- Email: [email protected]
High incidence of anal cancer among AIDS patients
High incidence of anal cancer among AIDS patients
Summary Until now, the only cancers that have been strongly associated with AIDS are Kaposi’s sarcoma and non-Hodgkin lymphoma. We used a linkage between AIDS (50 050 reports) and cancer (859 398 reports) registries in seven health departments in the USA to investigate the association between HIV infection and epidermoid anal cancer. We compared the numbers of observed cases and expected cases, calculated from general population rates with adjustment for age, sex, and race. The relative risk of anal cancer at and after AIDS diagnosis was 84·1 (95% Cl 46·4-152) among homosexual patients (11 cases) and 37·7 (9·4-151) among non-homosexual patients (2 cases). The relative risk of anal cancer up to 5 years before the AIDS diagnosis (23 cases) was also increased; it was 13·9 (6·6-29·2) in the period 2-5 years before AIDS and 27·4 (15·9-47·2) during the 2 years before AIDS diagnosis (p for trend=0·004). Among homosexual men, the relative risk of anal cancer was inversely related to age at AIDS onset (p for trend<0·001). Excess risks were found in all geographical areas.
This study establishes a strikingly increased risk of anal among people with AIDS. These data are consistent with a previously hypothesised association between HIVinduced immunodeficiency and anal cancer development, but because homosexual men were at increased risk of anal cancer even before the AIDS epidemic, we cannot say how much of the increased risk is attributable to HIV infection. Nevertheless, clinicians should be aware that AIDS patients have an increased risk of anal cancer. cancer
Lancet 1994; 343: 636-39
Epidemiology Science Centre, State Serum Institute, Copenhagen, Denmark (Prof M Melbye MD); Viral Epidemiology Branch, Division of Cancer Etiology (T R Coté MD, R J Biggar MD), Applied Research Branch, Division of Cancer Prevention and Control (L Kessler ScD), and Biostatistics Branch, Division of Cancer Etiology (M Gail MD), National Cancer Institute, Rockville, Maryland, USA The Danish
*
Members of the working group
are
listed at the end of the article.
Dr Timothy R Coté, Viral Epidemiology Branch, National Cancer Institute (NCI), EPN-434, 6130 Executive Blvd, Rockville, MD 20852, USA
Correspondence to:
636
Introduction Reports from Denmark, Sweden, and Connecticut, USA, have shown significant increases in the incidence of epidermoid anal cancer during the past 30 years.’-3 The increase has been more pronounced in women than in men and in urban than rural areas. Furthermore, black people are at higher risk than whites and never-married men at higher risk than ever-married men. These trends show that important behavioural and environmental changes were taking place before the beginning of the AIDS epidemic. Data from the Surveillance, Epidemiology and End Results Programme (SEER) in the USA have shown an especially rapid increase in epidermoid anal cancer incidence among men in the San Francisco Bay area.3 The incidence among these men was nearly 3 times higher than that among men living in other areas of the USA covered by the SEER programme or among Scandinavian men." Furthermore, the relative risk of anal cancer among never-married men (a surrogate marker for homosexuality) increased from 6-7 (95% CI 4-7-9-5) in 1979-84 to 10-3 (7-5-14-1) in 1985-89. Although these changes could reflect an influence by the growing AIDS epidemic, they could also be the result of an independent but contemporaneous influence of another aetiological agent with increasing prevalence among homosexual men during this period. Cervical cancer in an HIV-infected woman is now accepted as an AIDS-defining illness.4 Cervical and anal cancers share many characteristics;5-8 thus a link between AIDS and anal cancer is also plausible. In support of a direct influence by HIV and its associated immunodeficiency on anal cancer development, we and others have described anal intraepithelial abnormalities in a substantial proportion of HIV-positive homosexual men and shown a significant association between these abnormalities and decreasing immune competence.9-11 Although the anal lesions have not been shown to progress to invasive cancer, there is preliminary evidence of progression from mild more anal epithelial neoplasia among to severe immunocompromised homosexual men.12 Furthermore, among men with AIDS-related Kaposi’s sarcoma, most of whom probably were homosexual, there was a significant excess of anal cancer. 13 To investigate the possible association between the AIDS epidemic and the development of anal cancer, we have analysed data from a newly established registry that links AIDS and cancer surveillance systems in the USA.
Subjects and methods Until 1993, AID S registries in the United States recorded only two cancer types-Kaposi’s sarcoma and non-Hodgkin lymphoma. It was mandatory for these cancers to be reported to AIDS registries only when they were one of the initial AIDS-defining diseases; diagnoses of these cancers that followed an earlier AIDS diagnosis would not necessarily be recorded in AIDS registries. However, they would be registered by cancer registries whenever they occurred. To understand the relation between AIDS and cancer in
general, in 1992, one of us (TRC) set up a programme to link AIDS and cancer registries in nine regions of the USA/’* Between May and November, 1992, AIDS and cancer registries linked in California, Florida, Metropolitan Atlanta, and New Jersey. All registered individuals were followed until death or 2-25 years after the AIDS diagnosis, whichever came first. This restriction was applied to limit problems associated with loss to follow-up (eg, unregistered deaths or migration away from a region after a diagnosis of AIDS but before a cancer diagnosis); 2 25 years represents the average lifespan from AIDS to death. Only people younger than 70 years were included in the linkage. Because some cancer registries were established earlier or later than the AIDS registries, we included only periods in which both registries were were
Table 1: Relative risk (observed/expected ratio) of epidermoid anal and anorectal cancer among AIDS patients compared with population controls matched for age, sex, and race cancers
(two cancers adjacent to the anorectal area not thought to be
operating. Expected incident anal cancer cases after AIDS diagnosis were calculated by multiplying SEER age-specific incident rates by the corresponding person-years at risk after AIDS diagnosis. A person was at risk until anal cancer, death or censoring at 2 25 years, or
associated with HIV/AIDS).
