidemiological imagination were complex and equivocal. On the one hand, the epidemiologists' approach may have skewed the choice of models and hypotheses, determined which data were excluded from consideration until later in the epidemic, and offered scientific justification for popular prejudice, particularly against gay men. On the other hand, the epidemiological approach gave the new disease a human face. By defining the behaviors and the multiple social experiences of groups as risk factors for the disease, epidemiology countered attempts to reduce the etiology of HIV infection to a virus alone. In addition, epidemiology offered the possibility of primary prevention in the form of health education and follow-up, particularly important in the absence of a vaccine or a successful therapy.
The various characterizations of HIV infection examined in this essay will span the period from early 1981, when physicians first encountered anomalous medical facts, to mid-1990, when epidemiologists had attempted to define the distribution of the HIV across subpopulations; to project future rates of HIV infection and illness for the population as
a whole; and to establish wih some specificity the natural history of the new disease. This essay draws almost entirely on the medical literature of the period.
Epidemiology And Public Health
Epidemiology played a key role in the AIDS epidemic for at least two reasons—one institutional, the other scientific. The institutional link was the Centers for Disease Control (CDC) in Atlanta. Part of the Public Health Service, which falls under the jurisdiction of the U.S. Department of Health and Human Services, the CDC is responsible for monitoring morbidity and mortality trends in the United States and for responding to acute outbreaks of disease—infectious disease in particular. To fulfill its mission, the CDC depends heavily on case reports, surveillance, and epidemiological investigations.
Epidemiology is particularly well suited to explore, portray, and explain new medical phenomena. It seeks to measure and analyze the occurrence and distribution of diseases and other health-related conditions in human populations, acting both as a sentinel who warns of shifts in disease patterns and as a scout who seizes on such shifts to discover their etiology.
For example, by systematically collecting data on the frequency of disorders in populations or subgroups through surveillance programs, epidemiologists can discern changes in the distribution of diseases in the community. Observations of these distributions, and their variation in subgroups, lead to hypotheses concerning the relationship between the disease and variables that may affect its natural history and clinical course. Using different study designs, epidemiologists attempt to measure, reject, or refine the relative significance of such hypothetical associations. The ultimate objective of these studies is to isolate the causal variables of the disease in question. An intermediate goal is to discover a point in the natural history of the disease where intervention might alter its course, even if its etiology remains unknown.
Epidemiologists tend to believe in multifactorial disease models. They assume, that is, that intervention is possible at several points, even in the absence of a known "first cause." The major premise of the multifactorial model is, as the name implies, that a given disease may have a number of causes or antecedents, a combination of which may be needed to produce the disorder. The "web of causes," therefore, may be interdicted at more than one vulnerable point.[2]
Brian MacMahon and Thomas F. Pugh, Epidemiology (Boston: Little, Brown, 1970), p. 25. See also John M. Last, ed., A Dictionary of Epidemiology (New York: Oxford University Press, 1983), S.V. "multiple causation."
The power of the multifactorial model is that it can incorporate any measurable factor relevant to and statistically associated with the disease or disorder of interest. Unlike the reductionist paradigm of the germ theory, the multicausal model embraces a variety of environmental and social factors. The model's strength, however, is also its weakness. The multifactorial model allows the researcher to cast a very wide net. Scientists may attempt to incorporate many possible explanatory variables whose putative causal connections with the disease in question may be plausible for a number of reasons—scientific, logical, historical, experiential, and so forth. Variables may be drawn in (or left out) as a function of the social values of the scientists, the working group, or the society. When included in the model, embraced by the professionals, and published in the scientific press, such value judgments appear to be objective, well-grounded scientific statements.
Epidemiology is an applied science that responds to two kinds of disorder within the community: one caused by the disease directly, and the other the product of the fears it has aroused. Consequently, epidemiology bore the initial responsibility of outlining the direction of research, generating hypotheses, and synthesizing the results. In the face of a fatal disorder of unknown origin and indefinite proportions, epidemiology offered a set of procedures (for example, case definition, verification, and count) that swiftly generated results and then authenticated them, giving the public a sense of definite progress. The content of this science, by providing and naming concepts (for example, "risk groups"), made the epidemic potentially less frightening by making it appear more likely that the disease would eventually be understood and controlled.
Case Finding And Surveillance
The initial discoveries heralding a new disorder of unknown origin were made by physicians treating patients in Los Angeles hospitals. Michael Gottlieb and his colleagues alerted the CDC that between October 1980 and May 1981 five young, previously healthy homosexual men had been treated for biopsy-confirmed Pneumocystis carinii pneumonia (PCP). PCP is a protozoan-produced condition that occurs almost exclusively in persons with severely suppressed or defective immune systems. On June 5, 1981, a short paper describing the patients was published by the CDC in its Morbidity and Mortality Weekly Report (MMWR) .[3]
U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control, Reports on AIDS Published in the Morbidity and Mortality Weekly Report, June 1981 through February 1986 (Springfield, Va.: National Technical Information Service, 1986), pp. 1-2 (hereafter cited as MMWR).
Gottlieb's communication to the CDC was closely followed by another from New York City and San Francisco, which reported that, in the thirty months prior to July 1981, Kaposi's sarcoma (KS) had been diagnosed in twenty-six male homosexuals between twenty-six and fifty-one years of age.[4]
MMWR, pp. 2-4.
A rare cancer in the United States, KS occurred in this country primarily in elderly males and immunosuppressed transplant recipients. Its manifestation in a relatively large number of young men was considered highly unusual.An editorial note in the issue of MMWR that had published Gottlieb's paper hypothesized that "the fact that these patients were all homosexuals suggests an association between some aspect of a homosexual lifestyle or disease acquired through sexual contact and Pneumocystis pneumonia in this population."[5]
Ibid., p. 2.
The conjecture that some aspect of homosexuality predisposed the patients to immune dysfunction and infections was made on the basis of only five cases from a single community, a broad generalization to formulate from so small a sample.The basis for that sweeping hypothesis lay in a rough mixture of analysis and opinion. The CDC had just completed a cooperative study with a number of gay community health clinics. It was a multiyear, multisite study of risk factors for hepatitis B, a disease which can be sexually transmitted and whose prevalence is very high among homosexual men.[6]
David G. Ostrow, "Homosexuality and Sexually Transmitted Diseases," in Sexually Transmitted Diseases, ed. Yehudi M. Felman (New York: Churchill Livingston, 1986), p. 210. See also M. T. Schreeder et al., "Hepatitis B in Homosexual Men: Prevalence of Infection and Factors Related to Transmission," Journal of Infectious Diseases 146 (1982): 7-15.
In analyzing the interrelation of life-style and hepatitis B, the researchers found that blood markers for the disease were significantly associated with, among other factors, a large number of male sexual partners and with sexual practices that involved anal contact.The CDC-associated study took place against a background of other investigations that pointed to an increase in the incidence as well as the types of sexually transmitted diseases (STDs) in homosexual men.[7]
William W. Darrow, "Sexual Behavior in America," in Sexually Transmitted Diseases, ed. Felman, pp. 269-71.
Analysts linked this epidemic of STDs to gay liberation and the attendant life-style of bars, discos, and bathhouses and of anonymous sexual partners.[8]Terry Alan Sandholzer, "Factors Affecting the Incidence and Management of Sexually Transmitted Diseases in Homosexual Men, in Sexually Transmitted Diseases in Homosexual Men, ed. David G. Ostrow, Terry Alan Sandholzer, and Yehudi M. Felman (New York: Plenum Medical Book Company, 1983), p. 5.
The combination of the CDC's recent work on risk factors for hepatitis B transmission, which had increased its awareness of gay life-style and sexuality and its knowledge of the epidemicity of STDs among subgroups within the gay community, probably accounts in part for the hypothesis suggested in the MMWR . One might fairly infer that the CDC was prematurely ready to find the etiology of this mysterious disorder in an exotic subculture. This inference is strengthened by the ensuing scientific work of epidemiologists within and outside the CDC,
who found in gay culture—particularly in its perceived "extreme" and "nonnormative" aspects (that is, "promiscuity" and recreational drugs)—the crucial clue to the cause of the new syndrome.
Part of the reason for the CDC's speedy adoption of the "life-style" hypothesis was, most likely, that in certain previous outbreaks of diseases of uncertain origin (in particular, legionnaires' disease in 1976), CDC officials had been criticized for committing themselves to a microbial hypothesis without having paid sufficient attention to alternative causative theories.[9]
See U.S. Congress, House of Representatives, Subcommittee on Consumer Protection and Finance, Committee on Interstate and Foreign Commerce, Hearings on Legionnaires' Disease, November 23-24, 1976, 94th Cong. For a defense of the CDC, see Barbara J. Culliton, "Legion Fever: Postmortem on an Investigation That Failed," Science 194 (1976): 1025-27.
Such criticism probably influenced their desire to throw a causative net widely in the case of HIV infection.[10]Stephen Schultz, M.D., former deputy commissioner, New York City Department of Health, personal communication, July 22, 1987.
