Risk factors for human herpesvirus 8 seropositivity in the AIDS Cancer Cohort Study

Results Among non-KS subjects, HHV-8 seropositivity was 6%, 13% and 29% among women, heterosexual men and MSM, respectively. HHV-8 seropositivity was decreased in heavier (≥1/2 pack/day) compared to lighter smokers among women (5% versus 8%; adjusted OR (aOR) 0.4; 95% CI 0.2–0.8) and MSM (27% versus 32%; aOR 0.7; 95% CI 0.6–1.0), but not among heterosexual men (12% versus 16%; aOR 0.7; 95% CI 0.4–1.2). HHV-8 seroprevalence was increased in heavier (≥1 drink/day) compared to lighter consumers of alcohol among women (16% versus 4%; adjusted OR 5.2; 95% CI 2.3–12), but not among MSM (33% versus 28%; aOR 1.2; 95% CI 0.9–1.6) or heterosexual men (13% versus 13%; aOR 1.1; 95% CI 0.6–2.0). In analyses adjusted for smoking and drinking, HHV-8 seropositivity was positively associated with chlamydia infection (OR = 4.3; 95% CI 1.2–13) and with marital status among women p heterogeneity = 0.03, and with hepatitis (OR = 1.6; 95% CI 1.2–2.1), gonorrhea (OR = 1.5; 95% CI 1.1–1.9), genital warts (OR = 1.5; 95% CI 1.1–2.0) and nitrate inhalant use (OR = 1.7; 95% CI 1.3–2.3) among MSM. Conclusions Inverse association of HHV-8 seropositivity with cigarette smoking may indicate protective effect of tobacco smoke on HHV-8 infection, whereas positive associations with alcohol may reflect either behavioral factors or biological effects modulating susceptibility. Smoking and drinking may influence KS risk, at least in part, by altering the natural history of HHV-8 infection. Keywords Kaposi sarcoma Transmission Men who have sex with men Injection drug use United States 1 Introduction Human herpesvirus 8 (HHV-8, also called Kaposi sarcoma-associated herpesvirus) is accepted as the infectious cause of Kaposi sarcoma (KS) ( Boshoff and Weiss, 2001; Chang et al., 1994 ), the most common tumor among persons with AIDS ( Frisch et al., 2001; Mbulaiteye et al., 2003b ). HHV-8 seroprevalence is low (∼3%) in the United States ( Pellett et al., 2003 ) except among men who have sex with men (MSM; ∼30%) who also have a high risk of AIDS-related KS ( Martin et al., 1998 ). Seroprevalence is intermediate (5–10%) among HIV-positive heterosexual men, women and injection drug users ( Lennette et al., 1996; Martin, 2003 ). HHV-8 seropositivity is variably associated, particularly among MSM, with sexual exposures, sexually transmitted disease ( Martin et al., 1998 ), and injection and other drug use in some but not all studies ( Renwick et al., 2002 ). Cigarette smoking has been associated with decreased risks for AIDS-related KS in the U.S. ( Hoover et al., 1993; Nawar et al., 2005 ) and classical KS in Italy ( Goedert et al., 2002 ), suggesting that cigarette smoking may modulate the natural history of HHV-8 infection. However, findings were equivocal in two studies conducted in Uganda, where KS is more common but smoking is less prevalent ( Ziegler et al., 2003, 1997 ). Whether cigarette smoking decreases the risk of HHV-8 infection is unknown. We therefore evaluated the association of various behavioral factors, including cigarette smoking, with HHV-8 seropositivity among persons with AIDS in the U.S. 2 Methods 2.1 Study population and serology methods We studied 2795 patients with AIDS aged ≥18 years old participating in the National Cancer Institute's AIDS Cancer Cohort (NCI-ACC) study. The patients were enrolled at 24 AIDS treatment and clinical trial sites in the United States from October 1997 to January 2000. All patients met Centers the Disease Control and Prevention (CDC) criteria for AIDS diagnosis (1992) . At enrollment, the median (inter-quartile range) CD4 lymphocyte count was 204 (82–281) cells/μL and HIV viral load was 32,759 (3315–156,084) copies/μL. Interviewers used a Computer Assisted Personal Interview (CAPI) to obtain information on age, income, education, sexual behaviors, past medical history, lifetime use of cigarettes, alcoholic drink consumption in the past 12 months and use of recreational drugs in the past 12 months. Participants also gave a venous blood sample for HHV-8 testing. We tested for anti-HHV-8 antibodies using an enzyme-linked immunosorbent assay directed against the K8.1 glycoprotein antigen, as previously described ( Engels et al., 2000; Mbulaiteye et al., 2003a ). Ethical approval to conduct the study was granted by institutional review boards at the National Cancer Institute and at collaborating institutions. 