until cancer surveillance ended, whichever came first. The observed number of cases was then compared with the expected number to give observed/expected ratios (relative risk) and Poisson-distributed 95% CI.15 The frequency of anal cancer preceding AIDS was first known from the cancer registries. We provide an estimate of the interval-specific relative risk. However, calculating the expected rates was complicated because HIV-infected subjects who develop anal cancer might die before progression to AIDS and therefore would never be matched in the AIDS registry. To estimate the expected number of anal cancer cases that would have occurred among AIDS cases before their AIDS diagnosis (assuming that HIV-infected individuals have the same cancer rates and survival as the general population) we used modifications of published techniques.16 This approach estimates the probability of a person’s having had anal cancer diagnosed and surviving the necessary time to have developed AIDS, with incidence and survival rates from SEER data specific for age, sex, and period.1s We acknowledge that this approach is approximate. If HIV infection accelerates death from anal cancer, we may have underestimated the risk; conversely, if anal cancer accelerates the development of AIDS, we may have overestimated the risk. No information about these effects is available. Anal cancer is usually treated by surgery or local radiotherapy, which is unlikely to affect the course of HIV-related
The linkage analysis included 859 398 reports of cancer and 50 050 reports of AIDS. We found 39 AIDS patients with a diagnosis of epidermoid anal cancer (anal in 31, anorectal in 8), 3 of whom had their cancer diagnosed more than 5 years before AIDS was diagnosed. 21 other patients with anal-site cancers of other types were excluded, including 13 with malignant lymphoma (1Burkitt’s, 9 immunoblastic, 2 large-cell, 1 mixed-cell), and 8 with Kaposi’s sarcoma. All but 1 of the 39 patients were men. 36 were homosexual or bisexual, and 6 of these were also intravenous drug users; 2 others were intravenous drug users who did not report homosexual behaviour. For 1 case, risk behaviour was not known. To find out how serious these tumours were clinically thought to be, we examined treatment records for 30 patients; records for the patients diagnosed in Los Angeles were not available. No case was diagnosed at necropsy. Of the 30 patients with available treatment information, 21 had received radiotherapy or chemotherapy and 4 others had undergone operations. 3 of the 5 untreated patients died soon after the cancer diagnosis (10 days-3 months). Thus only 2 patients were untreated for unknown reasons. The relative risk of epidermoid anal cancer was 63 4 (95 % CI 36-8-109) after a diagnosis of AIDS compared with the general population. The relative risk was 841 (46 4—152) among homosexual or bisexual male patients and 37-7 (9 4-151 ) among non-homosexual patients. We found a 27-4 relative risk for anal cancer for the 2 years before an AIDS diagnosis and a somewhat lower risk (13-9) in the 2-5 years before AIDS diagnosis (table 1). Table 2 gives the observed/expected ratios for anal cancer after AIDS by age group and geographical distribution for homosexual men. Similar analyses for other risk groups were not possible because of small numbers. The relative
immunosuppression. To estimate the proportion of the general population of survivors to the time of AIDS diagnosis who had anal cancer diagnosed in a particular earlier time period, we used incidence data specific for sex, age, and period to count the number of cases incident in each month during the 5 years before the attained age at AIDS diagnosis. We then applied observed survival estimates to the cases lost to follow-up to calculate the fraction of cases still alive at the attained age. For each of the 60 months before AIDS, this estimate added to the number of cancer cases who were known to have survived to the attained age, and the sum was divided by the estimated population total to arrive at the proportion with an anal cancer diagnosed in an earlier month who survived to the attained age of AIDS diagnosis. To calculate the expected cancer counts in each month among patients in the AIDS registry, these proportions were multiplied by the corresponding person-years among AIDS cases. These monthly expected counts were then summed into longer periods before AIDS diagnosis and compared with observed counts for the same period. Tests for trend in relative risk were based on a score statistic for a Poisson model. Only cancers of the epidermoid type were studied in detail. This definition excluded AIDS-defining cancers (eg, Kaposi’s sarcoma and non-Hodgkin lymphoma) located in the anal region and focused the study on the histological type of tumour that has previously been reported to undergo significant changes.1-3 We used topography codes for epidermoid anorectal cancers 1541, 1542, 1543, and 1548, and histology codes 80513-23, 80703-63, 80903-53, and 81203-43, according to the International Classification of Diseases for Oncology (ICD-O).18 To seek possible ascertainment bias, a similar analytical approach was taken to assess the association with colon and prostate was
lymphoma histologies
were
Kaposi’s sarcoma and non-Hodgkin
again excluded.
Results
*Of anal cancer in homosexual men after AIDS diagnosis. tp for trend 0 0003. tFor San Francisco only 0/E= 134 (50 4-358).
Table 2: Relative risk of epidermoid anal homosexual men with AIDS
cancer
among
637
risk of developing anal cancer was highest for homosexual AIDS patients aged 20-29 years and was inversely correlated with age (p for trend 00003). Excess risks were found in all geographical areas. The relative risks of colon and prostate cancer were 0-7 and 22 at time of AIDS diagnosis and 0-6 and 1 -4 after the AIDS diagnosis.
Discussion This study documents an increased risk of anal cancer after AIDS diagnosis. Although the relative risks we report are high, especially for young homosexual men, they are based on only few cases. Therefore, the nature of the recordlinkage data needs to be carefully considered. We may have overestimated the relative risk. Mismatched cases could inflate apparent cancer risk. However, in a validation study of our linkage method in Los Angeles, all records linking AIDS and cancer diagnoses were reviewed by direct examination of the information at each registry. Only 9 (0-3%) of 2646 records were incorrectly linked (D Deapen, unpublished). All cancers in our study were histologically confirmed as invasive epidermoid cancers. Kaposi’s sarcoma and non-Hodgkin lymphoma are rarely found as primary tumours of the anorectal area; therefore, our cases were unlikely to be miscoded as these disorders. Nearly all of the cases found were treated with radiotherapy, chemotherapy, or surgery; we therefore doubt that the excess was due to misclassified trivial in-situ dysplasia. We were concerned that the excess might have been attributed to research interest in anal cancer in San Francisco, since this area has an active anal cancer research programme. 11,12 However, there was only a modest excess (1-33) of anal cancer diagnoses in San Francisco AIDS patients compared with those of other areas.