A special task force on KS and opportunistic infections was established at the CDC in mid-1981 and charged with the surveillance of all new cases. As a preliminary step, the CDC had to define what constituted a case. It initially described a case as "a person who 1) has either biopsy-proven KS or biopsy-proven, life-threatening opportunistic infection, 2) is under age 60, and 3) has no history of either immunosuppressive underlying illness or immunosuppressive therapy."[11]
MMWR, p. 9.
By September 1982, when the CDC first used the term AIDS in the MMWR , it refined this description to define an AIDS case as one with "a disease at least moderately predictive of a defect in cell-mediated immunity, occurring in a person with no known cause for diminished resistance to that disease." Included among the diseases were KS, PCP, and a specific list of "other opportunistic infections," a list which the CDC has amended over the years.[12]Ibid., pp. 18, 95-97.
On September 1, 1987, the CDC significantly modified its case definition. It not only included new medical conditions such as HIV-related encephalopathy (dementia) and wasting syndrome but, for the first time, counted as cases those who, along with a positive antibody test, have had only a presumptive (that is, non-laboratory-confirmed) diagnosis for certain diseases, such as PCP and KS. Preliminary evidence indicates that 12 percent of cases diagnosed during the four months after September 1, 1987, met only the new case definition.[13]
See John M. Karon, Timothy J. Dondero, Jr., and James W. Curran, "The Projected Incidence of AIDS and Estimated Prevalence of HIV Infection in the United States," Journal of Acquired Immune Deficiency Syndromes 1 (1988): 542-50.
What caused this disorder? With limited clinical data at hand, the CDC did a "quick and dirty" survey of 420 males attending venereal disease clinics in San Francisco, New York, and Atlanta, with the intention of finding cases with KS or PCP. The thirty-five cases culled from the sample (biased, since such patients are more active sexually than the general population) were interviewed on many subjects in the hope that a lead might be discovered.
The researchers found two patterns of behavior that "fell out": sex
and drugs. The cases, all homosexuals, had had many sexual partners in the past year (the median was eighty-seven) and had frequently used marijuana, cocaine, and amyl or butyl nitrite—inhalant sexual stimulants.[14]
Centers for Disease Control, Task Force on Kaposi's Sarcoma and Opportunistic Infections, "Epidemiologic Aspects of the Current Outbreak of Kaposi's Sarcoma and Opportunistic Infections," New England Journal of Medicine 306 (1982): 248 (hereafter cited as Task Force Report); see also Gerald Astor, The Disease Detectives (New York: New American Library, 1983), p. 56.
Were sex and drugs independent of each other, however? The rate of nitrite use, for example, was closely associated with the number of sexual partners, suggesting that nitrite inhalation might be associated with other hypothetical causal variables, including sexually transmitted diseases, the medications used to treat them, or types of sexual behavior.[15]Task Force Report, p. 252.
It was also possible that nitrite use was not an etiological factor, but appeared to be one because it was associated with a casual or "confounding" variable such as sex.Despite the dearth of evidence (the "quick and dirty" survey had found that 86.4 percent of homosexual or bisexual men, whether cases or not, had used nitrite in the previous five years), amyl nitrite (AN) did become one of the first hypothetical causal variables to be investigated.[16]
MMWR, pp. 4-5.
As a clue, amyl nitrite seemed worth pursuing. It appeared to be compatible with the gay life-style thesis posed by the MMWR and attractive to epidemiological researchers. Studies in which nitrite inhalant was a variable will be evaluated below.Scientific papers published in 1981 consisted mainly of case and surveillance reports, in which attempts were made to define the new syndromes and the patients—that is, to formulate what constituted a "case." By describing the population at risk in terms of person, place, and time, and by learning from physicians the clinical details of the disorder, epidemiologists could grope for etiological clues that they might use to design formal studies.
One of the first clinical clues pursued was the possibility that the new syndrome was caused by the cytomegalovirus (CMV), a microbe suspected of being both sexually transmitted and a cause of KS. In September the British medical journal The Lancet published a clinical study of KS in eight homosexual men in New York City; the investigation found that all four patients tested were positive for CMV.[17]
Kenneth B. Hymes et al., "Kaposi's Sarcoma in Homosexual Men—A Report on Eight Cases," Lancet 2 (1981): 598-600.
Three months later Michael Gottlieb and his colleagues reported that four previously healthy men with PCP were infected with CMV and also were suffering from a marked decrease in white blood cells, particularly of a kind known as T4 helper cells.[18]Michael S. Gottlieb et al., "Pneumocystis Carinii Pneumonia and Mucosal Candidiasis in Previously Healthy Homosexual Men," New England Journal of Medicine 305 (1981): 1430.
While acknowledging that CMV infection might result from T4-cell deficiency and the reactivation of a dormant infection, Gottlieb and his colleagues, basing their position on previous studies, preferred to hold CMV highly suspect.The CDC, in its year-end summary on the epidemic, also cited CMV
as one of three possible etiological agents.[19]
Task Force Report, pp. 251-52.
Other putative causes, perhaps more closely related to the life-style hypothesis, were amyl nitrite and opiate addiction. (A recent investigation in New York City of eleven immunocompromised men with PCP had found that seven of the patients, including five heterosexuals, were drug "abusers."[20]Henry Masur et al., "An Outbreak of Community-Acquired Pneumocystis Carinii Pneumonia," New England Journal of Medicine 305 (1981): 1431-38.
) Did any of these agents bear a relationship to any other? How did CMV fit into the life-style hypothesis? An editorial in the New England Journal of Medicine addressed these issues in December 1981.Ignoring the heterosexual cases of PCP and other opportunistic infections, the editorialist noted that "the question of cause is obviously central. What clue does the link with homosexuality provide?"[21]
David T. Durack, "Opportunistic Infections and Kaposi's Sarcoma in Homosexual Men," New England Journal of Medicine 305 (1981): 1466.
The answer was a high incidence of sexually transmitted diseases, including viral infections such as CMV and hepatitis B, that might cause immunosuppression and KS. But because neither homosexuality nor CMV is new, the author suggested that a new factor may have modified the host-agent relationship: recreational drugs, particularly amyl nitrite. On the basis of this reasoning, he postulated a possible multifactorial disease model.[22]A model can be defined as "a description, a collection of statistical data, or an analogy used to help visualize often in a simplified way something that cannot be directly observed"; see Webster's Third New International Dictionary of the English language Unabridged (1986), S.V. "model." According to Susser, a model is a system reduced to a set of related variables for the purpose of prediction or representation (Mervyn Susser, Causal Thinking in the Health Sciences [New York: Oxford University Press, 1973], p. 32). In the present essay the models discussed perform a representational function in that they "represent existing or postulated relationships in simplified form" (ibid., p. 33).
Specifically, he proposed that the joint effects of persistent, sexually transmitted viral infection (presumably from CMV) and a recreational drug such as amyl nitrite precipitated immunosuppression in genetically predisposed males. From this followed a clinical course that included minor illnesses, then KS or other neoplasms, and serious opportunisitic infections. In essence, the model was an elaboration of the hypothesis originally proposed in the editorial note appended to the first MMWR on the new disease.The Life-Style Hypothesis: Experimental Work
To refine hypotheses generated by case reports, "quick and dirty" surveys, and surveillance, researchers compared patients with the new syndrome to a group of healthy men possessing comparable sociodemographic characteristics, experiences, or behaviors. Such research designs, which begin with outcome (the disease) and attempt to discover factors retrospectively that can account for the different health status of the two groups, are known as case-control studies. The early case-control studies were meant, in part, to test whether suspected agents such as CMV or amyl nitrite might be causative factors.
One of the first such studies, by James Goedert and his colleagues at the National Institutes of Health (NIH) and the Uniformed Services
University of the Health Sciences, explored the relationship between KS and amyl nitrite.[23]
James J. Goedert et al., "Amyl Nitrite May Alter T Lymphocytes in Homosexual Men," Lancet 1 (1982): 412-16.
Goedert attempted to assess the new disorder (the outcome) by collecting clinical, virological, and immunological information on two male homosexuals with KS and fifteen healthy homosexual volunteers. The researchers hypothesized that CMV hyperinfection and/or the chronic use of amyl nitrite might be causal variables. In presenting their results and assessing the implications, the investigators suggested that amyl nitrite inhalation may predispose homosexual men to immune deficiency.This investigation had some serious limitations. The small number of subjects in the study, for example, deprived it of the power to find statistical significance if significance existed. Moreover, though amyl nitrite was correlated with immune defects, the researchers did not report controlling for the effects of possible "confounders"—that is, alternative causal variables, such as number of sexual partners or history of infectious diseases. Notwithstanding its defects, this study was cited by others as evidence for the plausibility of amyl nitrite as a causal variable, a tribute, in part, to the power of the life-style hypothesis.[24]
As a causal factor, nitrite continues to attract research attention. For a partial list of studies that tested the association of nitrites to AIDS, see Oppenheimer, "In the Eye of the Storm," note 34, p. 295.
