Introduction
Lung cancer is the third most common cancer among HIV-infected individuals in the USA, with incidence exceeded only by the two major AIDS-defining cancers, Kaposi sarcoma (KS) and non-Hodgkin lymphoma (NHL) [1]. Lung cancer risk has been estimated to be two to seven times higher among HIV-infected individuals than in the general population [2-11]. Survival following lung cancer is extremely poor in general, and prognosis is especially poor in HIV-infected individuals [12].
Because 60-80% of HIV-infected individuals in the USA are smokers [13], a high prevalence of smoking contributes to the excess lung cancer risk in this population. Nonetheless, it is uncertain whether other co-factors are involved in the etiology of lung cancer among HIV-infected individuals. Previous studies have attempted to account for differences in smoking behavior between HIV-infected individuals and the general population, either with directly measured smoking behavior data [14,15] or by using hypothetical smoking scenarios [11,13,16]. However, these studies included only a small number of lung cancer cases and produced contradictory results. Consequently, it remains unclear whether smoking entirely explains the substantial excess of lung cancer in HIV-infected individuals. Additionally, there are only limited data regarding the relationship between HIV-induced immunosuppression (as measured by CD4 cell counts) and lung cancer risk [17].
We used data from the HIV/AIDS Cancer Match Study to characterize patterns in lung cancer risk among people with AIDS (PWA) in the USA. The present study is the largest population-based evaluation of lung cancer risk among PWA to date. The degree to which smoking could account for the observed excess lung cancer risk in PWA was investigated by comparing lung cancer incidence in PWA to that predicted under various scenarios of prevalence, duration, and intensity of smoking. The relationship between immunosuppression and lung cancer risk was assessed using data on CD4 cell counts measured at AIDS onset.
Materials and methods
Study population and identification of lung cancers
During 2002-2005, AIDS surveillance data on N = 397 927 PWA was linked to population-based cancer registry data in six US states (Connecticut, Florida, Georgia, Massachusetts, Michigan, and New Jersey) and five metropolitan areas (Los Angeles, New York City, San Diego, San Francisco, and Seattle). The study was approved by institutional review boards at each registry. The present analyses included adolescent and adult PWA (aged 15 years or older at AIDS onset) diagnosed between 1980 and 2002. Data from cancer registries were used to identify lung cancer cases in the period spanning 60 months before to 60 months after AIDS onset, as described previously [4]. Evaluation of this entire 10-year period provides an overall estimate of lung cancer risk. However, within this period, the analyses were specifically focused on the 4-27-month period after AIDS onset to avoid potential biases arising from evaluation of prevalent lung cancers (60 to 7 months before AIDS), over-ascertainment at the time of AIDS onset (6 months before to 3 months after AIDS), and late losses to follow-up (28 to 60 months after AIDS) [4].
Follow-up for most analyses thus started at 4 months after AIDS onset and ended at the earliest of lung cancer, death, end of cancer registry coverage, or 28 months after AIDS onset. Lung cancers arising in this 2-year period were considered incident and were included in analyses. As the CD4 cell counts were largely unavailable before 1990, only people with AIDS onset in 1990-2002 were included for CD4 cell analyses. For individuals with AIDS onset since 1990 (N = 189 473), CD4 cell counts in the 6 months before to 3 months after AIDS were available from the AIDS registries for N = 122 130 (64.5%).
Pulmonary occurrences of KS and NHL, and cancers with non-specific International Classification of Diseases for Oncology (ICD-O3) codes 8000 to 8005, were not considered as lung cancers. Using ICD-O3 codes, lung cancer histologic subtypes were classified into squamous cell carcinoma, adenocarcinoma, bronchioloalveolar carcinoma, small cell carcinoma, large cell carcinoma, and other histologies [18].
