Risk of cancer in people with AIDS

Objectives: To determine whether the incidence of cancers other than the AIDS-defining cancers is increased in people with AIDS, and to determine if cancer incidence increases with time, a surrogate marker of declining immune function.

Design: Register-based retrospective cohort study in New South Wales, Australia. Age-, sex-, and period-adjusted standardized incidence ratios (SIR) were calculated for individual cancers occurring in 1980-1993 in people with AIDS registered before 1996.

Results: During the study period, 3616 people were registered with AIDS; 716 cases of AIDS-defining cancer and 62 cases of non-AIDS-defining cancer were identified. People with AIDS had a significantly increased incidence of Hodgkin‚s disease [SIR 18.3; 95% confidence interval (CI) 8.39-34.8], multiple myeloma (SIR 12.1; 95% CI 2.50-35.4), leukaemia (SIR 5.76; 95% CI 1.57-14.7), lip cancer (SIR 5.94; 95% CI 1.92-13.8) and lung cancer (SIR 3.80; 95% CI 1.39-8.29). The incidence of Hodgkin‚s disease increased significantly around the time of AIDS diagnosis (P=0.008 for trend with time), suggesting an association with immunodeficiency.

Conclusions: This study provides strong support for the hypothesis that Hodgkin‚s disease is an AIDS-associated condition. There was an increased incidence of several other forms of cancer, some of which are known to occur at increased rates in transplant recipients who have received immunosuppressive therapy. Improved survival in people with HIV infection may lead to increases in the number that develop these forms of cancer.

Introduction

The severe immunodeficiency caused by advanced HIV infection has been recognized as capable of causing three types of malignancy: Kaposi‚s sarcoma (KS), non-Hodgkin‚s lymphoma (NHL) and cervical cancer. Although these are the only forms of cancer that have been designated as ‚AIDS-defining‚ [¹], several other malignant diseases have been reported to occur more frequently following HIV infection than in its absence, including Hodgkin‚s disease [^2,3], anal cancer [⁴] and squamous cell carcinoma of the conjunctiva [⁵]. Medical immunosuppression has been shown to cause some of the same forms of cancer that occur in people with HIV infection, as well as causing an increased incidence of squamous cell carcinomas of the skin [⁶], lip [⁷] and anogenital region [⁶]. A recent study found increased rates of many other cancer types in transplant recipients [⁷] and raised the possibility that immunosuppression might increase the risk of a wide range of malignant disease.

In this study, we assessed cancer risk in a large population-based cohort of people with AIDS in New South Wales, Australia, with a particular focus on its relationship with severity of immune deficiency as reflected by proximity to AIDS diagnosis.

Methods

A retrospective cohort study using linkage between AIDS and cancer registers in New South Wales was performed. The details of the linkage procedure and its validation have been described in detail previously [⁸].

Linkage procedure

In New South Wales, AIDS has been a notifiable condition since 1984 [⁹], and notification of cancer is also a statutory requirement [¹⁰]. All cases of AIDS registered between 1984 and 1995 were matched against all cases of cancer registered between 1980 and 1993. To comply with the New South Wales Public Health Act, linkage was performed without names, using data only on name code (first two letters of first and last name), date of birth and sex. Linkage was assessed using the computer program AUTOMATCH [¹¹]. A match was accepted if there was an exact match on all three fields, or a near match supported by consistency between the registers in dates of death and area of residence [⁸]. The linkage procedure was approved by the New South Wales Cancer Council and the Confidentiality of Health Information Committee of the New South Wales Health Department.

The matching procedure had been previously validated for NHL [⁸]. Completeness of registration of KS to the cancer register was estimated as the proportion of cases of KS registered as the first AIDS-defining illness on the AIDS register between 1984 and 1993 that were also registered with the cancer registry. Cases of non-cutaneous KS reported to the cancer register for which there was no apparent match on the AIDS register were assumed to be failures of AIDS registration or matching, as non-AIDS KS rarely presents at non- cutaneous sites [¹²].