Almost simultaneously with the investigation by Goedert and his colleagues in Bethesda, Michael Marmor and his colleagues in New York City interviewed twenty gay men with biopsy-confirmed KS and forty gay male controls, matched for age and race, eliciting information on sociodemographic characteristics, medical history, sexual practices, and drug consumption. The cases were twenty of the twenty-one males with KS, aged fifty-two or younger, admitted to New York University Medical Center between March 1979 and August 1981. Controls were selected from the private patients of a Manhattan physician. (Since one-third of those asked to be controls refused, it is possible that the control group was skewed in some indeterminate way.) Using multivariate analysis, the investigators found that, of all the study variables, only amyl nitrite and "promiscuity" (as measured by number of different sexual partners per month in the year before onset of disease) appeared to have an independent, statistically significant association with KS.[25]
Michael Marmor et al., "Risk Factors for Kaposi's Sarcoma in Homosexual Men," Lancet 1 (1982): 1083-87.
In October 1981, approximately when the Marmor study began, the CDC undertook a multisite case-control investigation to identify risk factors for KS and PCP in gay men who lacked presdisposing clinical factors for either.[26]
Harold W. Jaffe et al., "National Case-Control Study of Kaposi's Sarcoma and Pneumocystis Carinii Pneumonia in Homosexual Men: Part I, Epidemiologic Results," Annals of Internal Medicine 99 (1983): 145-51.
The CDC chose as controls male homosexuals without KS or PCP, matched to the cases by age, race, and area of residence. Mindful that private-practice controls might not be drawn from preciselythe same population as the cases, the researchers used, where possible, multiple controls—that is, patients from both private practice and STD clinics.
Published in August 1983, the study found that KS and PCP were associated with certain aspects of male homosexuality—in particular, numerous sexual partners per year. Other significant variables were attendance at bathhouses, a history of syphilis, the use of illicit drugs (except nitrites), and exposure to feces during sex. The strong implication was that a subgroup of the male homosexual population, those who were most sexually active, was at greatest risk for KS or PCP. Taking into account the fact that AIDS had by then appeared in other segments of the U.S. population, including hemophiliacs, the authors concluded that an infectious agent might be the necessary cause of the syndrome.
Nonetheless, the CDC was unwilling to dismiss the life-style hypothesis and to commit itself completely to a microbe theory. In the second part of the study report, the authors summarized that position: "Although the cause of the acquired immune deficiency syndrome in homosexual men remains unknown, the study presented here and in the companion paper has identified a distinctive lifestyle as an important risk factor."[27]
Martha F. Rogers et al., "National Case-Control Study of Kaposi's Sarcoma and Pneumocystis Carinii Pneumonia in Homosexual Men: Part 2, Laboratory Results," Annals of Internal Medicine 99 (1983): 151.
The first heterosexual patients, including the first woman, were reported by the CDC in August 1981.[28]
MMWR, pp. 4-5.
The first clinical descriptions of immunosuppression in heterosexual intravenous (IV) drug users appeared in December 1981.[29]Masur et al., "An Outbreak."
By June 1982 the MMWR reported that 22 percent of patients with KS and/or PCP were heterosexuals, the majority IV drug users.[30]MMWR, p. 10.
Almost one-third of the heterosexual patients were women. Despite the early appearance and growing number of heterosexual patients, epidemiological studies of this group were significantly underrepresented in the literature prior to 1984.[31]For the articles published prior to 1984 on HIV infection in drug users and women, see Oppenheimer, "In the Eye of the Storm," note 51, pp. 296-97.
Would investigations of heterosexual patients, paralleling those of gays, have offered a different cast to the life-style model? We will never know for certain. Perhaps chemical toxicity or the immunosuppressive power of heroin, nitrites, and other drugs might have had more significance, at least at the start. But inasmuch as women—some of whom were not IV drug users—were among the first cases, investigators might have hypothesized much earlier that a microbe was the direct cause of the new disorder in all affected groups.
Why, we might ask, were heterosexual intravenous drug users not studied? There is no simple answer. One reason, a structural one, is that
at the federal level the National Institute on Drug Abuse (NIDA) had principal responsibility for investigating issues related to intravenous drug use and had a staff of epidemiologists just for that purpose. NIDA's traditional focus, however, was only on drug abuse; it eschewed investigations of diseases such as hepatitis B and endocarditis, which were endemic or epidemic in its target population. The leadership of NIDA decided that AIDS would be treated like any other disease, thereby leaving the research initiative to other centers at NIH or the CDC.[32]
Don C. Des Jarlais, Ph.D., former coordinator for AIDS Research, New York State Division of Substance Abuse Services, personal communication, January 15, 1988. As exceptions to that decision, NIDA funded some internal biomedical work in 1983, the same year it made a single extramural award to New York State to study risk factors for AIDS in drug users. In 1985 NIDA reversed itself and began to fund AIDS research extensively.
Unfortunately, the CDC, lacking previous experience and expertise, shied away from studying the drug-using population, leaving a lacuna.[33]Stephen Schultz, personal communication, July 27, 1987.
Another reason drug users were not studied is that only a relatively small number of research subjects were available, particularly outside the New York metropolitan area.[34]
Don C. Des Jarlais, personal communication, January 15, 1988.
That problem was alleviated, however, by the development during the summer of 1984 of a blood test measuring antibodies to the HIV. The test created a much larger pool of potential research subjects by identifying individuals who were infected but who did not have AIDS or serious related illnesses.[35]Ibid.
A final answer to the question posed is that epidemiologists were unwilling to study this group.[36]
Stephen Schultz, personal communication, July 22, 1987.
Partly justified by the disinclination of addicts to cooperate in interviews and with follow-up, their unwillingness may also, in part, be explained by a feeling among many clinicians and researchers (in this respect reflecting the attitudes of the public at large) that addicts are of less social consequence than other patients.[37]Ibid.
In a striking reflection of that lack of interest, at all levels of government and in the universities few epidemiologists had expertise in drug addiction when the HIV epidemic began.Despite its appeal, the life-style hypothesis was eventually undercut as a sufficient explanation. During 1982 epidemiological surveillance and case reports clearly indicated that others besides homosexual males were at risk. As an article in the Journal of the American Medical Association (JAMA) observed in September of that year, "if lifestyle is the key, the question still remains: Why has AIDS also occurred in heterosexual men (84 cases so far), women (32 cases so far), mostly heterosexual Haitians, and hemophiliacs?"[38]
Catherine Macek, "Acquired Immunodeficiency Syndrome Cause(s) Still Elusive," Journal of the American Medical Association 248 (1982): 1426.
A new model was required.An Unknown Transmissible Agent
On March 4, 1983, after a year of suggestive data, a Public Health Service Inter-Agency Report (published in the MMWR ) marked a major
shift in the conceptualization of the disorder.[39]
MMWR, pp. 32-34.
That shift was caused in part by the kind of evidence cited by JAMA: case reports to and surveillance by the CDC made it clear that the disease was more than a syndrome of homosexual men and promiscuous life-style.On July 9, 1982, the CDC had reported that thirty-two Haitian immigrants to the United States, seven of them women, showed immunological, morbidity, and mortality patterns similar to those in homosexual men and intravenous drug users.[40]
Ibid., pp. 12-13.
Although the MMWR had previously published two general updates on the increased incidence of the new disease—updates that had included data on heterosexual patients—the article on Haitians constituted the first complete report focusing directly on persons outside the "homosexual" category.A week later, and again in December 1982, the MMWR alerted its readers that patients with hemophilia but no other underlying disease had contracted PCP.[41]
Ibid., pp. 14-15, 24-26.
What the hemophilia patients shared was a dependence on Factor VIII, the clotting substance they lacked, usually derived from the pooled blood of two thousand to nearly twenty thousand donors.[42]Ibid., p. 47.
The possibility of blood as a vector for AIDS was heightened by a CDC report of unexplained immunodeficiency and opportunistic infection in a twenty-month-old infant who had received multiple transfusions from a donor subsequently diagnosed with AIDS.[43]
Ibid., pp. 26-27.
The sibling of the infant was in good health, and his parents were described as "heterosexual non-Haitians" without a history of intravenous drug use.Summing up the new cases, the March 4 MMWR observed that, according to current epidemiological data, four groups were at increased risk of AIDS: homosexual men with multiple sexual partners, users of intravenous drugs, Haitians who had emigrated to the United States in the previous few years, and hemophiliacs. In addition, unexplained immunodeficiency and life-threatening opportunistic infections had occurred in the female sexual partners of bisexual or intravenous drug-using men and in the children born of their unions.
Instead of life-style, the report suggested that the cases shared exposure to a transmissible agent. Though the agent was unknown, the pattern of cases mimicked that of a known pathogen, one that epidemiology had studied and helped control in the years before AIDS.[44]
W. Thomas London and Baruch S. Blumberg, "Comments on the Role of Epidemiology in the Investigation of Hepatitis B Virus," Epidemiologic Reviews 7 (1985): 59-79.