Statistical analyses
Expected numbers of lung cancers were calculated by applying registry, age, race, sex, and calendar-year-specific incidence rates from the general population to the person-time accrued among PWA. For periods before AIDS onset, expected rates were adjusted to account for mortality following a cancer diagnosis [4]. Standardized incidence ratios (SIRs) were calculated as the ratio of observed to expected number of cases, and 95% Poisson confidence intervals (CIs) around SIRs were derived using an exact method. After excluding the AIDS onset period (6 months before to 3 months after AIDS), the trend in SIRs across AIDS-relative time (60 to 25 months before AIDS, 24 to 7 months before AIDS, 4 to 27 months after AIDS, and 28 to 60 months after AIDS) was evaluated using mid-points of the respective intervals (-42.5, -15.5, 15.5, and 44.0 months) in a Poisson regression model [4]. Evidence for a linear trend in lung cancer incidence across CD4 cell count categories, after adjustment for age, sex, race, calendar period of AIDS, and HIV risk group, was assessed using Poisson regression.
Actual smoking behavior data were unavailable for PWA in this study. Thus, to investigate whether the increased lung cancer risk among PWA could be explained by smoking, the lung cancer incidence among PWA was compared with rates predicted under plausible hypothetical smoking scenarios. Specifically, the following models, developed by Flanders et al. and based on data from the Cancer Prevention Study II (CPS II) [19], which relate lung cancer mortality among 40 to 70 year-old current smokers to smoking duration (in years) and intensity (cigarettes smoked per day), were used:
Men aged 40-49 years: mortality = e-17.9 × duration1.9 × intensity0.95
Men aged 50-59 years: mortality = e-17.4 × duration2.6 × intensity0.52
Men aged 60-69 years: mortality = e-15.7 × duration2.4 × intensity0.37
Women aged 40-49 years: mortality = e-20.2 × duration2.8 × intensity0.96
Women aged 50-59 years: mortality = e-17.2 × duration2.2 × intensity0.75
Women aged 60-69 years: mortality = e-14.1 × duration1.5 × intensity0.78
As these models predict lung cancer mortality, they were modified to predict incidence by applying age and sex-specific incidence-mortality correction factors that were derived from incidence and mortality data from the Surveillance, Epidemiology and End Results (SEER) program [20]. The ability of these modified Flanders models to predict New Jersey (a non-SEER registry) lung cancer incidence rates was then confirmed (data not shown, details available from authors upon request).
To calculate predicted lung cancer rates among PWA under plausible smoking scenarios, two estimates were used for the proportion of current smokers among PWA (60 and 80%), assuming that the remainder were non-smokers and not at risk of lung cancer. For the proportion of PWA who were current smokers, various possible durations of smoking (based on smoking initiation at age 10, 15, and 20 years) and intensities of smoking (10, 20, 30, and 40 cigarettes per day) were considered. Using the modified Flanders models, age and sex-specific predicted lung cancer incidence rates were calculated for the 12 combinations of smoking duration and intensity (three combinations of duration × four combinations of intensity). Assuming that each of the 12 smoking duration and intensity combinations was equally common, the overall predicted lung cancer rates among PWA were calculated as the average of the 12 separate predicted rates. Observed lung cancer incidence among PWA was then compared with these averaged predicted rates by calculating an observed to predicted ratio and an exact 95% Poisson CI around the ratio.
Flanders et al. [19] developed the models based on data obtained predominantly from whites (93% whites in CPS II) [19]. As the lung cancer risk attributable to smoking may differ by race [21], a sensitivity analysis was also performed by repeating the above analyses restricted to white PWA.
Results
Table 1 describes the demographic characteristics of 397 927 PWA across three periods of AIDS diagnosis (1980-1989, when no antiretroviral therapy or monotherapy was available; 1990-1995, when monotherapy and dual therapy were available; and 1996-2002, when HAART was available). Over time, the proportion of women and non-white PWA increased, and the age at AIDS diagnosis increased. The proportion of PWA who were men who have sex with men declined, whereas the proportion who were heterosexual increased.