2.2 Statistical methods We performed analyses of HHV-8 seropositivity separately for women ( n = 514), heterosexual men ( n = 660) and MSM ( n = 1621). Eight additional subjects with inadequate samples ( n = 7) or incomplete data ( n = 1) were excluded. Men were classified as MSM if they reported having ever had sexual contact with men and as heterosexual otherwise. Subjects with KS at or before enrollment were excluded from analyses evaluating associations with asymptomatic HHV-8 seropositivity. We determined the crude and adjusted odds ratios (OR) of associations with HHV-8 seropositivity and associated 95% confidence intervals using logistic regression models. We specifically evaluated the relationship between cigarette smoking and HHV-8 seropositivity because prior studies have indicated an inverse association between cigarette smoking and KS ( Hoover et al., 1993; Nawar et al., 2005 ). However, because cigarette smoking and alcohol consumption tend to track together, we examined the association between these variables among HHV-8 seronegative subjects using frequency tables to determine if they fulfilled the definition of a classical confounder (i.e., associated with each other and also with the disease outcome, in this case, HHV-8 seropositivity) ( Hauck et al., 1991 ). In our data, cigarette smoking and alcohol consumption were associated. Thus, we controlled for alcohol consumption in the models estimating the independent association of cigarette smoking with HHV-8 seropositivity. Our primary multivariable model included cigarette smoking and alcohol consumption. To identify other variables that were significant predictors of HHV-8 seropositivity, we added all variables associated with HHV-8 seropositivity at p < 0.10 to the multivariable logistic models and determined the independent contribution of each individual variable to the full model using a stepwise procedure, with a p ≤ 0.05 based on the likelihood ratio test used as the stay or enter criteria. Because our analysis was primarily for hypothesis generation, we did not adjust for multiple comparisons. 3 Results One hundred thirty two (4.7%) of the subjects had KS at or before enrollment, including 2 women, 6 heterosexual men and 124 MSM ( Table 1 ). One hundred nine (83%) of the subjects with KS had HHV-8 antibodies, including both of the women, three of the six heterosexual men and 104 of the 124 MSM. Associations with asymptomatic HHV-8 seropositivity were evaluated among the 2663 subjects without KS, of whom 554 (21%) had HHV-8 antibodies. 3.1 HHV-8 seropositivity among women HHV-8 antibodies were detected among 6% of the women without KS. In univariate analyses, seropositivity was not significantly associated with age, race or income ( Table 2 ). Women who had never married were more likely to be HHV-8 seropositive as compared to women who were divorced, separated or widowed. Women with less than 12 years of education were more likely to be seropositive compared to women who had some college level education. A history of chlamydia infection, but not of other sexually transmitted diseases, was marginally associated with HHV-8 seropositivity. Use of crack cocaine or heroin within the past 12 months, but not of nitrate inhalants or injection drugs, was significantly associated with seropositivity ( Table 2 ). Slightly more than two-thirds of the women had ever smoked cigarettes and 62% had consumed alcohol in the past 12 months. Among HHV-8 seronegative women, those who had ever smoked cigarettes were more likely than those who had not to report alcohol consumption (65% versus 48%, p < 0.001). In univariate analyses, HHV-8 seropositivity was lower among women who smoked more than one-half pack of cigarettes per day compared to women who smoked less (5% versus 8%, p = 0.13; Table 2 ). Conversely, HHV-8 seropositivity was higher among women who consumed more than one alcoholic drink per day in the past 12 months compared to women who drank less (16% versus 4%, p < 0.001). Inverse HHV-8 associations with smoking became statistically significant in analyses that adjusted for the effects of alcohol consumption ( p = 0.02; Table 3 ). Likewise, the positive HHV-8 associations with alcohol consumption became accentuated by adjustment for the effects of smoking ( p < 0.001). In analyses that adjusted for effects of smoking and alcohol consumption, chlamydia infection in the past 12 months was significantly associated with HHV-8 seropositivity (OR = 4.3; 95% CI 1.2–13). Women who were divorced, separated or widowed were less likely to be HHV-8 seropositive as compared to never married women ( Table 3 ). 