The distribution of AIDS cases parallels that of hepatitis B virus infection, which is transmitted sexually and parenterally. Blood products or blood appear responsible for AIDS among hemophilia patients who require clotting factor replacement. The likelihood of blood transmission is supported by the
occurrence of AIDS among IV drug users. Many drug abusers share contaminated needles, exposing themselves to blood-borne agents, such as hepatitis B virus.[45]
MMWR, p. 33.
In adopting the hepatitis B analogy, epidemiologists posited an alternative organization of known variables, one that stressed a biological agent whose vector was blood and/or its constituents. Although lifestyle factors could be incorporated, they had lost some of their cachet. In the CDC national case-control study, for example, Harold W. Jaffe and his colleagues, reporting their results in August 1983, suggested that life-style factors are indirect causes of AIDS, with a microbe, probably a virus, as the direct cause.[46]
Jaffe et al., "National Case-Control Study," p. 149.
Although epidemiologists had not identified an agent, the model of hepatitis B supported the introduction of public health measures. That is, the model offered a putative point of intervention in the multifactorial "web of causes," even in the absence of a known pathogen. Applying recommendations developed for hepatitis B, the Public Health Service suggested that people avoid sexual contact with persons suspected or known to have AIDS. In addition, members of groups at risk were asked not to donate blood or plasma, and doctors were encouraged to recommend autologous transfusions to their patients. Finally, the Public Health Service called for the development of blood-screening procedures.
On March 4, 1983, for the first time in the MMWR , the CDC referred to high-risk groups, attesting to the spread of AIDS into multiple segments of the U.S. population and to the relationship between the concept of high-risk group and hepatitis B. High-risk groups were those whose members were at a greater risk of infection and of infecting others, carrying a microbe that was capable of spreading through sexual and blood-borne traffic. The MMWR underscored that "each group contains many persons who probably have little risk of acquiring AIDS."[47]
MMWR, p. 32.
Nonetheless, no calibration of degree of risk was introduced, so that no distinction could be drawn. Since no microbe had been isolated, risk designation was, in effect, regarded—even among scientists, not to speak of the news media and among the general public—as synonymous with carrier state.Some months later the CDC justified its use of risk groups, arguing that classification of individuals is intrinsic to any epidemiological investigation.[48]
Ibid., p. 45. Whatever the scientific basis for these high-risk groups, their existence was also open to negotiation. For a short discussion of the successful pressure applied by the Haitian government to have Haitians dropped as a risk group, see Dennis Altman, AIDS in the Mind of America (Garden City, N.Y.: Doubleday, 1986), pp. 71-73.
Classification should not be taken to mean, however, that groups at higher risk for AIDS could transmit the disease through nonintimate contact, since casual transmission was a view unsupported byavailable evidence. To use the likelihood of casual transmission as a basis for social and economic discrimination was unfair.
The apology of the CDC missed the point. Grouping individuals may be traditional in epidemiology, both as a means of intervention and as an analytic prerequisite. The political or social consequences of such grouping are rarely examined. In this instance, even if the fear of casual transmission could be eradicated, the groups identified would still be seen as bearing a strong negative relationship to the life-sustaining blood supply. They were created, qua groups, to signify their potential status as carriers of tainted blood and as contaminators. Moreover, the analogy with highly contagious hepatitis B reinforced the association of casual transmission, particularly for health care providers, because hepatitis B is a disease in which a virus is transmitted through close personal contact, through all secretions, and through wounds and lacerations.[49]
Abram S. Benenson, ed., Control of Communicable Diseases in Man, 12th ed. (Washington, D.C.: American Public Health Association, 1975).
A further consequence of creating high-risk groups was to reinforce the relationship between the disease and "marginal" members of the population. In the case of HIV, although each of the groups ostensibly threatened the remainder of the community through the medium of blood or sex, public health recommendations would inhibit such contamination. Consequently, the disorder could be contained at the boundaries, among people who were "different" from the majority but undifferentiated within each of the high-risk groups.
One of the dangers of a scientific classification of people based on stereotypes was that it defined the questions raised and thus answered. Such categorization created a procrustean mind-set that was evident from the beginning of the epidemic. For example, in early 1982 researchers, in an act of political and scientific oversimplification, designated the new disorder by the acronym GRID (gay-related immunodeficiency), even though the CDC and the New England Journal of Medicine had published reports of heterosexual IV drug users with the new syndrome. At a major conference Michael Gottlieb and his colleagues could report, in a paper entitled "Gay-Related Immunodeficiency (GRID) Syndrome: Clinical and Autopsy Observations," that of the ten adult males in the study with the syndrome, two were exclusively heterosexual.[50]
Michael S. Gottlieb et al., "Gay-Related Immunodeficiency (GRID) Syndrome: Clinical and Autopsy Observations," Clinical Research 30 (1982): 349A.
Ultimately, the hepatitis B metaphor assumed the existence of an infectious agent, probably a virus. Though some favored a new variant of the cytomegalovirus, others, including James W. Curran of the CDC Task Force, supported the notion of a new infections agent.[51]
Jean L. Marx, "A New Disease Baffles Medical Community," Science 217 (1982): 619; Robert C. Gallo, "The AIDS Virus," Scientific American 256 (1987): 48. James Curran was showing slides demonstrating the plausibility of a viral etiology at scientific meetings as early as February 1982 (Pauline Thomas, M.D., director of AIDS surveillance, New York City Department of Health, personal communication, July 28, 1987).
In thelong run, either hypothesis rested on the detection of a pathogen that had hitherto proved elusive.
Aids: "The Story Of A Virus"
From 1981 until the isolation of a new virus, epidemiology played a central role in the characterization of HIV infection. That discipline, using specific case definitions, surveillance, and case-control studies, identified high-risk groups and offered suggestive models and similes. Although epidemiology formulated the social context and morphology of the new disorder, it could not discover its microbial cause. That function was filled by virologists at the Pasteur Institute in Paris and in laboratories in the United States, at the National Cancer Institute (NCI) in particular.
In May of 1984 the journal Science published four reports authored by Robert C. Gallo of the NCI and his colleagues and a fifth by Luc Montagnier of the Pasteur Institute.[52]
Science 224 (1984): 497-508.
These reports established a strong case for a causal link between AIDS and a newly discovered retrovirus that the NCI called HTLV-III and the French called LAV. Later an international agreement was made to call the retrovirus human immunodeficiency virus (HIV).With the isolation of the HIV, the relative importance of epidemiology in the definition of the disease lessened. Epidemiologists continued to play an important, although somewhat more peripheral, role, providing supporting evidence for the viral hypothesis and developing information in areas outside the reach of microbiology and its techniques.
Increasingly, the "bench" scientists—virologists, immunologists, cancer researchers—determined the definition of HIV infection. In effect, they redefined AIDS as a set of biomedical problems open to a chemical resolution in the form of drugs and vaccines. These scientists removed the disorder to a considerable degree from the stigma of its original social matrix, placing it instead in a context resembling that of the supposedly more purely clinical crusades against cancer or polio.
The change in the types of professionals studying HIV infection and in their defined fields of observation and analysis effected a subtle shift in the characterization of the disorder. The disease was increasingly conceptualized in terms of the infections agent, the virus. Interest in cofactors or a multifactorial model diminished.
One marker of this shift was the title of a book published by the
Institute of Medicine and the National Academy of Sciences in 1986: Mobilizing against AIDS: The Unfinished Story of a Virus .[53]
Eve K. Nichols, Mobilizing against AIDS: The Unfinished Story of a Virus (Cambridge, Mass.: Harvard University Press, 1986).
Four years earlier, an article in JAMA had observed that "it seems unlikely that a virus alone is inducing AIDS."[54]Macek, "Acquired Immunodeficiency Syndrome Cause(s) Still Elusive," p. 1425.
Another marker was the dearth of studies of cofactors, of events or states independent of the virus but necessary to cause HIV infection in general or AIDS in particular. In early 1987 an article evaluating cofactors for HIV could cite only one published report on cofactors after 1984.[55]James J. Goedert et al., "Effect of T4 Count and Cofactors on the Incidence of AIDS in Homosexual Men Infected with Human Immunodeficiency Virus," Journal of the American Medical Association 257 (1987): 334.
A few months earlier, another volume by the Institute of Medicine and the National Academy of Sciences, although acknowledging the importance of cofactors, suggested that "there are no data to support the concept [of cofactors], with the possible exception of genital ulcers in Africa."[56]Institute of Medicine and National Academy of Sciences, Confronting AIDS (Washington, D.C.: National Academy Press, 1986), p. 45 (hereafter cited as Confronting AIDS).
The increasingly biological definition of the disease was reinforced by the successful development of serological procedures for the detection of antibodies to the virus. These tests—the enzyme-linked immunosorbent assay (ELISA) and the Western blot technique—allowed epidemiologists and other scientists to outline the biological boundaries of the new disorder.
IN July 1986 the CDC reported that epidemiologists, using the new blood tests, had confirmed that persons in the previously defined groups at higher risk of AIDS showed a greater prevalence of HTLV-III/LAV viral antibody.[57]
MMWR, p. 63.
