In linkages performed during 2003–2005 in 12 regions of the US with population-based cancer ascertainment, 85
268 women were followed for 665
987 person-years, measured from 5 years before to 10 years after an initial AIDS diagnosis. There were 51
616 women of African ancestry (61%), 17
705 women of Hispanic ancestry (21%), 15
019 non-Hispanic women of European ancestry (18%), and 928 women of other or missing ancestry (1%). At AIDS diagnosis, median age of the women was 36 years (interquartile range 31–43 years); 90% of the women were age 50 years or younger.
The incidence of breast and uterine corpus cancers, but not of ovary cancer, was significantly lower than in the general population. Specifically, linkage to the cancer registries revealed 313 cases of breast cancer (SIR 0.69; 95% CI, 0.62–0.77), 31 cases of uterine corpus cancer (SIR 0.57; 95% CI, 0.39–0.81), and 42 cases of ovary cancer (SIR 1.05; 95% CI, 0.75–1.42) among women with AIDS. Breast cancer SIR was reduced in all groups defined by age at AIDS onset, including those with AIDS onset before age 35 (SIR 0.69). As shown in , the SIR for breast and ovary cancers did not differ by age at cancer diagnosis, a surrogate for menopause status. In contrast, the uterine corpus cancer SIR differed by menopause status: 0.86 (95% CI, 0.54–1.32) before age 50 and 0.33 (95% CI, 0.16–0.61) at or after age 50. Standardized incidence ratio did not vary by racial ancestry for any of the three cancers (). Cancer and AIDS records matched exactly on social security number for 211 (67%) of the breast cancer cases; the records for the other 102 cases matched on other criteria. Breast cancer SIR did not vary by availability of social security number for matching (). By stage, breast cancer SIR from 4 to 60 months after AIDS onset was 0.49 (95% CI, 0.34–0.68) for locally invasive cancer, 0.61 (95% CI, 0.42–0.86) for regional dissemination, and 0.89 (95% CI, 0.40–1.68) for distant metastases (). Data for uterine corpus and ovary cancer were too sparse to evaluate by stage.
Table 1 Standardized incidence ratio (SIR) of cancer among 85268 women from 60 months before to 120 months after an initial AIDS-defining event, 1980–2002a
Standardized incidence ratio (SIR) of breast cancer among women with AIDS
We examined trends in SIR with respect to time relative to AIDS onset and CD4 count, as measures of immune deficiency (). Near the time of the initial AIDS-defining condition (−6 to +3 months), the SIR was 1.15 (95% CI, 0.87–1.50) for breast cancer, 1.19 (95% CI, 0.48–2.45) for uterine corpus cancer, and 2.90 (95% CI, 1.54–4.94) for ovary cancer. Excluding this interval of intensive diagnostic scrutiny, there was no significant trend in the breast cancer SIR from 60 months before to 120 months after AIDS onset (Ptrend=0.14, ). There was no trend with AIDS-relative time in the SIR for uterine corpus cancer or ovary cancer (Ptrend=0.72 and 0.94, respectively, data not presented). As with the AIDS-relative time analyses, CD4 lymphocyte count at AIDS onset was not significantly associated with SIR for breast cancer (Ptrend=0.47, ) or, based on sparse data, for uterine corpus or ovary cancer (Ptrend=0.91 and 0.89, respectively, data not presented).
Calendar time was used as a surrogate for availability of anti-HIV therapy – little or none pre-1990, single and dual reverse transcriptase inhibitors in 1990–1995, and HAART after 1995. As shown in , for breast cancer occurring 4 to 27 months after AIDS onset, no cases occurred (vs
4.7 expected) in the cohort with AIDS onset in 1980–1989, whereas the SIR was 0.38 with AIDS onset in 1990–1995, and the SIR was 0.83 with AIDS onset in 1996–2002 (Ptrend
=0.002). Excluding the 1980–1989 cohort that had no cases, the SIR still increased significantly (P
=0.01) from the 1990–1995 to the 1996–2002 cohort. The significant increase in breast cancer SIR also was found using single calendar years in a Poisson regression model (Ptrend
=0.003, ). Data were very sparse for years 2000–2002, with only 7.2 breast cancer cases expected. Comparing the 1990–1995 and the 1996–2002 cohorts, risk of local stage breast cancer increased from SIR 0.40 (95% CI, 0.22–0.66) to 0.61 (95% CI, 0.38–0.93). Similarly, risk of regional stage breast cancer risk increased from SIR 0.53 (95% CI, 0.29–0.88) to 0.77 (95% CI, 0.47–1.19). There was no trend with calendar time in the SIR for uterine corpus cancer, ovary cancer, or distant metastatic breast cancer (Ptrend
0.5) but data were sparse.
Figure 1 Observed and fitted (linear Poisson regression model) standardized incidence ratio (SIR) for breast cancer occurring 1–5 years after AIDS diagnosis (n=90 observed), by calendar year. A quadratic Poisson regression model did not fit the (more ...)
Indirect adjustment for age at first live birth, nulliparity and BMI had little effect on the breast cancer SIR estimate (SIRadj 0.72 vs 0.69 unadjusted). For uterine corpus cancer overall, there was little effect with indirect adjustment for parity and smoking (SIRadj 0.59) or for BMI (SIRadj 0.56) compared to the unadjusted estimate (SIR 0.57). Among postmenopausal women, the SIR for uterine corpus cancer was modestly attenuated with adjustment for parity and smoking (SIRadj 0.39) and for BMI (SIRadj 0.37, compared to the unadjusted SIR 0.33).