The increased medical attention given to AIDS patients probably created ascertainment bias, especially at the time of the initial AIDS diagnosis. This factor may be especially important for malignant disorders that have long latent periods and can therefore be detected through screening. However, we found little ascertainment bias around the time of AIDS diagnosis for two adjacent sites not suspected to have AIDS-associated malignant disorders (colon and
prostate). It is also possible that we underestimated the actual risk. For example, there may have been unreported cases or missed linkages. Furthermore, screening as part of examinations at the time of AIDS diagnosis could have detected cancers sooner than if the patient had not had AIDS. This factor would inflate the apparent risk at the time of AIDS, but it would also lower the observed rate in the follow-up period. In addition, AIDS has a short survival duration, and unregistered deaths (such as those lost
through migration)
will
spuriously
increase the
apparent size of the population under study, thus lowering calculated relative risks. The most important factor to take into account is that homosexual men were known to have an increased risk of anal cancer even before the AIDS epidemic.2,19-23 In Holly et al’s21 case-control study, homosexual men were 12 times more likely to develop anal cancer than age-matched non-homosexual men.21 That study was carried out in the San Francisco area during the 1980s and could have included homosexual men who were HIV-infected, which might have increased their risk. The relative risk we 638
recorded for homosexual men with AIDS was much higher (84-1). However, the subgroup of homosexual men at highest risk of AIDS may also have had the highest exposure to the unknown factors that increase the risk of anal cancer among homosexual men generally. Consistent with the hypothesis of an HIV association, we also found a significant excess risk of anal cancer in AIDS patients not identified as homosexual. However, this finding was based on only 2 patients (1male,1 female) and it is possible that the man, although reported as being an intravenous drug user, could have had misclassified exposure risks. We used a modification of a previously published method16 to estimate the risk of anal cancer in the 5-year period before the patients’ AIDS diagnoses. Although our estimates must be interpreted with caution, there was an increasing trend in the relative risk towards the time of AIDS diagnosis. Thus, the high risk we report could be due to HIV-induced immunodeficiency. This analysis cannot define what proportion of this risk, if any, is attributable to HIV and what proportion is due to a homosexual lifestyle. An aetiological association with HIV is biologically plausible and has long been suspected.24 In a study of severely immunosuppressed men with AIDS or other serious HIV disease, 15 % had anal epithelial neoplasia and 2% carcinoma-in-situ.11 Other studies of homosexual men have found high prevalence rates of anal epithelial atypia and dysplasia and have documented an association between HIV infection and increased prevalence and quantity of human papillomavirus (HPV) in anal samples.9.10 Such studies have also found significant associations between the presence of HPV and anal epithelial abnormalities9.1o.2s and shown anal epithelial abnormalities to be more frequent in men with both HIV and HPV than in those with either virus alone. Previous studies have shown an association between anal and cervical cancer and suggested a similar pathogenesis.S-8 Invasive cervical cancer is now included as an AIDSdefining illness in HIV-positive individuals, primarily because of the plausibility of an association.4However, the evidence for an association between invasive cervical cancer and HIV is weak,26 especially because the impact of confounding has not been adequately explored. In a preliminary analysis of linked records from these AIDS and cancer registries, invasive cervical cancer was only marginally increased over background.27 We propose two reasons why the relative risk of cervical cancer in AIDS patients is lower than that of anal cancer. First, cervical cancer is much more common than anal cancer in the general population, so small changes are more difficult to detect. Second, cervical dysplasia is detected by active screening programmes among HIV-positive women, which reduces the likelihood of progression to invasive cervical cancer. HIV-infected women have, in general, a higher rate of sexually transmitted diseases than other women 211 and will therefore be in closer contact with the health care system both before and after their HIV infection. The mechanism for a putative association between anal cancer and HIV remains unclear but the cancer types reported to be associated with AIDS are similar to those common to other forms of immunodeficiency.26 Excesses of anogenital cancers, including both cervical and anal cancer, have also been reported among renal transplant patients.29.30 So far, only a few cancer types have been associated with the HIV epidemic. This feature argues against the concept of general immune surveillance of
defence mechanism for cancer development. However, the tumours associated with the HIV epidemic are suspected to have a viral oncogenesis. Therefore, a potential influence for HIV-related immunodeficiency on tumour development could be through failing intracellular control of persisting viral
cellular control
as
a
genomes.
Clinicians dealing with HIV-infected homosexual men should be aware of an increased risk of anal cancer in these
patients. Fortunately, digital examination of the anorectal region is standard medical practice. If anal cancer is suspected on clinical grounds, assessment for cytological abnormalities is simple and inexpensive and, when necessary, treatment is often effective.
Implications for anal cancer screening programmes are complex. Although the relative risk of invasive anal cancer is increased, the absolute risk of this cancer in the group of homosexual men with AIDS we studied was about 1 per 1000 cases. However, anal dysplasia (including carcinomain-situ) may occur in up to 15 % of severely immunosuppressed HIV-infected homosexual men without signs or symptoms of anal cancer. Apparently, most AIDS patients diagnosed with anal dysplasia do not develop clinically significant invasive anal cancers before they die, at least with the current outlook for AIDS patients. Therefore, any recommendation to institute screening for anal dysplasia in symptom-free individuals will need to be based on a careful cost-benefit analysis. This study was supported by the National Cancer Institute and by a grant from the Danish Medical Research Council (J nr 5.22.98.62). We thank Ms Frances Yellin (ARC Professional Services, Rockville) for programming assistance; Cliff Notes Inc for the first name thesaurus used in the linkage; and Dr William Blattner, Dr Charles Rabkin, and Dr James J Goedert for helpful comments on the manuscript.