Epidemiologists also found that AIDS and a number of less full-blown conditions, including lymphadenopathy and AIDS-related complex (ARC), had the same underlying viral cause. In addition, antibody tests demonstrated the existence of the virus in persons without clinical symptomatology, a not unusual pattern in infectious disease epidemiology. These data suggested to the CDC a wide spectrum of human response to the virus, requiring careful study.[58]Ibid.
Standardized blood tests thus initially provided a biological justification for the previously defined high-risk groups. At the same time, antibody testing could determine which individuals within the risk groups were seropositive and which were not. As a result, group membership and carrier status could theoretically be separated. Given the logic of the biological model, moreover, the concept of high-risk membership should actually have withered away, replaced by the notion of high-risk activities that made infection more likely. Despite logic, a shift in emphasis from "status" to "act" did not occur until "mainstream" heterosexuals were targeted as a population at risk.[59]
See, for example, Confronting AIDS, pp. viii-ix.
Since 1984 epidemiologists have also contributed to knowledge of the natural history and transmission of HIV infection. The particular strength
of epidemiology in these areas has derived in part from the "bench" scientists' inability to uncover suitable nonhuman animal models and in part from epidemiologists' technical ability to transcend the ethical limitations on human experimentation by studying disease patterns occurring in populations.
Overall, these epidemiological studies are attempting to enlarge our knowledge of the biological and clinical dimensions of HIV infection, but to develop that knowledge, wherever possible, within the social matrix or behavioral history of the populations involved. By so doing, epidemiologists are maintaining the vitality of a multifactorial, social conception of AIDS in the face of a narrower biological definition.
To date, some of the most important epidemiological studies have prospectively followed defined cohorts of individuals—at first cohorts of homosexual men, but more recently cohorts of hemophiliacs, intravenous drug users, women, and children.[60]
For recent results of cohort studies involving one or more of these groups, see papers presented at the Sixth International Conference on AIDS, San Francisco, June 20-24, 1990.
The purpose of these investigations has been to establish the risk factors for HIV infection; the rate of, and time required for, seroconversion; the progression of pathology in those infected; and the proportion of the infected who eventually develop AIDS. In addition to defining the natural history of the disorder, the researchers aim to find determinative variables that may be open to clinical or social intervention. Finally, epidemiologists continue to develop more extensive and sophisticated means to measure the incidence and prevalence of HIV infection across subgroups in the American population.For example, a number of studies that followed gay or bisexual men over time in New York City,[61]
Cladd E. Stevens et al., "Human T-Cell Lymphotropic Virus Type III Infection in a Cohort of Homosexual Men in New York City," Journal of the American Medical Association 255 (1986): 2167-72.
Holland,[62]Godfried J. P. van Griensven et al., "Risk Factors and Prevalence of HIV Antibodies in Homosexual Men in the Netherlands," American Journal of Epidemiology 125 (1987): 1048-57.
and San Francisco[63]Warren Winkelstein, Jr., et al., "Sexual Practices and Risk of Infection by the Human Immunodeficiency Virus," Journal of the American Medical Association 257 (1987): 321-25.
isolated several possible risk factors for HIV infection. These included sexual contact with a person known to have AIDS and participation as the receptive partner in anal intercourse,[64]Stevens et al., "Human T-Cell Lymphotropic Virus," p. 2169; Winkelstein et al., "Sexual Practices," p. 323.
a risk that increased with the number of persons with whom one acted as the anal receptive partner.[65]Van Griensven et al., "Risk Factors," p. 1055.
These behaviors heightened the chance of viral transmission. Implicated as well was a history of anal douche use.[66]Winkelstein et al., "Sexual Practices," p. 324. For problems with the early cohort studies, see Oppenheimer, "In the Eye of the Storm," p. 288 and note 95, p. 299.
In the population studied, therefore, HIV infection is an STD in which anal mucosa appears to be an inefficient barrier to infection, especially when traumatized by frequent contact. These results, consistent over many epidemiological studies, offered the possibility of behavior intervention strategies.When epidemiologists have researched the natural history of HIV-associated disorders in infected persons, they have provided information on incidence and prevalence rates and, in the main, on biological markers and disease status. Their attempts to isolate cofactors for HIV
infection and progression have yielded, at best, some suggestive leads that must be interpreted with great caution. In addition, these investigations, like those discussed above, suffer from design flaws and biases. For example, most studies cannot specify the dates of HIV infection in their subjects. Consequently, endpoints (lymphadenopathy, for example, or AIDS) cannot be linked to and measured from precisely defined initiatory events. This lacuna often inhibits comparisons of findings across studies and prediction of time-measured outcomes. Recently, however, investigators have attempted both, using as their point of departure the few cohorts (primarily patients infected by blood products) with known or well-estimated dates of seroconversion.[67]
Andrew R. Moss and Peter Bacchetti, "Natural History of HIV Infection," AIDS 3 (1989): 56.
One of the first epidemiological studies of the course of HIV infection was that of Harold Jaffe and his colleagues, which followed a cohort of 6,875 male homosexuals and bisexuals recruited originally between 1978 and 1980 from STD patients at San Francisco City Clinic.[68]
Harold Jaffe et al., "The Acquired Immunodeficiency Syndrome in a Cohort of Homosexual Men," Annals of Internal Medicine 103 (1985): 210-11.
The researchers found that, by 1984, 87.4 percent of a putative random sample[69]About one-third of the sample refused to participate.
of the cohort were seropositive. More recently investigators have estimated that 54 percent of those seropositive for at least ten years will progress to AIDS.[70]A. Lifson et al., "The Natural History of HIV Infection in a Cohort of Homosexual and Bisexual Men: Clinical Manifestations, 1978-1989," paper presented at the Fifth International Conference on AIDS, Montreal, June 4-9, 1989.
These results suggest that without effective treatment a majority of those infected with HIV will eventually develop the last, usually fatal, stage of the disease.B. Frank Polk and his colleagues, unlike Jaffe and his colleagues, attempted to define predictors of AIDS in seropositive men by studying a cohort of 1,835 male homosexual volunteers recruited by centers in four cities: Los Angeles, Chicago, Pittsburgh, and Washington/Baltimore.[71]
B. Frank Polk et al., "Predictions of the Acquired Immunodeficiency Syndrome Developing in a Cohort of Seropositive Homosexual Men," New England Journal of Medicine 316 (1987): 61-66.
When each of the fifty-nine AIDS cases (developing over a median of fifteen months) was matched to five seropositive controls from the same study center, the researchers found three independent predictors of AIDS: a decreased number of T helper cells, a low level of HIV antibody, and a history of sex with someone who subsequently developed the syndrome. The first two predictors, however, are probably biological markers of disease progression to AIDS rather than determinants or causes of that progression. The last predictor may in fact be a marker of an infection longstanding enough for AIDS to develop in both partners. More recent epidemiological investigations of HIV-infected homosexual men and men with hemophilia have identified additional laboratory markers of progression to AIDS.[72]Andrew R. Moss, "Predicting Who Will Progress to AIDS," British Medical Journal 297 (1988): 1067-68; Moss and Bacchetti, "Natural History," pp. 57-58.
Cohort studies have also provided the basis for estimates of the "latency period," the median time between an initial infection and frank AIDS. In seropositive homosexual men, transfusion recipients, and hemophiliacs,
the latency period is an estimated seven to eleven years; and half of those infected are free of AIDS for an indefinitely longer term.[73]
Andrew R. Moss et al., "Seropositivity for HIV and the Development of AIDS or AIDS Related Condition: Three Year Follow Up of the San Francisco General Hospital Cohort," British Medical Journal 296 (1988): 745-50; Moss and Bacchetti, "Natural History," pp. 56-57; Alvaro Munoz et al., "Acquired Immunodeficiency Syndrome (AIDS)-Free Time after Human Immunodefiency Virus Type 1 (HIV-1) Seroconversion in Homosexual Men," American Journal of Epidemiology 130 (1989): 530-39.
This highly variable latency period will probably be extended further, moreover, with the prophylactic administration of AZT.[74]Gina Kolata, "Strong Evidence Discovered That AZT Holds Off AIDS," New York Times, August 4, 1989, p. A1; Philip J. Hilts, "Drug Said to Help AIDS Cases with Virus but No Symptoms," New York Times, August 18, 1989, p. A1.
In fact, new evidence appears to show that for some individuals the period between HIV infection and the appearance of persistent antibodies to HIV may be even longer than previously suspected. For some years, the normal period was thought to be three months or less.[75]
Moss and Bacchetti, "Natural History," p. 55.
Investigators in Los Angeles, however, have recently reported multiple instances of delayed seroconversion.[76]David T. Imagawa et al., "Human Immunodeficiency Virus Type 1 Infection in Homosexual Men Who Remain Seronegative for Prolonged Periods," New England Journal of Medicine 320 (1989): 1458-62.