In the 10-year period spanning 60 months before to 60 months after AIDS onset, n = 1489 lung cancer cases were observed. Compared to the general population, lung cancer risk among PWA was significantly elevated overall (SIR, 3.8; 95% CI, 3.6-4.1). With the exception of the period spanning 60 to 25 months before AIDS (SIR, 1.0; 95% CI, 0.7-1.3; n = 42 cases), lung cancer risk was significantly elevated in all periods relative to AIDS onset and was particularly high during the AIDS onset period: 24 to 7 months before AIDS (SIR, 2.6; 95% CI, 2.1-3.1; n = 123 cases), 6 months before to 3 months after AIDS (SIR, 10.5; 95% CI, 9.7-11.4; n = 629 cases), 4 to 27 months after AIDS (SIR, 2.9; 95% CI, 2.6-3.2; n = 393 cases), and 28 to 60 months after AIDS (SIR, 2.9; 95% CI, 2.5-3.2; n = 302 cases). Excluding the AIDS-onset period (6 months before to 3 months after AIDS), a significant increasing trend was observed in SIRs across AIDS-relative time (60 to 25 months before AIDS, 24 to 7 months before AIDS, 4 to 27 months after AIDS, and 28 to 60 months after AIDS; P-value for linear trend < 0.001).
The remaining analyses were limited to the 2-year period spanning 4 through 27 months after AIDS onset (N = 317 007 PWA with 477 012 person-years of follow-up). During this period, lung cancer incidence was 82.3 per 100 000 person-years. In comparison with the general population, lung cancer incidence was elevated in all demographic subgroups (Table 2). Although incidence was higher in men than in women and increased with age, SIRs indicated that risk relative to the general population was higher for women and especially high for younger individuals (10.4 for ages 15-29 years and 6.3 for ages 30-39 years; Table 2). Lung cancer risk was elevated for all HIV risk groups, particularly for injection drug users (SIR, 3.9). Risk among PWA was elevated across all calendar periods but was highest during 1990-1995.
The majority of lung cancers occurring in the 2-year follow-up period were diagnosed at an advanced stage. Stage information was unavailable from Florida (n = 74 cases); from the remaining registries (n = 319 cases), 11.2% were local, 18.2% regional, 50.5% distant, and 20.1% were unstaged (corresponding proportions from the SEER program were 18.5, 33.1, 36.9, and 11.5%, respectively) [20]. Similar results were observed for lung cancers diagnosed in the AIDS-onset period (6 months before to 3 months after AIDS: 11.0% local, 21.2% regional, 53.4% distant, and 14.4% unstaged). Adenocarcinoma was the most frequently observed subtype (34.0%), followed by squamous cell carcinoma (20.8%), large cell carcinoma (9.9%), small cell carcinoma (8.6%), and bronchioloalveolar carcinoma (2.0%). Approximately 24% of cases belonged to other histologic subtypes, which is similar to the corresponding proportion in the SEER program (23.3%) [20]. Risk of all histologic subtypes was similarly elevated among PWA compared to the general population (Table 2).
Lung cancer risk was significantly elevated in all CD4 cell count categories, except among PWA with CD4 cell count ≥ 300 cells/μl (Table 2). No significant trend in lung cancer incidence across CD4 cell count categories was observed (Fig. 1; P = 0.36 unadjusted, P = 0.23 after adjustment for demographic factors in Table 1). Similar analyses performed separately for the pre-HAART (1990-1995) and the HAART eras (1996-2002) showed no relationship with CD4 cell counts in either period (unadjusted P = 0.99 and P = 0.13, respectively).
The observed lung cancer incidence was compared with the average of the 12 predicted rates considering varying proportions of current smokers among PWA (80 and 60%) (Table 3). Regardless of smoking prevalence, observed lung cancer incidence was significantly higher than predicted among PWA aged 40-49 years. Even assuming that 80% of PWA were current smokers, lung cancer incidence was substantially elevated for 40-49-year-old PWA (observed/predicted = 5.03 for men and 1.88 for women). Observed rates were also higher than predicted among 50-59-year-old PWA, particularly for men, but the excess was less marked than for 40-49 year olds. Observed rates were similar to predicted among 60-69-year-old PWA. Similar results were observed in sensitivity analyses restricted to white PWA (Table 3).
Discussion
We found that lung cancer risk was substantially higher among PWA than in the general population. For younger PWA, lung cancer risk was especially elevated and the present analyses suggested that the observed lung cancer incidence was higher than could be predicted by the effects of smoking alone. These results point to the possibility that additional co-factors other than tobacco may be important in the development of this malignancy in HIV-infected individuals.
