| |
Smoking Increases Risk for Cervical Cancer in HIV+
|
| |
| |
"Relationship between Smoking and Human Papillomavirus Infections in HIV-Infected and -Uninfected Women"
The Journal of Infectious Diseases 2004;189:1821-1828
Howard Minkoff,1 Joseph G. Feldman,2 Howard D. Strickler,3 D. Heather Watts,6 Melanie C. Bacon,7 Alexandra Levine,8 Joel M. Palefsky,9 Robert Burk,4 Mardge H. Cohen,10 and Kathryn Anastos5
1Departments of Obstetrics and Gynecology, Maimonides Medical Center, Brooklyn, and 2Department of Preventive Medicine, SUNY Downstate, and Departments of 3Epidemiology and Population Health and 4Pediatrics, Albert Einstein College of Medicine, and 5Department of Medicine, Montefiore Medical Center, and Albert Einstein College of Medicine, Bronx, New York; 6Pediatric, Adolescent and Maternal AIDS Branch, National Institute of Child Health and Human Development, Rockville, Maryland; 7Department of Medicine, Georgetown University, Washington, DC; 8Department of Internal Medicine, University of Southern California School of Medicine, Los Angeles, and 9Department of Medicine, University of California, San Francisco; 10Department of Internal Medicine, Cook County Hospital, Chicago, Illinois
"......Among HIV-infected women, smoking is associated with a significantly higher prevalence and incidence of HPV infection. Smoking during HIV infection may alter the natural history of HPV infection and increase the risk of cervical disease....."
ABSTRACT
Smoking may increase the risk of cervical cancer, a disease that is related to human papillomavirus (HPV) infection. However, the effects of smoking on the natural history of HPV are poorly understood, especially in women coinfected with human immunodeficiency virus (HIV).
HIV-infected (n = 1797) and HIV-uninfected (n = 496) women were assessed every 6 months for type-specific HPV DNA. Smoking status was self-reported. Covariates included age, parity, sexual behavior, HIV load, CD4+ T cell count, and antiretroviral therapy.
Smoking was positively associated with HPV prevalence at baseline in HIV-infected women (P = .002) and was significantly associated with type-specific HPV detection (e.g., type 18, odds ratio [OR], 2.45; 95% confidence interval [CI], 1.863.22). In Cox models, detection of HPV was significantly associated with smoking in HIV-infected women (relative hazard [RH], 1.33; 95% CI, 1.101.60; P = .003), but HPV persistence was not (RH, 0.97; 95% CI, 801.16; P = .72). The overall likelihood of acquiring persistent HPV was higher in smokers (OR, 1.39; 95% CI, 1.051.86; P = .023) because of greater incidence.
Among HIV-infected women, smoking is associated with a significantly higher prevalence and incidence of HPV infection. Smoking during HIV infection may alter the natural history of HPV infection and increase the risk of cervical disease.
INTRODUCTION
Human papillomavirus (HPV) infections in women are both common and clinically important. They have been linked to the development of cervical intraepithelial lesions and to invasive cervical cancer. The frequency of HPV infections and related lesions has been shown, by us and by others, to be particularly high among HIV-infected women.
HPV-related cervical diseases have also been linked to smoking, which, like HIV, may play a role in the development of cervical disease, in part through an effect on the immune response to HPV. A review published more than a decade ago cited 33 epidemiological studies of the association between smoking and cervical cancer. The majority of those works found such an association, and many of those that did not were said to have methodological flaws. The author of the review concluded that "the evidence would seem to support the conclusion that the association between cigarette smoking and cervical cancer is causal" (p. 955). The results of subsequent work have supported these conclusions. In a recent study, which found a link between smoking and cancer among women with oncogenic HPV infection at baseline, the authors concluded that "subsequent studies should examine the role of smoking in the multistage pathogenesis of cervical cancer" (p. 1406).
It is possible that the association between smoking and cancer is mediated by an effect on the incidence and/or persistence of HPV infection, which has been clearly connected to cervical cancer. Because smoking has been linked to alterations in both systemic and local immunity, either of these effects is possible. However, most studies that have examined the relationship between smoking and cervical cancer have not included an assessment of the pattern of carriage of HPV infections. Castle et al.acknowledged that a limitation of their study was "that there was only a single baseline measurement of smoking and HPV DNA positivity" (p. 1412). In addition, longitudinal studies of HPV that have assessed smoking status have tended to focus on HIV-uninfected women.