Although HIV was isolated from 31 of 133 homosexual men, 27 of the 31 had no antibodies to HIV during the next thirty-six months of follow-up when their sera were tested by the ELISA and Western blot methods. Confirming the results of previous investigations,[77]Steve Wolinsky et al., "Polymerase Chain Reaction (PCR) Detection of HIV Provirus before HIV Seroconversion," paper presented at Fourth International Conference on AIDS, Stockholm, June 12-16, 1988; M. Loche and B. Mach, "Identification of HIV-Infected Seronegative Individuals by a Direct Diagnostic Test Based on Hybridisation to Amplified Viral DNA," Lancet 2 (1988): 418-21.
this study suggests that for an unknown number of individuals a "silent HIV infection," undetectable by conventional blood assays, may be, in fact, part of the latency period.These recent results carry several further implications. They raise questions about the limitations of current serum antibody tests, particularly worrisome if those with "silent" HIV infection can still transmit the virus. On a more positive note, these results suggest that some infected individuals have immune systems that successfully suppress the replication of HIV indefinitely. This finding has potentially profound implications for future drug research and therapy.
Why does HIV disease have such a variable incubation period? This question intrigues researchers and has renewed their interest in cofactors—exogenous or endogenous exposures that might modulate the rate of HIV-induced immunodeficiency.[78]
Confronting AIDS, p. 193.
Investigators have also hypothesized that cofactors may promote initial HIV infection. For example, some have suspected that a history of microbial infections, leading to immunological alterations, may put individuals at greater risk of HIV infection and of disease progression.[79]Thomas C. Quinn et al., "Serologic and Immunologic Studies in Patients with AIDS in North America and Africa," Journal of the American Medical Association 257 (1987): 2617-21.
There is growing evidence that sexually transmissible infections—particularly those that produce genital ulcerations, which, like douching and enemas, facilitate invasion of HIV—may be important cofactors.[80]Peter Piot et al., "Serum Antibody to Haemophilus Ducreyi as a Risk Factor for HIV Infection in Africa, but Not in Europe"; Edward E. Telzak et al., "A Prospective Cohort Study of HIV-1 Seroconversion in Patients with Genital Ulcer Disease in New York City"; and Robert Cannon et al., "Syphilis Is Strongly Associated with HIV Infection in Baltimore STD Clinic Patients Independent of Risk Group"—all presented at Fifth International Conference on AIDS, Montreal, June 4-9, 1989. See also Jacques Pepin et al., "The Interaction of HIV Infection and Other Sexually Transmitted Diseases: An Opportunity for Intervention," AIDS 3 (1989): 3-9.
Evidence is also accumulating that, for reasons not yet understood, lack of circumcision in African men may be a cofactor for HIV infection.[81]Jean L. Marx, "Circumcision May Protect against the AIDS Virus," Science 245 (1989): 470-71; J. Bongaarts et al., "The Relationship between Male Circumcision and HIV Infection in African Populations," paper presented at the Fifth International Conference on AIDS, Montreal, 1989.
According to some researchers, the simultaneous existence of genital ulcers in HIV-infected women and lack of circumcision in their partners may potentiate female to male transmission of the virus.[82]D. William Cameron et al., "Female to Male Transmission of Human Immunodeficiency Virus Type 1: Risk Factor for Seroconversion in Men," Lancet 2 (1989): 403-7.
There are also epidemiological indications that age-related variables may be cofactors for disease progression, sinceinfants and older homosexual men have higher rates of disease progression than other groups.[83]
Angelos Hatzakis et al., "Age at Time of HIV Infection as Cofactor of Progression to Advanced Immune Dysfunction and AIDS," paper presented at the Fifth International Conference on AIDS, Montreal, 1989; J. Roy Robertson et al., "Progression to AIDS in Intravenous Drug Users, Cofactors and Survival," paper presented at the Sixth International Conference on AIDS, San Francisco, 1990.
The possible role of cofactors testifies to the terrible complexity of HIV infection and justifies the reluctance of epidemiologists to reduce AIDS and related conditions to an agent-host phenomenon. Epidemiological researchers have consistently held up the possibility of nonviral factors to the "bench" scientists. Since 1981 they have rooted biological or clinical events in the matrices of human behavior and social experience. In one study of the role of cofactors in HIV infection, the authors put the epidemiologists' position quite well.[84]
Quinn et al., "Serologic and Immunologic Studies," pp. 2617, 2620.
Citing the viral etiology common to all patients with AIDS, they stressed the multiple determinants probably responsible for HIV infection and disease progression, including cultural differences, the presence of other endemic illnesses, and host and viral genetic factors. Their position reaffirms the multifactorial model as central to an understanding of HIV infection and to its control.From Aids To Hiv Infection: Tracking The Epidemic
While investigating the natural history of HIV infection, epidemiologists have continued to hold responsibility for an apparently mundane task: systematic surveillance. Since 1981 the CDC has both constructed the surveillance case definitions of AIDS and served as the national registry for all cases reported by the states, the District of Columbia, and the U.S. territories.[85]
AIDS has been a reportable condition since 1983, when the Council of State and Territorial Epidemiologists passed a resolution to that effect.
These data are used to monitor the spread of AIDS, project its future incidence and prevalence,[86]James W. Curran et al., "Epidemiology of HIV Infection and AIDS in the United States," Science 239 (1988): 610-16; Karon, Dondero, and Curran, "Projected Incidence of AIDS."
and provide the basis for health service planning and health education.In recent years, however, the systematic surveillance of cases has grown more problematic. Sources within and outside the CDC have observed that the true number of AIDS cases in the United States has been underreported, thereby weakening the epidemiological and policy functions the data serve. In addition, once the HIV virus was isolated, epidemiologists sought strategies to capture population-based information on HIV seroprevalence in general, not only on AIDS, the last stage of the disease. Methods developed by epidemiologists—the CDC in particular—to survey HIV prevalence put them at odds with other quantitative research workers and, for the first time, threatened the monopoly previously enjoyed by epidemiologists over the population-based definition of the disease.
In an editorial note in the MMWR of August 18, 1989, the CDC admitted that its AIDS case count was subject to error: "Because of the combination of underdiagnosis and underreporting of AIDS cases and severe manifestations of HIV infection that do not meet the CDC AIDS surveillance case definition, reported AIDS cases underestimate the number of persons severely affected by HIV since 1981."[87]
MMWR 38 (1989): 562.
Since the completeness of the case count varied by geographical region and patient population, the CDC surveillance system had captured only 70 to 90 percent of HIV-related deaths.In a separate assessment of the CDC's system for reporting AIDS cases, the Committee on AIDS Research and the Behavioral, Social and Statistical Sciences of the National Research Council (NRC) highlighted two problems: only 85 to 90 percent of cases are reported within one year of diagnosis, with a further decline expected; and the reliability and validity with which the mode of transmission of infection is established in each case have not been evaluated.[88]
Charles F. Turner, Heather G. Miller, and Lincoln E. Moses, AIDS, Sexual Behavior and Intravenous Drug Use (Washington, D.C.: National Academy Press, 1989), pp. 32-33.
Flaws in the CDC's methodology for establishing mode of transmission could affect the degree to which subpopulations are over- or underrepresented in the national surveillance system; such misreporting might have serious implications for identifying or tracking shifts in the spread of infection.Indeed, in a recent study the sociologist E. O. Laumann and his colleagues concluded that some segments of the U.S. population are systematically underrepresented. Arguing that the national reporting system is subject to systematic distortions because of "overt manipulations by interested parties" and the stigmatizing nature of HIV infection itself,[89]
E. O. Laumann et al., "Monitoring the AIDS Epidemic in the United States: A Network Approach," Science 244 (1988): 1186. By "overt manipulations" the authors mean that the highly decentralized CDC reporting system allows individual physicians and hospitals considerable opportunities to hide cases of AIDS if they have an interest in doing so (John H. Gagnon, personal communication, August 31, 1990).
the investigators used instead the 1988 General Social Survey (GSS), a national household survey in which respondents were asked to identify all those within their network of acquaintances who had either been a victim of homicide or had AIDS. When the GSS results were compared with official national statistics on homicide, the two were congruent. When a similar comparison was made between GSS survey data and those of the CDC, the investigators found that the national surveillance system significantly underestimated the prevalence of AIDS in white middle-class populations and in those living in the Midwest, while overstating the prevalence of that disease in blacks and latinos and in those living in the East. The researchers called for more prevalence studies independent of the CDC's surveillance network, in order to ensure a more accurate assessment of the social epidemiology of AIDS.The most critical evaluation of the national AIDS surveillance system is that of the U.S. General Accounting Office (GAO).[90]
U.S. General Accounting Office, AIDS Forecasting: Undercount of Cases and Lack of Key Data Weaken Existing Estimates (Washington, D.C.: General Accounting Office, June 1989). A study published some months later, albeit based on the experience of only one state, found that only an estimated 60 percent of AIDS cases in South Carolina were reported to the state's registry in 1986 and 1987. See George A. Conway et al., "Underreporting of AIDS Cases in South Carolina, 1986 and 1987," Journal of the American Medical Association 262 (1989): 2859-63.