Calculation of standardized incidence ratios

The follow-up period was from the beginning of 1980 or from 5 years prior to AIDS diagnosis (whichever occurred later) until the date of any cancer diagnosis, death or the end of 1993. For each type of cancer, the observed number of cases in people who developed AIDS was compared with the expected number based on age- and sex-specific annual cancer incidence rates in New South Wales [¹⁰] to give a standardized incidence ratio (SIR). Confidence intervals (CI) and tests for trend were based on the Poisson distribution [¹³]. For KS, because most cases in the study were AIDS related, expected numbers were based on age- and sex-specific average annual rates for the period 1975-1979, prior to the study period.

Calculation of the expected numbers of cancers for time periods preceding AIDS was complicated by the fact that some people with HIV infection may have developed cancer and died prior to an AIDS diagnosis, leading to an undercount of observed cases. Expected numbers of cancers were, therefore, adjusted to allow for this possibility by assuming that HIV infection has no effect on survival rates following cancer diagnosed prior to AIDS [⁴] and that South Australian cancer survival rates [¹⁴] applied in New South Wales. Using this approach, estimates are obtained of the probability that a person with HIV infection who is diagnosed with cancer survives long enough to develop AIDS.

To assess whether there was any association between cancer risk and increasing immunodeficiency, as indicated by proximity to diagnosis of AIDS, SIR was calculated for defined time periods before and after AIDS diagnosis. This analysis was confined to those forms of cancer with more than five cases or with significantly raised overall incidence.

Results

In total, 716 cases of AIDS-defining cancer and 62 cases of non-AIDS-defining cancer occurring between 1980 and 1993 were identified in 3616 people with AIDS by assessing linkage with the cancer registry. The median age at onset of AIDS was 37 years, and 96% of people with AIDS were male, of whom 91% reported homosexual contact.

Completeness of registration for Kaposi‚s sarcoma

Linkage analysis identified 511 cases of KS. In total, 528 people were reported to the AIDS registry with KS as their first AIDS-defining illness during 1980-1993, of whom 358 (68%) were reported to the cancer registry as having KS. Twelve cases of KS occurring at non-cutaneous sites during 1980-1993 were recorded on the cancer register but were not present on the AIDS register. Six cases were in the oral cavity, two occurred elsewhere in the gastrointestinal tract and four were at an unspecified site.

AIDS-defining cancers

Compared with the New South Wales population as a whole, the incidence of KS in people with HIV was increased by a factor of 72 695 and the risk of NHL was increased 97.3-fold (Table 1). There were no cases of cervical cancer reported in women with AIDS.

Table 1

Non-AIDS-defining cancers

For all non-AIDS-defining cancers combined, the SIR was 3.00. Significantly increased incidence was found for several forms of cancer that are not AIDS defining (Table 1). These were lip cancer (SIR 5.94), lung cancer (SIR 3.80), connective tissue cancer (SIR 9.21), Hodgkin‚s disease (SIR 18.3), multiple myeloma (SIR 12.1) and leukaemia (SIR 5.76). Available information on the histological subtype of these cancers is summarized in Table 2. The SIR values unadjusted for the decreased survival of people with cancer were also significantly raised for all the above forms of cancer, with the exception of lung cancer (unadjusted SIR 1.72; 95% CI 0.631-3.75).

Table 2

For Hodgkin‚s disease, the increased incidence was confined to the period immediately prior to and after AIDS diagnosis (P=0.008 for trend with time). For the other forms of cancer, there was no association between increased immunodeficiency and risk of cancer (Table 3).

Table 3

Discussion

This study has found increased rates in people with HIV infection of several forms of cancer that are not AIDS defining. It is the first population-based study to report significantly increased rates of lip cancer and supports a recent report that found increased rates of multiple myeloma and leukaemia in people with HIV [¹⁵]. Hodgkin‚s disease has been reported to occur at a higher incidence in people with HIV infection in several studies [^2,3,16-18], but this study is only the second to demonstrate that the incidence increases with the degree of immunodeficiency [¹⁵].