3.2 HHV-8 seropositivity among heterosexual men HHV-8 antibodies were detected among 13% of heterosexual men without KS, twice the prevalence among women ( p < 0.001). In univariate analyses, HHV-8 seropositivity was two-fold higher among heterosexual men aged 50 years or older compared to prevalence in younger men. HHV-8 seropositivity was unrelated to race, income or history of sexually transmitted diseases. Compared to heterosexual men who had never been married, those who were divorced, separated or widowed had a marginally elevated prevalence of HHV-8 seropositivity ( p = 0.12; Table 2 ). Eighty-three percent of heterosexual men had ever smoked cigarettes and 67% had consumed alcohol in the past year. Among HHV-8 seronegative heterosexual men, those who had ever smoked cigarettes were marginally more likely than those who had not to report alcohol consumption (68% versus 59%, p < 0.10). HHV-8 seropositivity was slightly decreased among smokers (12% versus 16%, p = 0.19); however, that association was not statistically significant, even with adjustment for alcohol drinking ( p = 0.18). HHV-8 seropositivity was unrelated to alcohol consumption and was slightly decreased among men reporting sexual activity in the past 12 months ( Table 2 ), but the latter difference was not significant in adjusted models. 3.3 HHV-8 seropositivity among MSM HHV-8 seropositivity was 29% among MSM without KS, more than double the prevalence among heterosexual men and more than four times the prevalence among women (all p < 0.001). In univariate analyses, seropositivity among MSM was positively associated with older age, Hispanic ethnicity, higher educational attainment, sexual activity in the past 12 months and history of hepatitis, syphilis, gonorrhea and genital warts ( Table 2 ). The use of nitrate inhalants and injection drugs, but not crack cocaine or heroin, in the past 12 months was significantly associated with HHV-8 seropositivity ( Table 2 ). Seventy-one percent of MSM had ever smoked cigarettes and 80% had consumed alcohol in the 12 months. Among HHV-8 seronegative MSM, those who had ever smoked cigarettes reported similar frequency of alcohol consumption as those who had never smoked cigarettes (79% versus 75%, p < 0.19). HHV-8 seropositivity was significantly decreased among MSM who were heavier smokers compared to MSM who were lighter smokers (27% versus 32%; p = 0.03, Table 2 ). However, the prevalence of HHV-8 seropositivity among MSM did not significantly differ according to alcohol consumption (33% versus 28%; p = 0.1; Table 2 ). Nonetheless, in analyses adjusting the effects of one for the other, the statistical significance of the HHV-8 associations with cigarette smoking was accentuated ( p = 0.02). In multivariable models, HHV-8 seropositivity was significantly associated with smoking, history of hepatitis, gonorrhea, genital warts and nitrate inhalant use, but not with sexual activity, injection drug, heroin, crack cocaine use or alcohol consumption ( Table 3 ). HHV-8 seropositivity among MSM light and heavier smokers with KS was similar (84% versus 83%). 4 Discussion In the NCI-ACC study, HHV-8 seropositivity was inversely associated with cigarette smoking and positively associated with alcohol consumption among women and MSM. The associations, particularly among women, were independent of sexual and recreational drug exposures. The magnitudes of the associations between cigarette smoking and alcohol consumption with HHV-8 seropositivity were accentuated in analyses that adjusted one for the other, indicating negative confounding. The associations were qualitatively similar among heterosexual men, but not statistically significant, perhaps because of confounding with unreported homosexual sexual activity. The inverse association between cigarette smoking and HHV-8 seropositivity is interesting because inverse associations have been previously reported between cigarette smoking and AIDS-related KS in the U.S. ( Hoover et al., 1993; Nawar et al., 2005 ) and