Given the high rate of smoking in many populations at risk for HIV- and HPV-related disease, as well as recent evidence linking the duration of HPV carriage with a risk for progression to cervical cancer, ascertaining the relationship between smoking and HPV infection has important public health consequences. If the relationship between smoking and cervical cancer is biological in nature, rather than due to uncontrolled residual confounding or other methodological research issues, it should be observed to have an effect on the natural history of HPV infection. Thus, prospective cohort studies of the effects of smoking on the prevalence, incidence, and duration of HPV infection are essential to prove a causal link between smoking and cervical cancer and to understand the mechanisms that likely underlie such an association. Here, we report the relationship between smoking and the carriage of HPV, using data from a longitudinal study of a large group of HIV-infected and at-risk HIV-uninfected women among whom smoking is common.
SUBJECTS
The Women's Interagency HIV Study (WIHS) is a prospective study of the natural history of HIV infection in women that enrolled 2059 HIV-seropositive and 569 at-risk HIV-seronegative participants at 6 clinical sites (Bronx and Brooklyn, NY; Chicago, IL; Los Angeles and San Francisco, CA; and Washington, DC) between October 1994 and November 1995. The recruitment methods and data collection procedures for WIHS have been described elsewhere. In brief, participants underwent visits every 6 months that included an interviewer-administered questionnaire, a physical examination, and the collection of specimens, including cervicovaginal lavage (CVL) for HPV DNA testing. Informed consent was obtained from all patients, and the human-experimentation guidelines of the authors' institutions were followed in the conduct of human research.
Twenty-five (4.4%) of 569 HIV-seronegative women and 71 (3.5%) of 2056 HIV-seropositive women did not have HPV testing performed. In addition, 48 HIV-seronegative women (8.4%) and 188 HIV-seropositive women (9.1%) had no follow-up and were excluded. Accordingly, the study population was composed of 1797 HIV-seropositive and 496 HIV-seronegative women who were tested for HPV infection during visits 15, from 3 October 1994 to 27 January 1998. At each of those visits, participants were asked about current smoking status; at the baseline visit, they were asked about their lifetime smoking history. Comparisons were made between HPV carriage patterns (defined below) of smokers and nonsmokers.
RESULTS
HPV data were available on each of the 5 visits for 40.9% of the women, on 4 visits for 24.3%, on 3 visits for 10%, on 2 visits for 8.9%, and on 1 visit for 15.8%. The distribution of the frequency of the HPV results was similar in the HIV-infected and HIV-uninfected groups, with data on at least 4 of the visits for 65% of each group. The 64 women for whom no data were available on follow-up visits were excluded from the analyses of incidence, and the 221 women for whom no data were available on >2 follow-up visits were excluded from the analyses of persistence. The women are categorized into (1) those who never smoked during follow-up (including those who had never smoked in their lifetime [N = 537] and (2) those who had smoked at some point before entering the study but not thereafter [N = 250]; no differences were seen between those 2 groups in any analyses, so they were combined [N = 787]), those who smoked occasionally (i.e., reported smoking during 1 or 2 visits; N = 282), and (3) consistent smokers (i.e., reported smoking during at least 3 visits; N = 1258). There was a higher percentage of HIV-infected women who never smoked. The age at first intercourse was almost 2 years later among those who had never smoked (P = .01), and the average number of sex partners during the preceding 5 years was lower, as was the average number of recent sex partners (P = .01). Among HIV-infected women, smokers had higher CD4+ cell counts and lower HIV loads at baseline P = .01). The number of women receiving HAART at baseline was <1%. HIV-uninfected women were significantly more likely to be smokers at baseline (62.8 vs. 55.9; P = .001), to acknowledge smoking at more of their visits (P = .01), and to be using oral contraceptives (8.8% vs. 4.6%; P = .01), and they were younger (age, 34.2 vs. 36.3 years; P = .01).
The prevalence of high-risk HPV was 28.4% in women who reported smoking at baseline and 25% in women who did not (odds ratio [OR], 1.23; 95% confidence interval [CI], 1.011.50; P = .05), and the prevalence of low-risk HPV types was 6.4% in smokers and 5.6% in nonsmokers (OR, 1.24; 95% CI, 0.861.78; P = .25). HIV-infected women were 3.9 times more likely (95% CI, 3.134.88; P < .001) than uninfected women to have a prevalent HPV infection at baseline. When the relationship between smoking at baseline and prevalence of HPV infection was examined separately among HIV-infected and -uninfected women, an association was found that was limited to HIV-infected women (P = .002). To take into consideration the repeated observations after baseline, using all available data, we then assessed the prevalence of HPV infection among the HIV-infected women at all 5 visits using a GEE analysis by specific HPV type, adjusted for age, oral contraceptive use, and CD4+ cell count and virus load at baseline. As discussed in Subjects, Materials, and Methods, the number of cigarettes smoked and pack years were not as highly associated with HPV occurrence as were the presence or absence of smoking at each visit (data not shown). Younger age, lower CD4+ cell count, and higher HIV RNA level were associated with an increased prevalence of almost all specific HPV types. The OR for smokers was elevated for each type of HPV for which there were adequate data to do the analysis. Thus, the OR for HPV type 16 was 1.44 (95% CI, 0.991.99; P = .047), and that for HPV type 18 was 2.45 (95% CI, 1.863.22; P = .001).