The SIR of 3.8 for the period spanning 60 months before to 60 months after AIDS onset is consistent with estimates from previous studies that have ranged from 1.9 to 7.4 [4-11]. Lung cancer risk in the 2-year period following AIDS onset was also significantly elevated. Although incidence increased appreciably with age, it was found that risk among PWA, relative to the general population, was remarkably high in the young (SIR, 10.4 for 15-29-year-old PWA; 6.3 for 30-39 year olds). The histologic distribution of lung cancers among PWA was similar to that observed in the general population, with a predominance of adenocarcinoma [2,3,18].
Cigarette smoking is the most important risk factor for lung cancer [19], and high lung cancer risk among HIV-infected subjects can be attributed in part to heavy smoking [13]. In the US 60-80% of HIV-infected individuals smoke compared with 20-30% in the general population [13,22]. While the majority of lung cancers among HIV-infected individuals occur in smokers [2,3,8,11], few epidemiologic studies have attempted to estimate the effect of HIV infection on lung cancer risk after accounting for smoking behavior. Two previous studies indirectly adjusted for smoking and reported that, even under the assumption of 100% smoking prevalence among HIV-infected individuals, lung cancer incidence was 2.0-2.5 times higher than predicted based on general population rates [11,16]. Two other studies reported no significant differences in lung cancer risk between similar cohorts of HIV-infected and HIV-uninfected women, although the number of cases was small in both studies [14,15].
We used the models developed by Flanders et al. [19] to take into account duration and intensity of smoking, which play a critical role in lung cancer etiology [19]. Under plausible smoking scenarios, observed incidence was significantly higher than predicted among 40-59-year-old men with AIDS and 40-49-year-old women with AIDS. Further indicating a substantial excess risk among younger PWA, the SIRs among PWA aged less than 40 years old were remarkably elevated (Table 2). In contrast, the present results indicated that plausible smoking could explain the observed excess of lung cancer among the oldest PWA.
The increased lung cancer risk among PWA may be attributed in part to intensive medical evaluations at and after AIDS onset, as evidenced by the extremely high risk of lung cancer observed in the AIDS onset period (6 months before to 3 months after AIDS; SIR, 10.4). However, cancers diagnosed through over-ascertainment would be expected to occur at earlier stages, whereas it was observed that, in both the AIDS onset period and in the 4-27 months after AIDS onset, lung cancers were actually diagnosed at a more advanced stage than is typically observed in the general population [20]. Furthermore, any heightened surveillance at AIDS onset would be expected to lead to a subsequent apparent decrease in lung cancer risk. Instead, lung cancer risk remained significantly elevated throughout the 5-year period after AIDS onset. These results thus argue against over-ascertainment as a potential explanation for the increased lung cancer risk among PWA.
The substantial lung cancer risk at younger ages coupled with an advanced stage at diagnosis suggests an acceleration of lung cancer pathogenesis among PWA. One possibility is that the effects of smoking could be more severe in HIV-infected individuals. Indeed, an accelerated form of smoking-related emphysema, a risk factor for lung cancer [23], also occurs among HIV-infected individuals [24]. Several lines of evidence point to altered lung function and enhanced lung damage among HIV-infected individuals. HIV infection may lead to increased oxidative stress in the lung mediated through deficiency of antioxidants such as glutathione [25]. HIV-infected individuals are also more susceptible to lung infections and pneumonia caused by a wide range of organisms such as Streptococcus pneumoniae, Chlamydia pneumoniae, Pneumocystis jirovecii, and Mycobacteria species. Repeated or chronic lung infections could increase lung cancer risk through chronic inflammation [26].
Although some of the aforementioned processes would be expected to be more severe with advancing HIV disease, the actual relationship between immunosuppression and lung cancer is uncertain. No association was found between lung cancer risk and immunosuppression, as measured by CD4 cell counts at AIDS onset (Fig. 1). However, because the relationship of lung cancer was only assessed with CD4 cell counts at AIDS onset, an association could have been missed if this relationship existed at higher CD4 cell counts or among people with early stage HIV-infection. In contrast, a significant increasing trend was found in SIRs from 60 to 25 months before AIDS onset to 28 to 60 months after AIDS onset, suggesting that lung cancer pathogenesis could still be related to more subtle disturbances of immunity or inflammation.