classical KS in Italy ( Goedert et al., 2002 ). In the U.S. Multicenter AIDS Cohort study, persons who smoked at least one-half pack of cigarettes per day experienced half the incidence of AIDS-related KS as compared to that in non-smokers ( Hoover et al., 1993 ). That study was unable to distinguish the effects of cigarette smoking on the risk of HHV-8 infection from those on KS risk in HHV-8 infected individuals. In the NCI-ACC study ( Nawar et al., 2005 ), Nawar et al. observed an inverse association between KS and cigarette smoking in an analysis restricted to HHV-8 seropositive MSM with AIDS. In Italy, patients with classical KS were one-fourth as likely to be current or former smokers as compared to HHV-8 seropositive controls ( Goedert et al., 2002 ). Together, these studies suggest that smoking may decrease the risk of KS among HHV-8 seropositive persons, perhaps through action of components of tobacco smoke. However, two studies conducted in Uganda, where KS is relatively common ( Ziegler et al., 2003, 1997 ) produced equivocal results. In one, an inverse association between endemic KS and current or prior smoking was reported in a case–control analysis of HIV-seronegative subjects with cancer, but those results were not statistically significant ( Ziegler et al., 2003 ). In the other, smoking was unrelated to AIDS-related KS ( Ziegler et al., 1997 ). The results from these studies two studies are inconclusive because smoking-associated cancers were not excluded from the analyses and because the prevalence of smoking is low in Uganda (approximately 20% among controls and cases), which reduced their statistical power. Our findings of inverse associations between HHV-8 seropositivity and cigarette smoking are intriguing. Similar findings have been reported by Baeten et al. who observed a lower HHV-8 seroprevalence among Kenyan truck drivers who smoked compared to those who did not ( Baeten et al., 2002 ). The observation of similar findings among the people who have different AIDS- and, most likely, HHV-8-risk profiles, may be interpreted as indicative of a specific biologic effect of smoking on HHV-8 seropositivity. Although smoking may decrease HHV-8 seropositivity by masking the humoral immune response to HHV-8, this explanation is not supported by our observation that HHV-8 seroprevalence was similar among MSM with KS who did and did not smoke. Instead, we speculate that components in tobacco smoke could modulate HHV-8 infection through their effects on leucocytosis ( Smith et al., 2003 ), secretion of oxygen radicals ( Sopori, 2002 ) and T helper 1 (T H 1)/T H 2 cytokine balance ( Leow and Maibach, 1998 ). Our finding of a positive association between alcohol consumption and HHV-8 seropositivity among the women is interesting. A similar finding was reported in a study of female commercial sex workers in Kenya, where current or past consumers of alcohol had significantly higher HHV-8 seropositivity compared to alcohol abstainers ( Lavreys et al., 2003 ). However, HHV-8 seropositivity was unrelated to alcohol consumption in the study of Kenyan truck drivers ( Baeten et al., 2002 ) and in the study of Ugandan HIV-negative cancer patients ( Ziegler et al., 2003 ), suggesting that this relationship is complex. Because alcohol consumption is associated with diverse adverse biological effects, e.g., on lymphocyte sub-populations, T H 1/T H -2 cytokine balance and nutrition ( Isaki and Kresina, 2000 ), it is conceivable that alcohol consumption could increase the risk of HHV-8 infection. Moreover, the potent behavioral modification associated with alcohol consumption may also influence the probability of exposure to HHV-8. The strengths of our study include access to a well-characterized cohort of AIDS patients who have diverse risk factors for HIV and HHV-8 infection. This allowed us to compare and contrast the epidemiology of HHV-8 in different populations. Our findings among women are particularly informative because they would not be confounded by unreported male homosexual exposures. Furthermore, because we had detailed drug and sexual exposure data, we were able to control for the effects of these exposures when estimating HHV-8 associations with smoking and alcohol consumption. Finally, recall bias is unlikely to explain our findings because all participants had AIDS at enrollment and were unaware