Members of the AIDS jCancer Working Group (for each city or state, the first member listed is from the AIDS registry and the second from the cancer registry). San Francisco (George Lemp, Dee West); California (Jim Singleton, John Young); Los Angeles (Peter Kerndt, Dennis Deapen); San Diego (Michelle Ginzberg, Hoda Anton-Culver); Florida (Spencer Lieb, Richard Hopkins); Georgia (Brian Williams, Jonathan Liff); New Jersey
(Douglas Morgan, William Parkin).
References 1
2 3
Goldman S, Glimelius B, Nilsson B, Paahlman L. Incidence of anal epidermoid carcinoma in Sweden 1970-1984. Acta Chir Scand 1989; 155: 191-97. Frisch M, Melbye M, Moller H. Trends in incidence of anal cancer in Denmark. BMJ 1993; 306: 419-22. Melbye M, Rabkin CS, Frisch M, Biggar RJ. Changing patterns of anal cancer incidence in the USA, period 1940-89. Am J Epidemiol (in
press). CDC. 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR 1992; 41/RR-17: 1-19. 5 Scholefield JH, Sonnex C, Talbot IC, et al. Anal and cervical intraepithelial neoplasia: possible parallel. Lancet 1989; ii: 765-69.
4
6
Melbye M, Sprogel P. Aetiological parallel between anal cancer and cervical cancer. Lancet 1991; 338: 657-59.
Rabkin CS, Biggar RJ, Melbye M, Curtis RE. Second primary cancers following anal and cervical carcinoma: evidence of shared etiologic factors. Am J Epidemiol 1992; 136: 54-58. 8 Scholefield JH, Hickson WGE, Smith JHF, Rogers K, Sharp F. Anal epithelial neoplasia: part of a multifocal disease process. Lancet 1992; 340: 1271-73. 9 Melbye M, Palefsky J, Gonzales J, et al. Immune status as a determinant of human papillomavirus detection and its association with anal epithelial abnormalities. Int J Cancer 1990; 46: 203-06. 10 Caussy D, Goedert JJ, Palefsky J, et al. Interaction of human immunodeficiency virus and papillomaviruses: association with anal epithelial abnormality in homosexual men. Int J Cancer 1990; 46: 214-19. 11 Palefsky JM, Gonzales J, Greenblatt RM, et al. Anal intraepithelial neoplasia and anal papillomavirus infection among homosexual males with group IV disease. JAMA 1990; 263: 2911-16. 12 Palefsky JM, Holly EA, Gonzales J, Lamborn K, Hollander H. Natural history of anal cytologic abnormalities and papillomavirus infection among homosexual men with group IV disease. J Acq Immun Defic Syndr 1992; 5: 1258-65. 13 Biggar RJ, Curtis RE, Coté TR, Rabkin CS, Melbye M. Risk of other cancers following Kaposi’s sarcoma: relationship to the acquired immunodeficiency syndrome. Am J Epidemiol 1994; 139: 362-68. 14 Coté TR. NCI plans linkage of AIDS and cancer registries. J Natl Cancer Inst 1993; 85: 1113-14. 15 Breslow NE, Day NE. Statistical methods in cancer research. Vol II: the design and analysis of cohort studies. Lyon: IARC, 1987: 48-79. 16 Feldman AR, Kessler L, Myers MH, Naughton MD. The prevalence of cancer: estimates based on the Connecticut Tumor Registry. N Engl J Med 1986; 315: 1394-97. 17 Ries LG, Pollack ES, Young JL. Cancer patient survival: surveillance, epidemiology, and end results program, 1973-79. J Natl Cancer Inst 1983; 70: 693-707. 18 WHO. ICD-O. International classification of diseases for oncology. Geneva: WHO, 1980. 19 Peters RK, Mack TM. Patterns of anal carcinoma by gender and marital status in Los Angeles County. Br J Cancer 1983; 48: 629-36. 20 Daling JR, Weiss NS, Hislop TG, et al. Sexual practices, sexually transmitted diseases, and the incidence of anal cancer. N Engl J Med 1987; 317: 973-77. 21 Holly EA, Whittemore AS, Aston DA, et al. Anal cancer incidence: genital warts, anal fissure or fistula, hemorrhoids, and smoking. J Natl Cancer Inst 1989; 81: 1726-31. 22 Scholefield JH, Jones HT, Cuzick J, Northover JMA. Anal cancer and marital status. Br J Cancer 1990; 62: 286-88. 23 Adam YG, Efron G. Current concepts and controversies concerning the etiology, pathogenesis, diagnosis, and treatment of malignant tumors of the anus. Surgery 1987; 101: 253-66. 24 Frazer IH, Medley G, Crapper RM, Brown TC, Mackay IR. Association between anorectal dysplasia, human papillomavirus, and human immunodeficiency virus infection in homosexual men. Lancet 1986; ii: 657-60. 25 Kiviat N, Rompalo A, Bowden R, et al. Anal human papillomavirus infection among human immunodeficiency virus seropositive and seronegative men. J Infect Dis 1990; 162: 358-61. 26 Rabkin CS, Blattner WA. HIV infection and cancers other than non-Hodgkins lymphoma and Kaposi’s sarcoma. Cancer Surveys 1991; 10: 151-60. 27 Coté TR, Schiffman M, Biggar RJ, et al. Invasive cervical cancer among women with AIDS: results of registry linkage. IX International Conference on AIDS, Berlin 1993, Abstr no PO-B14-1637. 28 Wasserheit JN. Epidemiological synergy. Interrelationships between human immunodeficiency virus infection and other sexually transmitted diseases. Sex Transm Dis 1992; 19: 61-77. 29 Penn I. Cancers of the anogenital region in renal transplant recipients. Cancer 1986; 58: 611-16. 30 Blohmé I, Brynger H. Malignant disease in renal transplant patients. Transplantation 1985; 39: 23-25. 7
639
Summary Until now, the only cancers that have been strongly associated with AIDS are Kaposi’s sarcoma and non-Hodgkin lymphoma. We used a linkage between AIDS (50 050 reports) and cancer (859 398 reports) registries in seven health departments in the USA to investigate the association between HIV infection and epidermoid anal cancer. We compared the numbers of observed cases and expected cases, calculated from general population rates with adjustment for age, sex, and race. The relative risk of anal cancer at and after AIDS diagnosis was 84·1 (95% Cl 46·4-152) among homosexual patients (11 cases) and 37·7 (9·4-151) among non-homosexual patients (2 cases). The relative risk of anal cancer up to 5 years before the AIDS diagnosis (23 cases) was also increased; it was 13·9 (6·6-29·2) in the period 2-5 years before AIDS and 27·4 (15·9-47·2) during the 2 years before AIDS diagnosis (p for trend=0·004). Among homosexual men, the relative risk of anal cancer was inversely related to age at AIDS onset (p for trend<0·001). Excess risks were found in all geographical areas.