We next assessed the relationship between current smoking status (yes/no) and the incidence of HPV infection. Incident cases included any HPV types that were not present at baseline but were detected at subsequent visits. Overall, HIV-infected women were 3.13-fold more likely to acquire an incident HPV infection (95% CI, 1.745.63; P = .001), although there was no interaction between smoking and HIV infection and the risk of acquiring an HPV infection (P = .89). Table 4 shows the results of a Cox model that examined the time to first occurrence of an HPV type among HIV-uninfected and -infected women, with parity and age at first intercourse treated as fixed covariates and with smoking, number of sex partners during the preceding 6 months, and oral contraceptive use as time-updated variables. In the bottom section of table 4, the analysis was also adjusted for ART and for CD4+ cell counts and virus load as time-updated variables. Smoking was significantly associated with the incidence of HPV among HIV-infected women (relative hazard [RH], 1.33; 95% CI, 1.101.60; P = .003). Although the RH of smoking among HIV-uninfected women was similar (1.62; 95% CI, 0.882.96; P = .12), given the smaller sample of HIV-uninfected women in the study and the smaller number of incident infections detected in these women, the relationship between smoking and detection of incident HPV infection did not reach significance at traditional values. Women with higher HIV RNA levels were more likely to have an incident HPV infection (RH, 1.06; 95% CI, 1.011.16; P = .01), whereas women with higher CD4+ cell counts had a significantly reduced risk (RH, 0.90; 95% CI, 0.830.97; P = .008). Among HIV-infected women, the RH for incidence of specific HPV types was elevated among smokers for all types but only achieved significance when high-risk types were combined (RH, 1.27; 95% CI, 1.061.54; P = .01) and for HPV type 16 alone (RH, 1.81; 95% CI, 1.093.00; P = .02).
Because cervical neoplasia has been linked specifically to HPV infections detected more than once in serial assessments, we next assessed the time until clearance of HPV infections. Table 5 shows a Cox proportional hazards model for time to clearance, with age, HIV status, smoking status, number of sex partners, and oral contraceptive use as time-updated variables. None of these factors were linked to the time until clearance of HPV. However, the time until clearance alone does not determine the likelihood that a persistent virus (i.e., an incident virus that is present at 2 consecutive visits) will be detected. That likelihood is related to both the rate of acquisition and the time until clearancethat is, even if the time until clearance is equal in 2 groups, the group that acquires more viruses (both those that persist and those that do not) will have more persistent viruses present at any given time. Table 6 shows a regression analysis of the relationship of various factors to the detection of a persistent HPV infection. Both HIV-infected women relative to HIV-uninfected women (OR, 5.63; 95% CI, 3.389.35; P = .001) and consistent smokers relative to nonsmokers (OR, 1.39; 95% CI, 1.051.86; P = .02) were significantly more likely to acquire a persistent HPV infection. Age was inversely associated with risk (OR, 0.981; 95% CI, 0.9640.998; P = .03).
DISCUSSION
We have found that smoking is associated with a significantly higher incidence and prevalence of HPV infection among HIV-infected women. However, the time to clearance of HPV was not related to smoking status.
Persistent HPV infections have been linked to cervical cancer. The number of persistent infections that are found in a population is a reflection of both the rate of acquisition of the virus (incidence) and the likelihood that an incident infection will persist. Even if the proportion of HPV infections that persist does not vary between populations (i.e., there is an equal time to clearance), if there is a difference in incidence rate, at any given time there will be more individuals with a persistent virus in the population with the higher incidence. Thus, in addition to the findings noted above, we found that both smokers and HIV-infected women were significantly more likely than nonsmokers and HIV-uninfected women, respectively, to have persistent viruses detected during the course of the study. These results may help explain why these 2 groups (smokers and HIV-infected women) are at risk for cervical cancer, and they suggest that HIV-infected smokers might be at particularly high risk. Furthermore, the demographics of the WIHS participants reflect those of the HIV-infected population in the United States. At baseline, 55% of participants reported smoking, which highlights the potentially large attributable risk of smoking for cervical diseases among HIV-infected women and reinforces the Surgeon General's recent report that maintained that reducing smoking is a critical health priority for women.