of their HHV-8 serostatus. Nonetheless, some limitations should be considered. Available HHV-8 serological assays are imperfect ( Rabkin et al., 1998 ). In our study, HHV-8 serology identified less than 100% of subjects with KS, a group thought to be universally HHV-8 infected, suggesting misclassification bias remains possible. Even so, misclassification of serostatus among subjects without KS would likely have been non-differential, which would bias our results toward the null. We performed multiple statistical comparisons; readers should interpret our results with caution. Other limitations include using cross-sectional data, which does not allow temporality to be determined. Moreover, we relied on self-reported behavioral data that likely over-simplified complex patterns of behavior, which tend to vary with time. Finally, it is possible that smoking exposure status may have been misclassified for participants who modified this behavior following the diagnosis of AIDS or KS. To conclude, our findings suggest that cigarette smoking is associated with lower risk and alcohol consumption with higher risk of HHV-8 infection. Identification of components in cigarette smoke protective against HHV-8 infection and/or KS may provide insight into pathogenesis and, potentially, novel approaches for prevention. Confirmation of HHV-8 associations with alcohol consumption would justify public health messages informing people with AIDS about this potentially serious adverse effect. Acknowledgements We are grateful to the study subjects and the staff at the clinics for participating. We also thank Tabitha Hendershot, Mary-Anne Ardini and Athan Gadonas for study management (Research Triangle Institute (RTI), Washington DC), Sylvia Cohn (RTI, Washington DC) for preparing the data files and to Christine Gamache (Science Applications International Corporation-Frederick) for running the serology assays. The Study was funded in part by Contracts N01-CP-81017 and N01-CO-12400 from the Intramural Research Program of the National Cancer Institute, Grant AI-25868-15 from the National Institute of Allergy and Infectious Diseases AIDS Clinical Trials Group and Grants RR00046 and MO1 RR00046-41S1 from the General Clinical Research Center Program of the Division of Research Resources, National Institutes of Health; Department of Health and Human Services. Appendix A Collaborators of the AIDS Cancer Cohort (ACC) Study were: Peter Frame, M.D., University of Cincinnati Medical Center, Cincinnati (Ohio); William G. Powderly, M.D., Washington University ACTU, St. Louis (Missouri); Michael Lederman, M.D., Case Western Reserve University, Cleveland (Ohio); Robert Kalayjian, M.D., MetroHealth Medical Center, Cleveland (Ohio); Denise Signs, M.D., Summa Health Systems, Akron (Ohio); Richard Reichman, M.D., Strong Memorial Hospital, Rochester (New York); Charles van der Horst, M.D., Chapel Hill (North Carolina); Martin Hirsch, M.D., Massachusetts General Hospital, Boston (Massachusetts); Kathleen Squires, M.D., University of Alabama, Birmingham (Alabama); Mitchell Goldman, M.D., University Hospital, Indianapolis (Indiana); James A. Thommes, M.D., Pacific Oaks Medical Group, Beverley Hills (California); Edwin DeJesus, M.D., IDC Research, Altamonte Springs (Florida); Christopher Lahart, M.D., Thomas Street Clinic and Houston Veterans’ Affairs Medical Center, Houston (Texas); G. Stephen Bowen, M.D., HIV Clinical Research, Ft. Lauderdale (Florida); Cynthia Gibert, M.D., Veterans’ Affairs Medical Center, Washington, D.C.; Michael Rigsby, M.D., Yale University, New Haven (Connecticut); Michael Sampson, M.D., AIDS Healthcare Foundation, Los Angeles (California); David Butcher, M.D., Chase Brexton Health Services, Baltimore (Maryland); Gervais Frechette, M.D., Grossman Clinic, New York (New York); Susan Krown, M.D., Memorial Sloan-Kettering Cancer Center, New York (New York); Juan Lertora, M.D., Ph.D, Tulane University Medical Center, New Orleans (Louisiana); Judith L. Neidig, RN, PhD., Ohio State University Medical Center, Columbus (Ohio). References AIDS diagnosis, 1992 AIDS diagnosis 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults MMWR Recomm Rep 41 1992 1 19 Baeten et al., 2002 J.M. Baeten B.H. Chohan L. Lavreys J.P. Rakwar R. Ashley B.A. Richardson Correlates of human herpesvirus 8 seropositivity among heterosexual men in Kenya AIDS 16 2002 