This study establishes a strikingly increased risk of anal among people with AIDS. These data are consistent with a previously hypothesised association between HIVinduced immunodeficiency and anal cancer development, but because homosexual men were at increased risk of anal cancer even before the AIDS epidemic, we cannot say how much of the increased risk is attributable to HIV infection. Nevertheless, clinicians should be aware that AIDS patients have an increased risk of anal cancer. cancer
Lancet 1994; 343: 636-39
Epidemiology Science Centre, State Serum Institute, Copenhagen, Denmark (Prof M Melbye MD); Viral Epidemiology Branch, Division of Cancer Etiology (T R Coté MD, R J Biggar MD), Applied Research Branch, Division of Cancer Prevention and Control (L Kessler ScD), and Biostatistics Branch, Division of Cancer Etiology (M Gail MD), National Cancer Institute, Rockville, Maryland, USA The Danish
*
Members of the working group
are
listed at the end of the article.
Dr Timothy R Coté, Viral Epidemiology Branch, National Cancer Institute (NCI), EPN-434, 6130 Executive Blvd, Rockville, MD 20852, USA
Correspondence to:
636
Introduction Reports from Denmark, Sweden, and Connecticut, USA, have shown significant increases in the incidence of epidermoid anal cancer during the past 30 years.’-3 The increase has been more pronounced in women than in men and in urban than rural areas. Furthermore, black people are at higher risk than whites and never-married men at higher risk than ever-married men. These trends show that important behavioural and environmental changes were taking place before the beginning of the AIDS epidemic. Data from the Surveillance, Epidemiology and End Results Programme (SEER) in the USA have shown an especially rapid increase in epidermoid anal cancer incidence among men in the San Francisco Bay area.3 The incidence among these men was nearly 3 times higher than that among men living in other areas of the USA covered by the SEER programme or among Scandinavian men." Furthermore, the relative risk of anal cancer among never-married men (a surrogate marker for homosexuality) increased from 6-7 (95% CI 4-7-9-5) in 1979-84 to 10-3 (7-5-14-1) in 1985-89. Although these changes could reflect an influence by the growing AIDS epidemic, they could also be the result of an independent but contemporaneous influence of another aetiological agent with increasing prevalence among homosexual men during this period. Cervical cancer in an HIV-infected woman is now accepted as an AIDS-defining illness.4 Cervical and anal cancers share many characteristics;5-8 thus a link between AIDS and anal cancer is also plausible. In support of a direct influence by HIV and its associated immunodeficiency on anal cancer development, we and others have described anal intraepithelial abnormalities in a substantial proportion of HIV-positive homosexual men and shown a significant association between these abnormalities and decreasing immune competence.9-11 Although the anal lesions have not been shown to progress to invasive cancer, there is preliminary evidence of progression from mild more anal epithelial neoplasia among to severe immunocompromised homosexual men.12 Furthermore, among men with AIDS-related Kaposi’s sarcoma, most of whom probably were homosexual, there was a significant excess of anal cancer. 13 To investigate the possible association between the AIDS epidemic and the development of anal cancer, we have analysed data from a newly established registry that links AIDS and cancer surveillance systems in the USA.
Subjects and methods Until 1993, AID S registries in the United States recorded only two cancer types-Kaposi’s sarcoma and non-Hodgkin lymphoma. It was mandatory for these cancers to be reported to AIDS registries only when they were one of the initial AIDS-defining diseases; diagnoses of these cancers that followed an earlier AIDS diagnosis would not necessarily be recorded in AIDS registries. However, they would be registered by cancer registries whenever they occurred. To understand the relation between AIDS and cancer in
general, in 1992, one of us (TRC) set up a programme to link AIDS and cancer registries in nine regions of the USA/’* Between May and November, 1992, AIDS and cancer registries linked in California, Florida, Metropolitan Atlanta, and New Jersey. All registered individuals were followed until death or 2-25 years after the AIDS diagnosis, whichever came first. This restriction was applied to limit problems associated with loss to follow-up (eg, unregistered deaths or migration away from a region after a diagnosis of AIDS but before a cancer diagnosis); 2 25 years represents the average lifespan from AIDS to death. Only people younger than 70 years were included in the linkage. Because some cancer registries were established earlier or later than the AIDS registries, we included only periods in which both registries were were
Table 1: Relative risk (observed/expected ratio) of epidermoid anal and anorectal cancer among AIDS patients compared with population controls matched for age, sex, and race cancers
(two cancers adjacent to the anorectal area not thought to be
operating. Expected incident anal cancer cases after AIDS diagnosis were calculated by multiplying SEER age-specific incident rates by the corresponding person-years at risk after AIDS diagnosis. A person was at risk until anal cancer, death or censoring at 2 25 years, or
associated with HIV/AIDS).