The accuracy of our estimation of cancer risks depended on the completeness of cancer and AIDS registration, and the accuracy of linkage. In this respect, the largest inaccuracy was in reporting of KS to the cancer register. Only 68% of KS cases that occurred as the first AIDS-defining illness were reported on the cancer register. This may have been related to confusion over whether cutaneous KS is a cancer requiring registration, as other non-melanoma skin cancers do not require registration in New South Wales. In contrast, we have previously shown that 95% of AIDS-related NHL in Sydney was reported to the cancer register [⁸], suggesting that the serious degree of under-registration was probably confined to KS. Our data suggest that AIDS registration was 98% complete. There were only 12 cases of non-cutaneous KS, which may have been AIDS-related, that were present on the cancer register but not the AIDS register, compared with 511 cases that were linked (2% of linked cases). Previous validation for NHL showed that linkage was 99% sensitive and 100% specific in identifying cases of NHL occurring in people who had been reported to both registers [⁸]. Cases of cancer occurring in HIV-infected individuals who have not yet developed AIDS were not identified by this linkage.

Apart from a real association between HIV and cancer development, it is important to consider the possibility that increased medical surveillance in people with HIV may have led to the diagnosis of cancers that would otherwise remained inapparent. This effect would be expected to be greatest for those forms of cancer with a long latency that may be detected by screening and for those that are most clinically apparent, such as cutaneous tumours. The absence of increased risk for melanoma argues against increased surveillance as a major bias in this study, although it is possible that it may be partially responsible for the small increases in incidence seen across a wide range of cancer types.

The method of adjustment for decreased survival after cancer diagnosis was approximate and could have resulted in either under- or overestimates of the expected numbers of cancers. If HIV infection had, in fact, resulted in poorer survival following diagnosis of a cancer, then the method would give underestimates of the relative risk. If diagnosis of cancer resulted in faster progression to AIDS, then the relative risk would be overestimated. Relative risks, both adjusted and unadjusted, were significantly increased for lip cancer, Hodgkin‚s disease, myeloma and leukaemia.

Rates of lip cancer were clearly increased overall in people with HIV infection but did not increase with more advanced immunodeficiency. Although not previously reported in association with HIV, lip cancer does occur at greatly increased rates in transplant recipients [^6,7]. Some of the cases reported in this study may have been misclassified squamous cell carcinoma of the skin, which also occurs at increased rates in transplant recipients [⁶] but is not recorded by the New South Wales Cancer Registry.

There have been numerous case reports of myeloma and leukaemia in people with AIDS [¹⁹] but only one report of significant increases in incidence in people with AIDS [¹⁵]. Human herpes virus 8 infection was linked with multiple myeloma in one study [²⁰], although this finding has not been replicated. In the present study, two of the three men with myeloma developed KS within a few months of registration of myeloma.

The increased rate we found for cancers of connective tissue is probably a result of misdiagnosis of KS as haemangiosarcoma, which may histopathologically resemble KS [²¹]. Both men with haemangiosarcoma were later registered as developing KS.

Increased incidence of Hodgkin‚s disease has been previously associated with HIV, both in cohort studies [^2,3,16-18] and ecological studies [²²]. The present study adds support to a recent large population-based linkage analysis that reported increased incidence of Hodgkin‚s disease and that the risk of Hodgkin‚s disease was greater in people with advanced immunodeficiency, as indicated by proximity to the time of AIDS diagnosis [¹⁵]. The available histological data suggest that these cases were not misclassified NHL. Of the six cases for which histological subtype was available, four were of histologies previously described as predominating in HIV-related Hodgkin‚s disease [²³]. These data indicate that Hodgkin‚s disease is associated with immunodeficiency in people with HIV infection and that it should probably qualify as an AIDS-defining illness.

No significantly increased rates were found for the two most common cancer types in young men: testicular cancer and melanoma. Based on the 95% CI, it was possible to exclude a relative risk as low as 2.8 for melanoma and 4.9 for testicular cancer. For neither of these forms of cancer was there any association with immunodeficiency.

Until recently, most studies of cancer in transplant recipients have suggested that only certain cancers, mainly those associated with viral infection, occur at increased rates in immunosuppressed people. A recent large study of transplant recipients has suggested that a wide range of cancer types may occur at increased rates in immunosuppressed people [⁷]. Our study is the first to find increased rates of lip cancer, which is strongly associated with iatrogenic immunosuppression, in people with AIDS. As treatments for AIDS improve, leading to prolonged survival of immunodeficient individuals, it is possible that other cancers occurring at increased rates will be identified.