Moving toward Clarity on 2 Fronts in the Epidemiology of Kaposi Sarcoma–Associated Herpesvirus Infection

The first several years after the discovery of Kaposi sarcoma (KS)–associated herpesvirus (KSHV, also known as human herpesvirus 8) [1] featured a flurry of epidemiologic work establishing its causal role in KS as well as its major routes of spread. On a variety of findings, there was an uncommon degree of consensus among investigators. In subsequent years, however, after the low-lying epidemiologic inferences were harvested, there has often been more controversy than consensus. In this issue of the Journal, 2 articles fortunately move 2 controversies closer to clarity. In the first study, Engels et al. [2] examined KSHV transmission among US adults. Interestingly, 17 years earlier, epidemiologists predicted the transmission of the yet-to-be-discovered causal agent of KS. That is, the disproportionate incidence of AIDS-related KS among homosexual men suggested that the agent was infectious, sexually transmitted, predominantly found in homosexual men, and not ubiquitous in the general population (at least in industrialized settings) [3]. Shortly after KSHV was discovered, this prediction was confirmed. Specifically, KSHV was predominantly found in homosexual men, in whom it is related to the number of sex partners, which established intimate contact between men as a major route of spread. What was not clear from the early work is to what extent KSHV is transmitted sexually between heterosexual individuals. The documented spread by intimate contact among homosexual men logically suggests that sexual spread occurs in other sexually active populations—in fact, that saliva is the body fluid that most commonly harbors KSHV implies that heterosexual spread should easily occur. Yet epidemiologic investigation in nonendemic settings, such as the United States and northern Europe, has yielded mixed findings. For example, a large sexually transmitted disease clinic study found no evidence of spread among heterosexual adults [4]. Where heterosexual transmission has been suggested [5], confounding by injection drug use transmission might provide an alternative explanation [6]. To bring clarity to the field, Engels et al. [2] conducted one of the largest and most representative studies of KSHV seroprevalence to date. They tested 13,894 adults from the National Health and Nutrition Examination Survey (NHANES) III, a probability sample of the US general population. Because of the field's lack of a standardized and commercially available KSHV antibody assay, the authors were cautious in establishing assay cut points for seropositivity. Fortunately, the validity of their inferences was strengthened by their use of highly specific cut points, an approach known to result in unbiased relative seroprevalences between groups, regardless of assay sensitivity [7]. Engels et al. determined first that homosexual men have by far the highest KSHV seroprevalence in the United States, nearly 6 times higher than that in heterosexual individuals. This confirms earlier population-based studies of more concentrated geographic areas [8]. The advance in the article stems not from what was found in homosexual men but rather what was not found in heterosexual persons. Specifically, among 7594 women, there was no association between KSHV seropositivity and duration of sexual activity, number of sex partners, or the presence of several sexually transmitted infections. The overall KSHV seroprevalence among 6231 men who denied homosexual behavior was identical to that in women. Although there was some evidence in these men of an association between sexual behavior and KSHV seropositivity, the face-to-face nature of the NHANES interview likely caused underreporting of homosexual behavior. Indeed, the authors note that the percentage of homosexual men was less than half that of other surveys; hence, homosexual men misclassified as heterosexual may account for whatever associations with sexual behavior that were present. The work of Engels et al., therefore, helps bring clarity to the epidemiology of KSHV in the United States by determining that the public health relevance of KSHV infection is squarely concentrated among homosexual men. Although the study cannot conclude that heterosexual transmission never occurs, it is substantially less efficient than that between homosexual men—an inference that should firmly shape the future research agenda. The focus now must turn to why homosexual men are disproportionately infected, to how KSHV is spread among homosexual men, and to alerting homosexual men of their notable risk. These 3 objectives are interrelated, in that ignorance of how KSHV is transmitted among homosexual men has likely weakened attempts at community prevention. Whatever the cause, homosexual men in general have scant awareness of this organism [9]. Finally, it should be noted that no single absolute KSHV seroprevalence in the study by Engels et al. can be accepted as truth. Indeed, seroprevalence in homosexual men ranged from 8.2% to 25.2%, depending on the cut point. Instead, only comparisons between groups have meaning. Although reliance on between-group comparisons has served epidemiologists well in generating our current understanding of KSHV, this has diminishing value for the next set of questions. Understanding how KSHV is transmitted among homosexual men will require accurate ascertainment of within-subject seroconversion. Moreover, to move KSHV from a research-level organism to one that can be tested for in the clinic (e.g., to test donors or recipients of solid organ transplants) will also require highly accurate, commercially available diagnostic assays. In the other KSHV-related work in this issue, Jenkins et al. [10] sought to extend their earlier observation [11] regarding an association between KSHV and prostate cancer. Given the enormity of what a causal association would mean for this common malignancy, it is important to examine the entire spectrum of evidence. To start, the finding