until cancer surveillance ended, whichever came first. The observed number of cases was then compared with the expected number to give observed/expected ratios (relative risk) and Poisson-distributed 95% CI.15 The frequency of anal cancer preceding AIDS was first known from the cancer registries. We provide an estimate of the interval-specific relative risk. However, calculating the expected rates was complicated because HIV-infected subjects who develop anal cancer might die before progression to AIDS and therefore would never be matched in the AIDS registry. To estimate the expected number of anal cancer cases that would have occurred among AIDS cases before their AIDS diagnosis (assuming that HIV-infected individuals have the same cancer rates and survival as the general population) we used modifications of published techniques.16 This approach estimates the probability of a person’s having had anal cancer diagnosed and surviving the necessary time to have developed AIDS, with incidence and survival rates from SEER data specific for age, sex, and period.1s We acknowledge that this approach is approximate. If HIV infection accelerates death from anal cancer, we may have underestimated the risk; conversely, if anal cancer accelerates the development of AIDS, we may have overestimated the risk. No information about these effects is available. Anal cancer is usually treated by surgery or local radiotherapy, which is unlikely to affect the course of HIV-related
The linkage analysis included 859 398 reports of cancer and 50 050 reports of AIDS. We found 39 AIDS patients with a diagnosis of epidermoid anal cancer (anal in 31, anorectal in 8), 3 of whom had their cancer diagnosed more than 5 years before AIDS was diagnosed. 21 other patients with anal-site cancers of other types were excluded, including 13 with malignant lymphoma (1Burkitt’s, 9 immunoblastic, 2 large-cell, 1 mixed-cell), and 8 with Kaposi’s sarcoma. All but 1 of the 39 patients were men. 36 were homosexual or bisexual, and 6 of these were also intravenous drug users; 2 others were intravenous drug users who did not report homosexual behaviour. For 1 case, risk behaviour was not known. To find out how serious these tumours were clinically thought to be, we examined treatment records for 30 patients; records for the patients diagnosed in Los Angeles were not available. No case was diagnosed at necropsy. Of the 30 patients with available treatment information, 21 had received radiotherapy or chemotherapy and 4 others had undergone operations. 3 of the 5 untreated patients died soon after the cancer diagnosis (10 days-3 months). Thus only 2 patients were untreated for unknown reasons. The relative risk of epidermoid anal cancer was 63 4 (95 % CI 36-8-109) after a diagnosis of AIDS compared with the general population. The relative risk was 841 (46 4—152) among homosexual or bisexual male patients and 37-7 (9 4-151 ) among non-homosexual patients. We found a 27-4 relative risk for anal cancer for the 2 years before an AIDS diagnosis and a somewhat lower risk (13-9) in the 2-5 years before AIDS diagnosis (table 1). Table 2 gives the observed/expected ratios for anal cancer after AIDS by age group and geographical distribution for homosexual men. Similar analyses for other risk groups were not possible because of small numbers. The relative
immunosuppression. To estimate the proportion of the general population of survivors to the time of AIDS diagnosis who had anal cancer diagnosed in a particular earlier time period, we used incidence data specific for sex, age, and period to count the number of cases incident in each month during the 5 years before the attained age at AIDS diagnosis. We then applied observed survival estimates to the cases lost to follow-up to calculate the fraction of cases still alive at the attained age. For each of the 60 months before AIDS, this estimate added to the number of cancer cases who were known to have survived to the attained age, and the sum was divided by the estimated population total to arrive at the proportion with an anal cancer diagnosed in an earlier month who survived to the attained age of AIDS diagnosis. To calculate the expected cancer counts in each month among patients in the AIDS registry, these proportions were multiplied by the corresponding person-years among AIDS cases. These monthly expected counts were then summed into longer periods before AIDS diagnosis and compared with observed counts for the same period. Tests for trend in relative risk were based on a score statistic for a Poisson model. Only cancers of the epidermoid type were studied in detail. This definition excluded AIDS-defining cancers (eg, Kaposi’s sarcoma and non-Hodgkin lymphoma) located in the anal region and focused the study on the histological type of tumour that has previously been reported to undergo significant changes.1-3 We used topography codes for epidermoid anorectal cancers 1541, 1542, 1543, and 1548, and histology codes 80513-23, 80703-63, 80903-53, and 81203-43, according to the International Classification of Diseases for Oncology (ICD-O).18 To seek possible ascertainment bias, a similar analytical approach was taken to assess the association with colon and prostate was
lymphoma histologies
were
Kaposi’s sarcoma and non-Hodgkin
again excluded.
Results
*Of anal cancer in homosexual men after AIDS diagnosis. tp for trend 0 0003. tFor San Francisco only 0/E= 134 (50 4-358).
Table 2: Relative risk of epidermoid anal homosexual men with AIDS
cancer
among
637
risk of developing anal cancer was highest for homosexual AIDS patients aged 20-29 years and was inversely correlated with age (p for trend 00003). Excess risks were found in all geographical areas. The relative risks of colon and prostate cancer were 0-7 and 22 at time of AIDS diagnosis and 0-6 and 1 -4 after the AIDS diagnosis.
Discussion This study documents an increased risk of anal cancer after AIDS diagnosis. Although the relative risks we report are high, especially for young homosexual men, they are based on only few cases. Therefore, the nature of the recordlinkage data needs to be carefully considered. We may have overestimated the relative risk. Mismatched cases could inflate apparent cancer risk. However, in a validation study of our linkage method in Los Angeles, all records linking AIDS and cancer diagnoses were reviewed by direct examination of the information at each registry. Only 9 (0-3%) of 2646 records were incorrectly linked (D Deapen, unpublished). All cancers in our study were histologically confirmed as invasive epidermoid cancers. Kaposi’s sarcoma and non-Hodgkin lymphoma are rarely found as primary tumours of the anorectal area; therefore, our cases were unlikely to be miscoded as these disorders. Nearly all of the cases found were treated with radiotherapy, chemotherapy, or surgery; we therefore doubt that the excess was due to misclassified trivial in-situ dysplasia. We were concerned that the excess might have been attributed to research interest in anal cancer in San Francisco, since this area has an active anal cancer research programme. 11,12 However, there was only a modest excess (1-33) of anal cancer diagnoses in San Francisco AIDS patients compared with those of other areas.