The GAO hasfound that the system substantially undercounts the number of AIDS cases in the United States. It attributes that problem to essentially four sources, some already identified above. In essence, these sources are (1) the CDC's surveillance definition, which specifies those illnesses that qualify a case as AIDS and thereby excludes a considerable number of fatal HIV-related cases—in particular, young intravenous drug users—who never contract the required diseases;[91]
Rand L. Stoneburner et al., "A Larger Spectrum of Severe HIV-1 Related Disease in Intravenous Drug Users in New York City," Science 242 (1988): 916-19.
(2) the CDC's test criterion, which excludes from the national surveillance system all cases of AIDS diagnosed without HIV test results—despite the fact that such presumptive diagnoses are not rare and are increasing as physicians become more experienced with AIDS and as patients insist that no test results be attached to their medical charts; (3) physician error, as a consequence of which AIDS cases go undiagnosed or are diagnosed late; and (4) surveillance system breakdown, in which diagnosed cases are never reported or are reported late.The GAO estimates that, because of these and other sources of error, the national surveillance system may have counted only two-thirds of the cases of AIDS and other HIV-related fatal illnesses in the United States—an estimate that is lower than the CDC's own estimate of 70 to 90 percent. Whatever the precise shortfall, the combined results of the GAO and other studies suggest that the current AIDS case count may be sufficiently flawed to affect health planning or estimates of future cases, particularly for subpopulations or specified regions of the country. Unfortunately, similar, though perhaps more profound, flaws may be vitiating the recent HIV surveillance projects.
The need to monitor HIV infection rather than only AIDS was clearly adumbrated by the Committee on AIDS Research and the Behavioral, Social and Statistical Sciences of the NRC:
Counts of AIDS cases are out-of-date indicators of the present state of the epidemic. There is a long, asymptomatic latency period between HIV infection and the development of AIDS (in most persons). Consequently, the statistics on new AIDS cases reflect old cases of HIV infection. … [In addition,] persons whose life spans are significantly shortened by HIV infection do not always manifest sufficient symptoms to be captured by the AIDS reporting system. … [Finally,] the future magnitude of the AIDS epidemic will be determined primarily by the current extent and future spread of HIV infection in the population.[92]
Turner, Miller, and Moses, AIDS, pp. 31-32.
The CDC recognized the need for HIV seroprevalence data quite early. In the fall of 1985, six months after the ELISA was licensed, the CDC proposed that selected "sentinel" hospitals across the country provide
sera for "blinded" seroprevalence surveys—surveys that use anonymous samples of blood and therefore do not require informed consent, so that they are relatively free of self-selection bias.[93]
Ronald Bayer, L. H. Lumey, and Lourdes Wan, "The American, British and Dutch Responses to Unlimited Anonymous HIV Seroprevalence Studies: An International Comparison," AIDS 4 (1990): 283-90.
Once initiated, this plan was followed by another, outlined in September 1987 and implemented thereafter,[94]
U.S. Department of Health and Human Services, Public Health Service, Centers for Disease Control, Human Immunodeficiency Virus Infections in the United States: A Review of Current Knowledge and Plans for Expansion of the HIV Surveillance Activities, a Report to the Domestic Policy Council (Washington, D.C.: DHHS, November 30, 1987).
to develop a "comprehensive family of complementary HIV surveys" that would capture seroprevalence information on pregnant women, those at high risk of HIV infection, and selected subgroups within the general population.[95]Timothy J. Dondero, Jr., Marguerite Pappaioanou, and James W. Curran, "Monitoring the Levels and Trends of HIV Infection: The Public Health Service's HIV Surveillance Program," Public Health Reports 103 (1988): 213-20.
Specifically, the CDC agreed to provide technical and financial support to thirty large metropolitan areas across the United States.[96]Of the thirty large metropolitan areas, twenty are cities that report 75 percent of the current cases of AIDS; the remaining ten were selected from cities with moderate to low prevalence of AIDS.
In each of these urban areas, the federal government, in collaboration with state and local agencies, selected in a nonrandom fashion one or more of six types of health care institutions or groups: sentinel hospitals, newborn infants, tuberculosis clinics, STD clinics, drug treatment centers, and women's health centers. Only "blinded" surveys are conducted in the first three; "blinded" and "unblinded" studies in the last. (Such "unblinded" studies allow investigators to ask in-depth questions, but they require informed consent of the respondents and run the risk, as recent studies have shown, of self-selection bias—the nonrandom refusal of some, perhaps those at greatest risk, to participate.[97]Harry F. Hull et al., "Comparisons of HIV-Antibody Prevalence in Patients Consenting to and Declining HIV-Antibody Testing in an STD Clinic," Journal of the American Medical Association 260 (1988): 935-38.
) According to the CDC, the family of surveys is central to defining and managing the problems presented by HIV infection: "Information on current levels and trends of HIV infection is needed to follow the course of the epidemic, to help project future trends in AIDS incidence, and to target and evaluate the impact of AIDS/HIV preventive programs."[98]U.S. Centers for Disease Control and National Institute on Drug Abuse, "Proposal for Monitoring HIV Seroprevalence in Intravenous Drug Users in Treatment, National HIV Seroprevalence Surveys," CDC Protocol No. 840, 1988.
The CDC has elected to use health care institutions to capture prevalence information, a traditional epidemiological strategy. The types of facilities selected allow it to obtain seroprevalence data on those at greatest risk of infection: the sexually active (STD clinics), intravenous drug users (drug treatment and tuberculosis clinics), and childbearing or reproductive-age women in lower socioeconomic strata (newborn screening and women's health centers). The CDC admits that the survey design for each of these subpopulations is flawed. It hopes, however, to analyze and evaluate the biases in each design and to compensate for them statistically, so that it can provide accurate prevalence estimates.[99]
Karon, Dondero, and Curran, "Projected Incidence of AIDS," p. 547.
The Committee on AIDS Research and the Behavioral, Social and Statistical Sciences of the NRC, advised by a panel of statisticians and demographers, has examined the six surveys in depth. It found that, contrary to the CDC's expectations, the "comprehensive family of surveys" is sufficiently flawed in research design to prevent it from accurately
measuring, with knowable margins of error, the incidence or prevalence of HIV infection in the subpopulations of interest. Central to the committee's criticism is that the CDC is using nonrandom samples—samples of convenience—in all surveys except newborn screening.[100]
Turner, Miller, and Moses, AIDS, pp. 52-62.
The committee's subsequent conclusions are unequivocally critical of the CDC:The committee has listened with interest to arguments that population-based estimates of HIV incidence and prevalence are unnecessary from a public health perspective. Rather, it has been suggested that targeted samples of convenience could suffice to provide "sentinels" that could be used to guide the nation's response to the AIDS epidemic. The committee concludes that it would be a serious mistake for the Public Health Service to continue to "make do" with estimates derived from convenience samples . … Now is the time to prepare for the future, and good data will be indispensable in future efforts to control the epidemic. No postponement should be accepted.[101]
Ibid., pp. 68-70.
To meet its objections, the committee suggests that the CDC reconstitute each of the seroprevalence surveys as probability samples, despite the administrative, political, and financial difficulties involved. In reformulating the surveys, the committee urges the CDC to draw on the expertise of the National Center for Health Statistics (recently made a part of the CDC), which employs statisticians, demographers, and other social scientists.[102]
Ibid., p. 7.
The committee does not comment, however, on another significant limitation of the "comprehensive family of surveys." These surveys are limited to groups historically at risk of HIV infection and to a special subgroup, the hospitalized sick, which only in part includes those at low risk. The CDC's surveys do not, however, measure seroprevalence in the population at large and therefore cannot estimate, with known margins of error, the prevalence of HIV in the United States. In addition, the surveys cannot monitor the incidence of HIV in new, previously unknown, risk groups.[103]Other federal agencies measure seroprevalence in segments of the general population—for example, civilian applicants for military service, active-duty military personnel, and Job Corps entrants; but the results obtained are flawed by self-selection bias.
Responding to the need to measure HIV prevalence in the general population, the National Center for Health Statistics (NCHS) has sponsored a National Household Seroprevalence Survey (NHSS), contracting with a private research organization, the Triangle Research Institute (TRI), to conduct feasibility tests. The ultimate objective of the NHSS will be to survey 50,000 anonymous household respondents concerning factors that might put them at risk for HIV infection and to take a blood sample from each participant. These respondents are to be randomly selected on the basis of probability sampling; the result should
be an estimate of HIV prevalence in the total U.S. population. Before the government approves the survey, however, a pilot stage must successfully demonstrate that the study is feasible and can generate new and useful data.[104]
Research Triangle Institute, "National Household Seroprevalence Survey, Pilot Study Summary Report," Contract No. 200-88-0605, Research Triangle Park, N.C., April 1989, p. 1.
Specifically, the pilot involves a careful evaluation of all field procedures and research methodologies, including sampling strategies, protection of the respondents, blood collection methods, survey design, and development of community support.After an aborted start in Washington, D.C., where local officials and community groups rejected the project, TRI successfully piloted the NHSS in Allegheny County, Pennsylvania, in January 1989; it initiated a second study in Dallas in September of the same year. The results from Pennsylvania show that, of 308 randomly selected households with an eligible respondent (a civilian, permanent resident, eighteen to fifty-four years of age), 85 percent agreed to participate in the study.[105]
Ibid., p. 7.