The strengths of the present study include its large size and population-based design (i.e., including all lung cancers diagnosed in a large and representative fraction of all US AIDS cases), and the comprehensive evaluation of relationships with demographic factors, smoking, and immune function. The limitations of the study should also be noted. Most importantly, actual measured smoking information on the participants was not available. As lung cancer is a somewhat rare outcome, this limitation is a general problem, in that it is difficult to obtain smoking information from HIV-infected and HIV-uninfected cohorts of adequate size that simultaneously provide robust data on cancer incidence. Although the use of lung cancer prediction models made it possible to take into account the intensity and duration of smoking, these analyses involved several assumptions. Additionally, because these models were based on individuals who smoked ≤ 40 cigarettes per day, it was not possible to evaluate more extreme smoking behaviors [19]. Although expected rates of lung cancer were corrected to account for mortality in the periods preceding AIDS, the SIRs in the pre-AIDS periods and trends in SIRs across AIDS-relative time should be interpreted with caution because they also involve several assumptions [27].
In conclusion, the results of the present study indicate that lung cancer risk is significantly elevated among PWA and that smoking does not readily explain all of the excess lung cancer risk. Additional research is needed to understand the etiology of lung cancer among HIV-infected individuals.
Acknowledgements
HIV/AIDS and cancer registries in the following regions participated in the HIV/AIDS Cancer Match Study: the states of Connecticut, Florida, Georgia, Massachusetts, Michigan, and New Jersey; and the metropolitan areas of Los Angeles, San Diego, and San Francisco (California), New York City (New York), and Seattle (Washington). We are grateful for the many staff members at these registries who collected the data, prepared data files for the matches, and facilitated the record linkages.
We also thank our study managers Norma Kim and Emily Moser (RTI International, Rockville, Maryland). We are grateful to Dr Neil E. Caporaso and Dr Jay H. Lubin (Division of Cancer Epidemiology and Genetics, National Cancer Institute) for critical reading of the manuscript.
Sponsorship: This research was supported by the Intramural Research Program of the National Cancer Institute, National Institutes of Health.
References
1. Engels EA, Pfeiffer RM, Goedert JJ, Virgo P, McNeel TS, Scoppa SM, et al. Trends in cancer risk among people with AIDS in the United States 1980-2002. AIDS 2006; 20:1645-1654.
2. Powles T, Nelson M, Bower M. HIV-related lung cancer - a growing concern? Int J STD AIDS 2003; 14:647-651.
3. Lavole A, Wislez M, Antoine M, Mayaud C, Milleron B, Cadranel J. Lung cancer, a new challenge in the HIV-infected population. Lung Cancer 2005; 51:1-11.
4. Frisch M, Biggar RJ, Engels EA, Goedert JJ. Association of cancer with AIDS-related immunosuppression in adults. JAMA 2001; 285:1736-1745.
5. Herida M, Mary-Krause M, Kaphan R, Cadranel J, Poizot-Martin I, Rabaud C, et al. Incidence of non-AIDS-defining cancers before and during the highly active antiretroviral therapy era in a cohort of human immunodeficiency virus-infected patients. J Clin Oncol 2003; 21:3447-3453.
6. Gallagher B, Wang Z, Schymura MJ, Kahn A, Fordyce EJ. Cancer incidence in New York State acquired immunodeficiency syndrome patients. Am J Epidemiol 2001; 154:544-556.
7. Dal Maso L, Polesel J, Serraino D, Franceschi S. Lung cancer in persons with AIDS in Italy, 1985-1998. AIDS 2003; 17:2117-2119.
8. Bower M, Powles T, Nelson M, Shah P, Cox S, Mandelia S, et al. HIV-related lung cancer in the era of highly active antiretroviral therapy. AIDS 2003; 17:371-375.
9. Clifford GM, Polesel J, Rickenbach M, Dal Maso L, Keiser O, Kofler A, et al. Cancer risk in the Swiss HIV Cohort Study: associations with immunodeficiency, smoking, and highly active antiretroviral therapy. J Natl Cancer Inst 2005; 97:425-432.