The increased medical attention given to AIDS patients probably created ascertainment bias, especially at the time of the initial AIDS diagnosis. This factor may be especially important for malignant disorders that have long latent periods and can therefore be detected through screening. However, we found little ascertainment bias around the time of AIDS diagnosis for two adjacent sites not suspected to have AIDS-associated malignant disorders (colon and
prostate). It is also possible that we underestimated the actual risk. For example, there may have been unreported cases or missed linkages. Furthermore, screening as part of examinations at the time of AIDS diagnosis could have detected cancers sooner than if the patient had not had AIDS. This factor would inflate the apparent risk at the time of AIDS, but it would also lower the observed rate in the follow-up period. In addition, AIDS has a short survival duration, and unregistered deaths (such as those lost
through migration)
will
spuriously
increase the
apparent size of the population under study, thus lowering calculated relative risks. The most important factor to take into account is that homosexual men were known to have an increased risk of anal cancer even before the AIDS epidemic.2,19-23 In Holly et al’s21 case-control study, homosexual men were 12 times more likely to develop anal cancer than age-matched non-homosexual men.21 That study was carried out in the San Francisco area during the 1980s and could have included homosexual men who were HIV-infected, which might have increased their risk. The relative risk we 638
recorded for homosexual men with AIDS was much higher (84-1). However, the subgroup of homosexual men at highest risk of AIDS may also have had the highest exposure to the unknown factors that increase the risk of anal cancer among homosexual men generally. Consistent with the hypothesis of an HIV association, we also found a significant excess risk of anal cancer in AIDS patients not identified as homosexual. However, this finding was based on only 2 patients (1male,1 female) and it is possible that the man, although reported as being an intravenous drug user, could have had misclassified exposure risks. We used a modification of a previously published method16 to estimate the risk of anal cancer in the 5-year period before the patients’ AIDS diagnoses. Although our estimates must be interpreted with caution, there was an increasing trend in the relative risk towards the time of AIDS diagnosis. Thus, the high risk we report could be due to HIV-induced immunodeficiency. This analysis cannot define what proportion of this risk, if any, is attributable to HIV and what proportion is due to a homosexual lifestyle. An aetiological association with HIV is biologically plausible and has long been suspected.24 In a study of severely immunosuppressed men with AIDS or other serious HIV disease, 15 % had anal epithelial neoplasia and 2% carcinoma-in-situ.11 Other studies of homosexual men have found high prevalence rates of anal epithelial atypia and dysplasia and have documented an association between HIV infection and increased prevalence and quantity of human papillomavirus (HPV) in anal samples.9.10 Such studies have also found significant associations between the presence of HPV and anal epithelial abnormalities9.1o.2s and shown anal epithelial abnormalities to be more frequent in men with both HIV and HPV than in those with either virus alone. Previous studies have shown an association between anal and cervical cancer and suggested a similar pathogenesis.S-8 Invasive cervical cancer is now included as an AIDSdefining illness in HIV-positive individuals, primarily because of the plausibility of an association.4However, the evidence for an association between invasive cervical cancer and HIV is weak,26 especially because the impact of confounding has not been adequately explored. In a preliminary analysis of linked records from these AIDS and cancer registries, invasive cervical cancer was only marginally increased over background.27 We propose two reasons why the relative risk of cervical cancer in AIDS patients is lower than that of anal cancer. First, cervical cancer is much more common than anal cancer in the general population, so small changes are more difficult to detect. Second, cervical dysplasia is detected by active screening programmes among HIV-positive women, which reduces the likelihood of progression to invasive cervical cancer. HIV-infected women have, in general, a higher rate of sexually transmitted diseases than other women 211 and will therefore be in closer contact with the health care system both before and after their HIV infection. The mechanism for a putative association between anal cancer and HIV remains unclear but the cancer types reported to be associated with AIDS are similar to those common to other forms of immunodeficiency.26 Excesses of anogenital cancers, including both cervical and anal cancer, have also been reported among renal transplant patients.29.30 So far, only a few cancer types have been associated with the HIV epidemic. This feature argues against the concept of general immune surveillance of
defence mechanism for cancer development. However, the tumours associated with the HIV epidemic are suspected to have a viral oncogenesis. Therefore, a potential influence for HIV-related immunodeficiency on tumour development could be through failing intracellular control of persisting viral
cellular control
as
a
genomes.
Clinicians dealing with HIV-infected homosexual men should be aware of an increased risk of anal cancer in these
patients. Fortunately, digital examination of the anorectal region is standard medical practice. If anal cancer is suspected on clinical grounds, assessment for cytological abnormalities is simple and inexpensive and, when necessary, treatment is often effective.
Implications for anal cancer screening programmes are complex. Although the relative risk of invasive anal cancer is increased, the absolute risk of this cancer in the group of homosexual men with AIDS we studied was about 1 per 1000 cases. However, anal dysplasia (including carcinomain-situ) may occur in up to 15 % of severely immunosuppressed HIV-infected homosexual men without signs or symptoms of anal cancer. Apparently, most AIDS patients diagnosed with anal dysplasia do not develop clinically significant invasive anal cancers before they die, at least with the current outlook for AIDS patients. Therefore, any recommendation to institute screening for anal dysplasia in symptom-free individuals will need to be based on a careful cost-benefit analysis. This study was supported by the National Cancer Institute and by a grant from the Danish Medical Research Council (J nr 5.22.98.62). We thank Ms Frances Yellin (ARC Professional Services, Rockville) for programming assistance; Cliff Notes Inc for the first name thesaurus used in the linkage; and Dr William Blattner, Dr Charles Rabkin, and Dr James J Goedert for helpful comments on the manuscript.
