Non-acquired immune deficiency syndrome-defining malignancies in a hospital-based cohort of human immunodeficiency virus-infected patients: Bordeaux, France, 1985-1991. Groupe d'Epidémiologie Clinique du SIDA en Aquitaine

The role of the human immunodeficiency virus (HIV) in carcinogenesis is not clearly understood (1). Two types of malignant tumors are included in the Centers for Disease Control (CDC) definition of acquired immunodeficiency syndrome (AIDS) (2): Kaposi’s sarcoma and non-Hodgkin’s lymphoma. However, researchers have suggested that other malignancies might occur with an increased frequency in HIV-infected patients (3–11).

In this communication, we describe and estimate the incidence of and mortality from non–AIDS-defining malignancies in a hospital-based cohort of HIV-infected patients, regardless of the clinical stage of disease or the category of disease transmission.

Data for this study were obtained from the Groupe d’Epidémiologie Clinique du SIDA en Aquitaine (GECSA), a hospital-based surveillance system created in 1987 for reporting HIV infection and follow-up of patients. Aquitaine, a region in southwest France, ranks third among 22 French regions for reports of AIDS (12). Adults, aged 15 years or over, are included in GECSA after they give written informed consent. At each follow-up consultation or hospital admission, physicians complete anonymous standardized questionnaires (13).

Non–AIDS-defining malignancies were reported to the GECSA through a system of passive surveillance, that is, with spontaneous notification. For all reported non–AIDS-defining malignancies, we reviewed hospital records and interviewed the physicians of the patients. Incident cancers were defined as primary non–AIDS-defining malignancies diagnosed during the follow-up, excluding Kaposi’s sarcoma, non-Hodgkin’s lymphoma, carcinoma in situ, and non–AIDS-defining malignancies present when the patient entered the cohort. All incident cancers were confirmed by histopathology.

Statistical comparisons were made by computing Poisson distribution-based standardized incidence ratios and standardized mortality ratios, which are ratios of observed to expected numbers of events (14). Expected numbers of cancers and deaths by cancer, given the amount of follow-up in the cohort, were obtained by indirect standardization, that is, by multiplying the number of person-years of follow-up of HIV-infected individuals (specific for gender and age) by gender- and age-specific reference rates (per 100000 per year) (15). These reference rates were the 1978–1982 cancer incidence rates in the general French population (16) and the 1987–1990 cancer mortality rates in Aquitaine (17).

By May 1, 1991, 1899 HIV-infected patients had been included in the GECSA. Among those patients, 177 cancers have been reported: 112 Kaposi’s sarcomas, 38 non-Hodgkin’s lymphomas, four carcinomas in situ, and 23 non–AIDS-defining malignancies. Nine non–AIDS-defining malignancies were present at entry in the cohort (including three cases of Hodgkin’s disease), leaving 14 incident cases for this analysis (Table 1).

Table 1.

Characteristics of incident non–AIDS-defining malignancies, Groupe d’Epidémiologie Clinique du SIDA en Aquitaine, Bordeaux, France, 1985–1991

Patient No.	Type and location of tumor*	Age, y†	Transmission category	CD4+ lymphocytes, mm3†	Other AIDS-related events†,‡	Date of diagnosis, mo/y		Date of death, mo/y
						HIV infection	Non–AIDS-defining malignancy
Women
1	Breast adenocarcinoma, grade II	27	Blood recipient	154	None	2/88	8/89	Alive
2	Breast adenocarcinoma, grade II	30	Intravenous drug user	754	None	12/87	12/90	Alive
3	Breast adenocarcinoma, grade II	60	Unknown	6	Toxo, CMV, oral thrush	5/89	12/90	Alive
4	Breast adenocarcinoma, grade II	65	Heterosexual	235	None	10/90	4/91	Alive
5	Urothelial papilloma	43	Heterosexual	236	None	2/86	5/86	8/86
Men
6	Astrocytoma	43	Homosexual	407	None	2/89	4/90	4/90§
7	Hodgkin’s type III, stage IV	26	Intravenous drug user	10	None	1/88	1/90	3/90
8	Hodgkin’s type III, stage IV	27	Intravenous drug user	397	None	7/86	1/90	Alive
9	Hodgkin’s type III, stage II	27	Blood recipient	255	None	12/87	9/90	Alive
10	Hodgkin’s type III, stage IV	33	Homosexual	343	None	9/88	8/89	10/91§
11	Hodgkin’s type III, stage II	49	Blood recipient	253	None	7/88	9/91	Alive
12	Kidney adenocarcinoma	55	Homosexual	426	None	7/89	1/90	5/90§
13	Small-cell lung carcinoma	50	Heterosexual	15	TB, cand, herpes zoster	1/87	12/88	8/89§
14	Generalized adenocarcinoma	45	Homosexual	49	PCP, HSV	5/88	9/90	10/90§

^*Grading and staging systems for breast adenocarcinoma and Hodgkin’s disease are per (a) Bloom HJG, Richardson WW: Histological grading and prognosis in breast cancer. Br J Cancer 11:359–377, 1957; (b) American Cancer Society and National Cancer Institute: Staging in Hodgkin’s disease. Cancer Res 31:1707–1870, 1971; and (c) Lukes RJ: Criteria for involvement of lymph node, bone marrow, spleen, and liver in Hodgkin’s disease. Cancer Res 31:1755–1767, 1971.