10. Grulich AE, Wan X, Law MG, Coates M, Kaldor JM. Risk of cancer in people with AIDS. AIDS 1999; 13:839-843.
11. Engels EA, Brock MV, Chen J, Hooker CM, Gillison M, Moore RD. Elevated incidence of lung cancer among HIV-infected individuals. J Clin Oncol 2006; 24:1383-1388.
12. Biggar RJ, Engels EA, Ly S, Kahn A, Schymura MJ, Sackoff J, et al. Survival after cancer diagnosis in persons with AIDS. J Acquir Immune Defic Syndr 2005; 39:293-299.
13. Giordano TP, Kramer JR. Does HIV infection independently increase the incidence of lung cancer? Clin Infect Dis 2005; 40:490-491.
14. Hessol NA, Seaberg EC, Preston-Martin S, Massad LS, Sacks HS, Silver S, et al. Cancer risk among participants in the women's interagency HIV study. J Acquir Immune Defic Syndr 2004; 36:978-985.
15. Phelps RM, Smith DK, Heilig CM, Gardner LI, Carpenter CC, Klein RS, et al. Cancer incidence in women with or at risk for HIV. Int J Cancer 2001; 94:753-757.
16. Parker MS, Leveno DM, Campbell TJ, Worrell JA, Carozza SE. AIDS-related bronchogenic carcinoma: fact or fiction? Chest 1998; 113:154-161.
17. Mbulaiteye SM, Biggar RJ, Goedert JJ, Engels EA. Immune deficiency and risk for malignancy among persons with AIDS. J Acquir Immune Defic Syndr 2003; 32:527-533.
18. Jemal A, Travis WD, Tarone RE, Travis L, Devesa SS. Lung cancer rates convergence in young men and women in the United States: analysis by birth cohort and histologic type. Int J Cancer 2003; 105:101-107.
19. Flanders WD, Lally CA, Zhu BP, Henley SJ, Thun MJ. Lung cancer mortality in relation to age, duration of smoking, and daily cigarette consumption: results from Cancer Prevention Study II. Cancer Res 2003; 63:6556-6562.
20. National Cancer Institute, DCCPS Surveillance Research Program Cancer Statistics Branch. Surveillance, Epidemiology, and End Results (SEER) Program (
www.seer.cancer.gov) SEER*Stat Database: Incidence-SEER 9 Regs Public-Use, NOV 2004 Sub (1973-2002), National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch, released April 2005, based on the November 2004 submission. 4-1-2005.
21. Haiman CA, Stram DO, Wilkens LR, Pike MC, Kolonel LN, Henderson BE, et al. Ethnic and racial differences in the smoking-related risk of lung cancer. N Engl J Med 2006; 354:333-342.
22. Burns DN, Kramer A, Yellin F, Fuchs D, Wachter H, DiGioia RA, et al. Cigarette smoking: a modifier of human immunodeficiency virus type 1 infection? J Acquir Immune Defic Syndr 1991; 4:76-83.
23. Schabath MB, Delclos GL, Martynowicz MM, Greisinger AJ, Lu C, Wu X, et al. Opposing effects of emphysema, hay fever, and select genetic variants on lung cancer risk. Am J Epidemiol 2005; 161:412-422.
24. Diaz PT, King MA, Pacht ER, Wewers MD, Gadek JE, Nagaraja HN, et al. Increased susceptibility to pulmonary emphysema among HIV-seropositive smokers. Ann Intern Med 2000; 132:369-372.
25. Pacht ER, Diaz P, Clanton T, Hart J, Gadek JE. Alveolar fluid glutathione decreases in asymptomatic HIV-seropositive subjects over time. Chest 1997; 112:785-788.
26. Littman AJ, Jackson LA, Vaughan TL. Chlamydia pneumoniae and lung cancer: epidemiologic evidence. Cancer Epidemiol Biomarkers Prev 2005; 14:773-778.
27. Engels EA, Frisch M, Biggar RJ, Goedert JJ. Re: Estimation of risk of cancers before occurrence of acquired immunodeficiency syndrome in persons infected with human immunodeficiency virus. Am J Epidemiol 2003; 157:955-956.
© 2007 Lippincott Williams & Wilkins, Inc.