Members of the AIDS jCancer Working Group (for each city or state, the first member listed is from the AIDS registry and the second from the cancer registry). San Francisco (George Lemp, Dee West); California (Jim Singleton, John Young); Los Angeles (Peter Kerndt, Dennis Deapen); San Diego (Michelle Ginzberg, Hoda Anton-Culver); Florida (Spencer Lieb, Richard Hopkins); Georgia (Brian Williams, Jonathan Liff); New Jersey
(Douglas Morgan, William Parkin).
References 1
2 3
Goldman S, Glimelius B, Nilsson B, Paahlman L. Incidence of anal epidermoid carcinoma in Sweden 1970-1984. Acta Chir Scand 1989; 155: 191-97. Frisch M, Melbye M, Moller H. Trends in incidence of anal cancer in Denmark. BMJ 1993; 306: 419-22. Melbye M, Rabkin CS, Frisch M, Biggar RJ. Changing patterns of anal cancer incidence in the USA, period 1940-89. Am J Epidemiol (in
press). CDC. 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR 1992; 41/RR-17: 1-19. 5 Scholefield JH, Sonnex C, Talbot IC, et al. Anal and cervical intraepithelial neoplasia: possible parallel. Lancet 1989; ii: 765-69.
4
6
Melbye M, Sprogel P. Aetiological parallel between anal cancer and cervical cancer. Lancet 1991; 338: 657-59.
Rabkin CS, Biggar RJ, Melbye M, Curtis RE. Second primary cancers following anal and cervical carcinoma: evidence of shared etiologic factors. Am J Epidemiol 1992; 136: 54-58. 8 Scholefield JH, Hickson WGE, Smith JHF, Rogers K, Sharp F. Anal epithelial neoplasia: part of a multifocal disease process. Lancet 1992; 340: 1271-73. 9 Melbye M, Palefsky J, Gonzales J, et al. Immune status as a determinant of human papillomavirus detection and its association with anal epithelial abnormalities. Int J Cancer 1990; 46: 203-06. 10 Caussy D, Goedert JJ, Palefsky J, et al. Interaction of human immunodeficiency virus and papillomaviruses: association with anal epithelial abnormality in homosexual men. Int J Cancer 1990; 46: 214-19. 11 Palefsky JM, Gonzales J, Greenblatt RM, et al. Anal intraepithelial neoplasia and anal papillomavirus infection among homosexual males with group IV disease. JAMA 1990; 263: 2911-16. 12 Palefsky JM, Holly EA, Gonzales J, Lamborn K, Hollander H. Natural history of anal cytologic abnormalities and papillomavirus infection among homosexual men with group IV disease. J Acq Immun Defic Syndr 1992; 5: 1258-65. 13 Biggar RJ, Curtis RE, Coté TR, Rabkin CS, Melbye M. Risk of other cancers following Kaposi’s sarcoma: relationship to the acquired immunodeficiency syndrome. Am J Epidemiol 1994; 139: 362-68. 14 Coté TR. NCI plans linkage of AIDS and cancer registries. J Natl Cancer Inst 1993; 85: 1113-14. 15 Breslow NE, Day NE. Statistical methods in cancer research. Vol II: the design and analysis of cohort studies. Lyon: IARC, 1987: 48-79. 16 Feldman AR, Kessler L, Myers MH, Naughton MD. The prevalence of cancer: estimates based on the Connecticut Tumor Registry. N Engl J Med 1986; 315: 1394-97. 17 Ries LG, Pollack ES, Young JL. Cancer patient survival: surveillance, epidemiology, and end results program, 1973-79. J Natl Cancer Inst 1983; 70: 693-707. 18 WHO. ICD-O. International classification of diseases for oncology. Geneva: WHO, 1980. 19 Peters RK, Mack TM. Patterns of anal carcinoma by gender and marital status in Los Angeles County. Br J Cancer 1983; 48: 629-36. 20 Daling JR, Weiss NS, Hislop TG, et al. Sexual practices, sexually transmitted diseases, and the incidence of anal cancer. N Engl J Med 1987; 317: 973-77. 21 Holly EA, Whittemore AS, Aston DA, et al. Anal cancer incidence: genital warts, anal fissure or fistula, hemorrhoids, and smoking. J Natl Cancer Inst 1989; 81: 1726-31. 22 Scholefield JH, Jones HT, Cuzick J, Northover JMA. Anal cancer and marital status. Br J Cancer 1990; 62: 286-88. 23 Adam YG, Efron G. Current concepts and controversies concerning the etiology, pathogenesis, diagnosis, and treatment of malignant tumors of the anus. Surgery 1987; 101: 253-66. 24 Frazer IH, Medley G, Crapper RM, Brown TC, Mackay IR. Association between anorectal dysplasia, human papillomavirus, and human immunodeficiency virus infection in homosexual men. Lancet 1986; ii: 657-60. 25 Kiviat N, Rompalo A, Bowden R, et al. Anal human papillomavirus infection among human immunodeficiency virus seropositive and seronegative men. J Infect Dis 1990; 162: 358-61. 26 Rabkin CS, Blattner WA. HIV infection and cancers other than non-Hodgkins lymphoma and Kaposi’s sarcoma. Cancer Surveys 1991; 10: 151-60. 27 Coté TR, Schiffman M, Biggar RJ, et al. Invasive cervical cancer among women with AIDS: results of registry linkage. IX International Conference on AIDS, Berlin 1993, Abstr no PO-B14-1637. 28 Wasserheit JN. Epidemiological synergy. Interrelationships between human immunodeficiency virus infection and other sexually transmitted diseases. Sex Transm Dis 1992; 19: 61-77. 29 Penn I. Cancers of the anogenital region in renal transplant recipients. Cancer 1986; 58: 611-16. 30 Blohmé I, Brynger H. Malignant disease in renal transplant patients. Transplantation 1985; 39: 23-25. 7
639