of KSHV nucleic acid in prostate tissue [12] was sufficient to pose the question regarding an etiologic role for KSHV. This was buttressed by reports linking sexual behavior to prostate cancer [13]. However, there were holes in the preliminary argument for KSHV as a prostate cancer agent. Specifically, if KSHV were causative, we would expect that homosexual men, or conditions caused by KSHV, would be linked to prostate cancer. Instead, unmarried men (a surrogate for homosexual men) do not have a higher risk [14]. Furthermore, registry studies have not found associations between KS and prostate cancer [15]. Finally, analogous to KS, it might be expected that a high incidence of prostate cancer would be observed in HIV disease. Yet this has not been the case [16]. Therefore, even before direct studies of KSHV and prostate cancer were evaluated, the probability for an association was low. Because of the relative rarity of prostate cancer, all epidemiologic studies directly examining a role for KSHV have featured case-control designs. Although such studies can yield valid inferences when well designed, they are prone to bias when not, which is typically the result of improper control selection. Thus, a critical determinant of the validity of a case-control study is whether control subjects are sampled from the population that ultimately gave rise to the case patients; this is the study base principle [17]. The first case-control study of KSHV and prostate cancer was conducted in South Africa, where 202 inpatients with prostate cancer were compared with blood donors as well as with inpatients with a variety of other malignancies [18]. KSHV seroprevalence did not significantly differ in the case patients, compared with any of the control groups, with the exception of KS. Although the control subjects were not directly sampled from the population that gave rise to the case patients, it is difficult to explain in this instance why this would falsely underestimate an association, especially considering the many different control groups assessed. The first of 2 reports by Jenkins et al. was actually 2 separate studies [11]. In the first, in Trinidad and Tobago, patients with prostate cancer were all from Tobago, but their exact selection was not clearly described. The control group consisted of self-referred men from the greater Tobago community, a population of largely African descent. Case patients had a 2.24-fold significantly greater odds of KSHV seropositivity. Of note, age was the only factor formally controlled for. This is important, because a higher KSHV seroprevalence among the case patients could reflect more sexual exposure and perhaps a yet-to-be-discovered sexually transmitted agent of prostate cancer [8]. The second study had as its case patients men with prostate cancer at the University of Pittsburgh. Two control groups were used, but it was only in comparison to blood donors that case patients had a significantly higher KSHV seroprevalence. However, as the authors acknowledged, it is unlikely that blood donors are representative of the population that gave rise to the cases. The best designed study to date, by Korodi et al., used a nested case-control approach [19]. A cohort was assembled by virtue of serum collection among 20,243 men between 1968 and 1972, from which 165 cases of prostate cancer later arose as identified by the nationwide cancer registry. Each case patient was compared with age-matched control subjects formally selected from the underlying cohort by incidence density sampling. Another advantage of this design is that KSHV serostatus could be determined in blood drawn well before prostate cancer onset. That study found no evidence of an association between KSHV and prostate cancer. The most recent report, again by Jenkins et al. in this issue of the Journal, consisted of 3 separate case-controls studies [10]. The first 2 compared patients with prostate cancer with control subjects with benign prostatic hypertrophy (BPH); one study was in Italy and the other in the United States. Whether persons with BPH are representative of the population that gave rise to the cases of prostate cancer is in question. The ⩾19% KSHV seroprevalence seen among the control subjects with BPH from the United States, which is far higher than most general population estimates, substantiates this concern. The third study compared patients with prostate cancer ascertained in population-based cancer registries in the United States with population-based control subjects. Despite better control selection, this study, like the first 2 in the report, did not find evidence for an association between KSHV and prostate cancer. Thus, with the new studies by Jenkins et al. in hand, there is increasing clarity in the controversy regarding KSHV and prostate cancer. After discounting the blood donor control group in Hoffman et al. [11], only 1 of 7 studies showed an association. Hence, this picture demonstrates similarities to the last time KSHV was linked to a major malignancy (multiple myeloma), where an initial report generated much excitement but was subsequently refuted. These scenarios remind us how treacherous the road can be in epidemiologic work and how judgments cannot be made until many studies have been examined. Although the final verdict cannot be given for KSHV in prostate cancer, it is reasonable to conclude from current evidence that the virus is not the major culprit in prostate cancer in the United States and likely Europe as well. Research should now focus back on Tobago to attempt to substantiate the original finding [11], but now with careful attention to confounding, more specific nucleic acid-based measurement of KSHV infection, and perhaps a longitudinal design. Finally, just as the door begins to close for KSHV in prostate cancer, a new candidate viral etiologic agent has surfaced [20], but, as always, we should wait until all of the evidence has been scrutinized before drawing conclusions. I thank Dennis Osmond and Sheila Dollard for their helpful comments.