In Dallas a survey of 1,715 eligible households, completed in December 1989, achieved an overall response rate of 84 percent (90 if one includes those who completed the questionnaire but refused to be bled). Reaching that number proved somewhat more difficult than anticipated, because the leading gay political and service organization in the city, the Dallas Gay Alliance, actively campaigned against the survey.[106]On the Dallas survey, see National Center for Health Statistics, "Report on the Dallas County Household HIV Survey," Hyattsville, Md., May 1990; and Bruce Lambert, "Dallas AIDS Survey Is Begun amid a Furor over Its Worth," New York Times, September 27, 1989, p. A1.
With the feasibility studies completed, the CDC, along with other federal bodies, must now decide whether a national seroprevalence survey is technically and politically possible. The fierce local controversies in Washington and Dallas make political considerations important; so, too, do actions in Congress, where in July 1989 conservative members of the House Appropriations Committee were able to delete the $11 million required to fund a Public Health Service survey of sexual behavior in the United States.[107]
Michael Specter, "Funds for Sex Survey Blocked by House Panel," Washington Post, July 26, 1989, p. A3.
However, the CDC reportedly has an antipathy to the NHSS that predates and is independent of these political considerations.[108]Privileged communication.
The epidemiologists of the Centers for Disease Control had argued early on that the study, requiring blood samples and a survey of sex- and drug-related behaviors, would be vitiated by nonresponse bias; it would be bad science. They also insisted that the NHSS was politically untenable, in that it needed substantial outreach in the face of community opposition. Finally, the NHSS would consume funds that were better spent on the family of surveys, which, with its use of "blinded" seroprevalence studies, was unbiased (good) science.The arguments raised by the CDC regarding the scientific and political feasibility of the NHSS are somewhat disingenuous, in that they hide a struggle on the part of the CDC to maintain the hegemony of its
own mission and culture over the HIV "territory."[109]
Privileged communication.
The CDC has dominated the population-based study of AIDS since 1981. It has defined the disease for surveillance purposes, directed the national AIDS-reporting system, and designed the "comprehensive family of surveys" to expand that system to the whole spectrum of HIV infection. That design was based on traditional medical epidemiology; to measure rates of disease, the CDC has used patient data captured within health care institutions. After almost a decade of work and achievement, "accomplished in the face of considerable adversity on a number of fronts—physical, diplomatic, political and administrative"[110]Turner, Miller, and Moses, AIDS, p. 70.
—it would be strange if the CDC did not feel that if "owns" to a large degree the population-based study of HIV (as does epidemiology, through it). It would be surprising for the CDC to easily relinquish its funding, political power, and high visibility.The CDC experiences as an incursion the criticism and critical work of quantitative social scientists, most of whom are relatively new to AIDS research. These social scientists' insistence on population probability sampling—the General Social Survey or the NHSS, for example—as the basis of good science at least temporarily excludes the CDC. The CDC has little experience with the methodologies involved, and the CDC staff in Atlanta includes no sampling statisticians and precious few quantitative social scientists on the Ph.D level.[111]
Ibid., p. 24.
To alter course now requires the CDC to change both corporate strategy and corporate culture and to allow non-epidemiologists, with their own mission and culture, to participate in the population-based definition of HIV. The CDC is loath to share this territory, and a certain degree of inflexibility, even dogmatism, has followed. For example, the leadership of the CDC has made its calculation of 1 to 1.5 million HIV seropositive individuals in the United States an article of faith, despite the fact that the figure is only an estimate, based on much-criticized parameters.[112]Privileged communication. In 1986 the CDC estimated that 1 to 1.5 million people in the United States were infected with the HIV, a range that it modified the next year to between 945,000 and 1.4 million; see Institute of Medicine and National Academy of Sciences, Confronting AIDS: Update 1988 (Washington, D.C.: National Academy Press, 1988), p. 51. Early in 1990 the CDC changed the estimate slightly to between 800,000 and 1.3 million.
The social scientists who criticize the manner in which the CDC defines cases or collects the data are demanding something more than greater methodological purity—although that is important. They are also frustrated by the dominant role played by the CDC and other epidemiologists in defining and managing the HIV epidemic. Their criticism of the CDC is only the most public expression of anger at the power and apparent insensitivity of epidemiologists, who are seen as excluding, devaluing, or co-opting social science methodologies and objectives. Social scientists argue there are sound reasons for multiple approaches to studying the epidemic. Such approaches would, for example,
enable researchers to analyze with greater sophistication the personal (particularly the sexual) behavior of individuals; to measure the unique, local configurations and manifestations of the HIV epidemic; and to develop models of the political economy of that epidemic.[113]
John H. Gagnon, personal communication, August 31, 1990.
In brief, social scientists want badly to broaden the theoretical and empirical basis for the study and management of the epidemic. Such a change, in which they would have a greater voice in defining public policy, would enhance the power and prestige of these professionals and might (although this is not certain) increase the amount of research dollars available to them.Is the role of epidemiology, of the CDC, in defining HIV infection coming to an end? Most certainly not. That role, however, may be undergoing a subtle shift—not so dramatic as when the HIV was discovered, but still a change of position to make room for the social scientists. The degree of that displacement will depend on a number of issues, not all in the epidemiologists' control. How will the CDC, for example, incorporate the statisticians and social scientists at NCHS into its HIV data collection projects? Will the CDC benefit from the reluctance of political conservatives to survey Americans in their homes about sex- and drug-related activities? Will the growing insistence by clinicians and some public health officials that all pregnant women, surgical patients, and hospital patients undergo "unblinded" serotesting politicize and undercut all seroprevalence studies? Only time, that most confounding of variables, will tell.
Conclusion
I have outlined how epidemiologists, drawing on the unique perspectives of their profession, reacted to the outbreak of a new disease of unknown origin. By responding early to the epidemic, epidemiologists defined the syndrome first—an act of scientific acumen and power. Over time, however, investigators using other techniques have challenged the primacy of the epidemiologists' construction of the disorder, both of the disease itself and of the hypothetically infected population. To the extent that these challenges were successful, the definition of the disorder has changed, and with it the relative standing of epidemiologists (the CDC in particular). The history of the epidemic demonstrates that the construction of HIV infection was and is a dynamic process in which different scientific specialties negotiated definitions that, to a degree, reflected their relative power.
In the process, the legacy of epidemiologists remains significant. From the beginning of the epidemic, epidemiologists conceptualized HIV infection as a complex social phenomenon, with dimensions that derived from the social relations, behavioral patterns, and past experiences of the population at risk. On the one hand, the epidemiologists' approach may have skewed the choice of models and the hypotheses pursued and may have offered some justification for homophobia. On the other, by defining HIV infection as a multifactorial phenomenon, with both behavioral and microbial determinants, epidemiologists offered the possibility of primary prevention, a traditional epidemiological response to infectious and chronic diseases. Epidemiologists, in effect, established the basis for an effective public health campaign and—through publications, conferences, and the continuous collection of surveillance data—helped make AIDS a concern of policymakers and the public.
Primary prevention—including blood screening, health education, and behavior modification—is currently the only effective social response to the spread of HIV infection. Evidence from several sites indicates that the rate of HIV infection among some groups of homosexual males and IV drug users has begun to decline, possibly because of a reduction in high-risk activities.[114]
Marshall A. Becker and Jill G. Joseph, "AIDS and Behavioral Change to Reduce Risk: A Review," American Journal of Public Health 78 (1988): 394-410; Andrew Moss et al., "Seroconversion for HIV in Intravenous Drug Users in Treatment in San Francisco, 1985-1990," paper presented at the Sixth International Conference on AIDS, San Francisco, 1990.
These results, hopeful signs, have not yet been linked to a decrease in HIV-associated mortality. They may presage, however, a parallel between HIV and past infectious disease experiences.Historical epidemiology has shown that medical interventions, both chemotherapeutic and prophylactic, have had little impact on the overall decline in infectious disease mortality in this century. For example, John and Sonja McKinlay found that since 1900 new medical measures have had almost no detectable effect on U.S. disease-specific mortality rates, because such measures usually occurred some decades after significant declines in death rates had already set in.[115]
John B. McKinlay and Sonja M. McKinlay, "The Questionable Contribution of Medical Measures to the Decline of Mortality in the United States in the Twentieth Century," Milbank Memorial Fund Quarterly 55 (1977): 425.
Thomas McKeown and his colleagues obtained similar results in a study of the mortality trends in England and Wales. According to McKeown, the observed secular decline was mainly attributable to community factors, particularly better nutrition and hygiene.[116]Thomas McKeown et al., "An Interpretation of the Decline of Mortality in England and Wales during the Twentieth Century," Population Studies 29 (1975): 391-422.
It remains to be seen whether HIV-related mortality will also decline as a result of community-directed hygiene (condoms, clean needles, blood screening) before a vaccine or new chemotherapy can be introduced. If it does, the history of HIV infection will offer a powerful vindication of the epidemiologists' multifactorial social definition of disease and of the public health actions that followed from it.