Several of our findings regarding the patterns of carriage of HPV are consistent with and expand on the findings of other researchers. In a previous report on a smaller cohort, we noted that HIV-infected women were significantly more likely to acquire HPV than were HIV-uninfected women. Ahdieh et al. found that HIV-positive women were 1.8, 2.1, and 2.7 times more likely to have high-, intermediate-, and low-risk HPV infections, respectively, compared with HIV-negative women. In that analysis, persistence was more common when the subject had a lower CD4+ cell count. Thus, HIV infection and immunosuppression play important roles in modulating the natural history of HPV infection. Our data confirm those observations and suggest that smoking has similar effects on the incidence and prevalence of HPV infection.
Many studies have suggested that smoking may have adverse effects on immunity, both systemic and mucosal. Zeidel et al. reported that asymptomatic smokers exhibit impaired NK cell cytotoxic activity (NKA) in peripheral blood and an unbalanced systemic production of pro- and anti-inflammatory cytokines. Yovel et al. also reported, among both patients with schizophrenia and control subjects, that smokers and women showed lower NKA. Bjil et al. found that smoking was associated with increased Fas expression on peripheral blood lymphocytes in general and on B cells in particular; they concluded that this might render these cells more susceptible to apoptosis and contribute to the reduced humoral immune response observed in smokers. With regard to mucosal immunity, several studies have demonstrated that smoking may be associated with a decrease in the number of Langerhans cells, and squamous intraepithelial lesions have been reported to have particularly low numbers of Langerhans cells.
Thus, it is not surprising to find that smoking can influence the carriage of HPV. Of interest, although we did not find that smoking increased the persistence of HPV infection, there have been studies that have actually found that smoking "protects" from HPV. These studies differed from ours in several important ways. First, they were performed in HIV-uninfected populations, and some of the effect of smoking on HPV may be limited to women who are coinfected with HIV. Second, the women in our study were at high risk for HPV and had many more infections. Additionally, the women in the various studies may well have different frequencies of genetic polymorphisms that determine an individual's ability to metabolize nicotine and other metabolic by-products of smoke. This may result in differential effects of smoking among different populations. In any case, to our knowledge, our study is the first to assess the summary relative odds of a persistent HPV infection after considering the possible higher incidence of HPV infection among smokers.
The higher incidence of HPV infection that we noted among smokers may represent one of several phenomena. As was noted by Ahdieh et al., incident infections might represent higher rates of exposure, an increased likelihood of reactivation of latent infections, or an increased susceptibility to infection. The first explanation seems unlikely, because we controlled for numbers of sex partners during both the preceding 5 years and the preceding 6 months. Also, the other groups that we found to have relatively high incidence rates were women with more-advanced HIV disease (lower CD4+ cell counts). It is counterintuitive to think that those women who had fewer sex partners and fewer pregnancies (data not shown) had more frequent sexual contact. Rather, it suggests that immune compromise may contribute to higher incidence rates. Although we must acknowledge that residual confounding is a concern, we believe that either of Ahdieh et al.'s other 2 explanationsan increased likelihood of reactivation of latent infections or an increased susceptibility to infectionseems more reasonable. Either would reflect an effect of smoking on the host's immune system that facilitates the establishment of a new infection or interferes specifically with the host's ability to maintain an established infection at an undetectable level.
A few limitations of our findings should be noted. First, no objective measure of smoking, such as serum cotinine levels, was used. However, we did perform repeated interviews and, more importantly, it is most likely that misclassification would have been in the direction of understating the frequency of smoking, a direction that would have biased toward the null. Similarly, as noted above, we did not assess the pharmacogenetics of the population, and some portion of the population might be relatively less affected by the toxic effects of smoke. Again, by including those women in the analyses, we would have biased toward the null. Thus, to the degree that these additional considerations might be important, they indicate that our results more likely underestimate, rather than overestimate, the effects of smoking on HPV infection. Finally, we were unable to determine whether incident cases represent new acquisitions of virus or reemergence of preexisting virus. However, the clinical consequence of the repeated detection of a virus over time may be independent of which phenomenon is being observed. Previous studies that linked the persistence of HPV to cancer did not distinguish between acquisition and reemergence but did demonstrate that the detection of virus over time was clearly associated with the development of cervical neoplasia.
In summary, we have found that smoking is associated with an increased prevalence and incidence of HPV among HIV-infected women. This suggests that smoking and HIV infections, in conjunction, may alter the natural history of HPV infection and thereby increase the risk of cervical disease. Finally, these data add to the litany of reasons for making smoking cessation a cornerstone of women's health.
|
|
| |
| |
|
|
|