^†At diagnosis of non–AIDS-defining malignancies.

^‡Toxo = central nervous system toxoplasmosis; CMV = cytomegalovirus infection; TB = tuberculosis; cand = esophageal candidiasis: PCP = pneumocystosis; HSV = herpes simplex virus infection.

^§Death attributable to non–AIDS-defining malignancy.

All cancers were diagnosed at the Bordeaux University hospital. At the time the non–AIDS-defining malignancies were diagnosed, 11 patients were asymptomatic [group II of the CDC classification (18)], and three had AIDS.

The risk of non–AIDS-defining malignancies, adjusted for age and gender, was 5.4 times higher for the whole cohort than in the general population in France (Table 2). The gender-specific standardized incidence ratios were 4.4 in males and 9.0 in females. When patients with Hodgkin’s disease were excluded, the overall standardized incidence ratio was 3.4 (95% confidence interval = 1.5–6.4; P<.001).

Table 2.

Standardized incidence ratio and standardized mortality ratio for non–AIDS-defining malignancies in 1899 HIV-infected patients: Groupe d’Epidémiologie Clinique du SIDA en Aquitaine, Bordeaux, France, 1985–1991^*

Age, y	Gender	Person-years of follow-up	Cancer incidence rate,* France, 1978–1982	Observed No. of non–AIDS-defining malignancies	Cancer-related mortality rate,† Aquitaine, 1987–1990	No. of deaths from non–AIDS-defining malignancies
15–24	Men	375.8	23	0	6	0
	Women	211.5	17	0	4	0
25–34	Men	852.5	42	4	14	0
	Women	327.5	46	2	11	0
35–44	Men	388.8	118	1	52	2
	Women	68.8	154	1	41	0
45–54	Men	112.8	446	3	208	2
	Women	18.3	354	0	116	0
55–64	Men	44.4	867	1	600	1
	Women	26.8	541	1	246	0
65–74	Men	14.2	1547	0	1053	0
	Women	0.5	768	1	426	0
75–84	Men	2.3	1944	0	2043	0
	Women	5.3	921	0	889	0
All men. . . . . . . . . . . . . . . . . . . . . . . . .		1790.8	447	9	367	5
All women. . . . . . . . . . . . . . . . . . . . . .		658.7	327	5	195	0
Standardized incidence ratio = 5.4			95% confidence interval = 2.9–9.0		P<.001‡
Standardized mortality rate = 4.0			95% confidence interval = 1.3–9.3		P<.05‡

^*Of the 1899 patients, 1381 were men and 518 were women.

^†Carcinoma in situ and skin cancers excluded. Rates per 100 000 and per year.

^‡Poisson distribution.

Of the 14 patients we studied, seven died. Five of those deaths, all in men, were attributed to non–AIDS-defining malignancies (Table 1). The risk of death by non–AIDS-defining malignancies was 4.0 times higher among the group we analyzed than among the general population of Aquitaine (Table 2).

This study is the first that quantified the increased risk of non–AIDS-defining malignancies in a prospective cohort and included all transmission categories of HIV infection.

As in all cohort studies using standardization, a major design issue was the appropriateness of the comparison with an external reference population (15). Mortality in the general population of the Aquitaine region during the follow-up period was the best possible reference. Unfortunately, similar figures were not available for incidence of cancers or transmission categories of HIV. Thus, we used national data, which were available only for the years 1978–1982. We believe that use of these data was appropriate because figures for this period can be considered unaffected by the AIDS epidemic. Moreover, standardization techniques would correct for important differences between the reference and the follow-up periods as previously described (15).

The use of the Poisson distribution guarantees valid statistical inferences when only few events are observed (14). However, the low number of non–AIDS-defining malignancies precluded subgroup analyses by categories of transmission or types of cancers. Thus, it was not possible to assess whether the increased risk of non–AIDS-defining malignancies was entirely linked to HIV or was reflective, in part, of the known increased risk of cancers in homosexuals (19) or intravenous drug users (20). Nevertheless, because our system is based on passive surveillance, the observed figures probably underestimate the incidence of non–AIDS-defining malignancies, as illustrated by the low number of carcinomas in situ in the cohort. Therefore, further studies will be needed. In these studies, the risk of non–AIDS-defining malignancies in HIV-infected patients should be compared with that in uninfected individuals showing the same at-risk behaviors.

In conclusion, our study reinforces the hypothesis that HIV has a role in carcinogenesis. In addition, it provides a direct estimate of the incidence of and the mortality from non–AIDS-defining malignancies in HIV-infected individuals, regardless of the category of transmission. Active surveillance of cancers among several cohorts’ of HIV-infected patients and uninfected populations at risk of HIV will be necessary to more accurately determine the risk of non–AIDS-defining malignancies linked to HIV.