2073 2078 Boshoff and Weiss, 2001 C. Boshoff R.A. Weiss Epidemiology and pathogenesis of Kaposi's sarcoma-associated herpesvirus Philos Trans R Soc Lond B Biol Sci 356 2001 517 534 Chang et al., 1994 Y. Chang E. Cesarman M.S. Pessin F. Lee J. Culpepper D.M. Knowles Identification of herpesvirus-like DNA sequences in AIDS-associated Kaposi's sarcoma Science 266 1994 1865 1869 Engels et al., 2000 E.A. Engels D. Whitby P.B. Goebel A. Stossel D. Waters A. Pintus Identifying human herpesvirus 8 infection: performance characteristics of serologic assays J Acquir Immune Defic Syndr 23 2000 346 354 Frisch et al., 2001 M. Frisch R.J. Biggar E.A. Engels J.J. Goedert Association of cancer with AIDS-related immunosuppression in adults JAMA 285 2001 1736 1745 Goedert et al., 2002 J.J. Goedert F. Vitale C. Lauria D. Serraino M. Tamburini M. Montella Risk factors for classical Kaposi's sarcoma J Natl Cancer Inst 94 2002 1712 1718 Hauck et al., 1991 W.W. Hauck J.M. Neuhaus J.D. Kalbfleisch S. Anderson A consequence of omitted covariates when estimating odds ratios J Clin Epidemiol 44 1991 77 81 Hoover et al., 1993 D.R. Hoover C. Black L.P. Jacobson O. Martinez-Maza D. Seminara A. Saah Epidemiologic analysis of Kaposi's sarcoma as an early and later AIDS outcome in homosexual men Am J Epidemiol 138 1993 266 278 Isaki and Kresina, 2000 L. Isaki T.F. Kresina Directions for biomedical research in alcohol and HIV: where are we now and where can we go? AIDS Res Hum Retroviruses 16 2000 1197 1207 Lavreys et al., 2003 L. Lavreys B. Chohan R. Ashley B.A. Richardson L. Corey K. Mandaliya Human herpesvirus 8: seroprevalence and correlates in prostitutes in Mombasa, Kenya J Infect Dis 187 2003 359 363 Lennette et al., 1996 E.T. Lennette D.J. Blackbourn J.A. Levy Antibodies to human herpesvirus type 8 in the general population and in Kaposi's sarcoma patients Lancet 348 1996 858 861 Leow and Maibach, 1998 Y.H. Leow H.I. Maibach Cigarette smoking, cutaneous vasculature, and tissue oxygen Clin Dermatol 16 1998 579 584 Martin, 2003 J.N. Martin Diagnosis and epidemiology of human herpesvirus 8 infection Semin Hematol 40 2003 133 142 Martin et al., 1998 J.N. Martin D.E. Ganem D.H. Osmond K.A. Page-Shafer D. Macrae D.H. Kedes Sexual transmission and the natural history of human herpesvirus 8 infection N Engl J Med 338 1998 948 954 Mbulaiteye et al., 2003a S.M. Mbulaiteye R.J. Biggar P.M. Bakaki R.M. Pfeiffer D. Whitby A.M. Owor Human herpesvirus 8 infection and transfusion history in children with sickle-cell disease in Uganda J Natl Cancer Inst 95 2003 1330 1335 Mbulaiteye et al., 2003b S.M. Mbulaiteye D.M. Parkin C.S. Rabkin Epidemiology of AIDS-related malignancies: an international perspective Hematol Oncol Clin North Am 17 2003 673 696 Nawar et al., 2005 E. Nawar S.M. Mbulaiteye J.E. Gallant D.A. Wohl M. Ardini T. Hendershot Risk factors for Kaposi's sarcoma among HHV-8 seropositive homosexual men with AIDS Int J Cancer 115 2005 296 300 Pellett et al., 2003 P.E. Pellett D.J. Wright E.A. Engels D.V. Ablashi S.C. Dollard B. Forghani Multicenter comparison of serologic assays and estimation of human herpesvirus 8 seroprevalence among US blood donors Transfusion 43 2003 1260 1268 Rabkin et al., 1998 C.S. Rabkin T.F. Schulz D. Whitby E.T. Lennette L.I. Magpantay L. Chatlynne Interassay correlation of human herpesvirus 8 serologic tests. HHV-8 Interlaboratory Collaborative Group J Infect Dis 178 1998 304 309 Renwick et al., 2002 N. Renwick N.H. Dukers G.J. Weverling J.A. Sheldon T.F. Schulz M. Prins Risk factors for human herpesvirus 8 infection in a cohort of drug users in the Netherlands, 1985–1996 J Infect Dis 185 2002 1808 1812 Smith et al., 2003 M.R. Smith A.L. Kinmonth R.N. Luben S. Bingham N.E. Day N.J. Wareham Smoking status and differential white cell count in men and women in the EPIC-Norfolk population Atherosclerosis 169 2003 331 337 Sopori, 2002 M. Sopori Effects of cigarette smoke on the immune system Nat Rev Immunol 2 2002 372 377 Ziegler et al., 2003 J. Ziegler R. Newton D. Bourboulia D. Casabonne V. Beral E. Mbidde Risk factors for Kaposi's sarcoma: a case–control study of HIV-seronegative people in Uganda Int J Cancer 103 2003 233 240 Ziegler et al., 1997 J.L. Ziegler R. Newton E. Katongole-Mbidde S. Mbulataiye K. De Cock H. Wabinga Risk factors for Kaposi's sarcoma in HIV-positive subjects in Uganda Aids 11 1997 1619 1626