Risk of invasive cervical cancer among women with, or at risk for, HIV infection
Although invasive cervical cancer (ICC) has been included among the AIDS-defining conditions since 1993, it remains controversial whether HIV infection increases the risk of developing such neoplasm. In this study, ICC risk was longitudinally investigated among 1,340 HIV-positive intravenous drug user (IDU), 811 HIV-negative IDU, and 801 HIV-positive heterosexual women. These women, aged 15–49 years, were followed up at the Italian HIV Seroconverter Study, at the San Patrignano Community (Rimini, North Italy), and in South-eastern France (the DMI-2 study). The number of observed cases of ICC was compared with the expected one, based on ICC incidence rates among women of the same age in the general population of Italy or France, and standardized incidence ratios (SIR) were computed; 9,070 person-years of observation were accumulated among HIV-positive women and 2,310 among HIV-negative ones. Ten cases of ICC were diagnosed among HIV-positive women (SIR = 12.8): ICC risk was apparently higher among HIV-positive IDU (SIR = 16.7) than among heterosexual women (SIR = 6.7). No cases of ICC were diagnosed among HIV-negative IDU women admitted to the San Patrignano Community (0.15 cases were expected). Our findings confirm previous suggestions showing an increased risk of ICC among HIV-infected women and have important implications at the individual and public health levels. Int. J. Cancer 82:334–337, 1999. © 1999 Wiley-Liss, Inc.
The AIDS epidemic is increasingly affecting European women, particularly in the South, where yearly AIDS incidence rates nearly doubled between 1991 and 1995 (Franceschi et al.,1998b), and where Italy and France account for nearly 50% of female cases (European Centre for the Epidemiological Monitoring of AIDS, 1998). The growth of AIDS among European women paralleled that observed in the United States, and has been minimally influenced by the inclusion, in 1993, of invasive cervical cancer (ICC) among the AIDS-defining diseases (European Centre for the Epidemiological Monitoring of AIDS, 1998).
Several studies have documented an increased risk for cervical intraepithelial neoplasm (CIN), the precursor lesion of ICC, among HIV-infected women, but the role of HIV in ICC development is still discussed (International Agency for Research on Cancer, 1996). A lack of clear impact of the spread of HIV infection on cervical cancer incidence trends and/or increased cervical cancer risk in women with HIV/AIDS has been reported in the United States (Rabkin et al.,1993; Goedert et al.,1998) and Africa (Bassett et al.,1995). However, women with, or at risk for, HIV infection may die before CIN progresses to ICC, or they may be regularly screened and treated for cervical dysplasia, thus preventing progression to ICC.
Among women with AIDS reported to the Italian AIDS Registry between 1993 and 1995, the frequency of ICC as AIDS-defining disease was nearly 3 times higher among intravenous drug users (IDU) than among those infected through heterosexual contacts (Serraino et al.,1996), suggesting that the practice of at-risk sexual behavior (e.g., prostitution) and/or the lack of appropriate screening may favor the occurrence of ICC among HIV-infected IDU women. We have conducted the present study to assess ICC risk among HIV-positive IDU and heterosexual women, and among HIV-negative IDU women.
MATERIAL AND METHODS
In this investigation, longitudinal data from a South-eastern French HIV clinical database (the DMI-2 study), from the Italian HIV Seroconverter Study (ISS) and from the clinical database of former IDUs admitted to the San Patrignano Community, Northern Italy, were combined in order to assess ICC (International Classification of Disease–9th Revision, code 180) risk among women with, or at risk for, HIV infection. The analysis was restricted to women aged between 15 and 49 years who were IDU or who acquired HIV infection through heterosexual intercourse.
DMI-2 HIV clinical database
Included in this study were 1,218 women followed between 1 January 1988 and 31 January 1998 at the Nice University Hospital. Details on the DMI-2 HIV clinical database have been published (Pradier et al.,1998). At baseline, information was collected on sociodemographic characteristics, HIV exposure category, date of first HIV-positive test, biological markers (e.g., the CD4+ cell count and plasma viral load) and clinical conditions. Follow-up visits were scheduled at least every 6 months, although the timing could vary according to the patient's immunological status. Person-years at risk were computed from the first visit up to the date of death, or of ICC diagnosis or of last follow-up. All ICC diagnoses were histologically confirmed by the Pathology Department of Nice University Hospital. For patients who missed the follow-up visits for more than 1 year, information on AIDS and on vital status was actively elicited from either the French National AIDS Registry, clinical records or the Census Bureau of the town of residence.
The ISS is an ongoing longitudinal investigation conducted by 16 clinical centers (for details of the overall study design, see Rezza et al.,1989; and Serraino et al.,1997). Included in this study were 484 women (272 IDUs and 212 heterosexuals) who had had a documented HIV-seronegative test followed by a positive confirmed one. The maximum accepted lag between the 2 HIV tests was 2 years, and the midpoint between the 2 dates was used to estimate the seroconversion period. Person-years at risk for ICC were computed from the estimated HIV seroconversion period to death, ICC diagnosis, or to a cutoff date (i.e., 30 June 1997). ICC diagnoses were histologically confirmed by the collaborating pathologists. To reduce follow-up losses, a linkage with the Italian National AIDS Registry was carried on for AIDS-free individuals at the visit that preceded the cutoff date. Further information on vital status was also ascertained through the census bureau of the town of residence.
San Patrignano Community clinical database
The third group under study consisted of 1,250 formerly IDU women admitted between 1982 and 1997 to the San Patrignano Community, Rimini, North Italy. These women injected heroin for at least 1 year, and spent at least 6 months in the San Patrignano Community. At admission, 439 (35.1%) were HIV-positive and 811 HIV-negative. Person-years at risk of developing ICC were computed from the date of admission to the San Patrignano Community to the date of ICC diagnosis, or the date of death, or the date of leaving the San Patrignano Community, or to 30 September 1997. All ICC diagnoses were histologically confirmed by the collaborating pathologists of the Medical Centre of the San Patrignano Community.
In both the DMI-2 database and the ISS cohort, women were classified as either IDUs or heterosexuals according to the classification adopted by the European Centre for the Epidemiological Monitoring of AIDS (1998). To avoid the inclusion of prevalent cases, 1 woman diagnosed with ICC within 2 months from admission to the Nice University Hospital and 1 woman diagnosed with ICC within 2 months from admission to the San Patrignano Community were excluded from our study.
A person-year analysis was conducted: in each of the at-risk groups (i.e., HIV-positive and HIV-negative IDU, and HIV-positive heterosexuals), the expected number of cases of ICC was computed based on ICC incidence rates registered, for the period 1988–1992, among women 15–49 years of age in the population-based cancer registries of Italy or France (Parkin et al.,1997). Data were from all the 13 Italian and from all the 8 French population-based cancer registries, since the women enrolled in the 3 studies resided in different Italian or French areas not covered by cancer registration. The average annual, age-standardized, ICC incidence rates were multiplied by the number of person-years of observation to obtain the expected number of ICC. The observed number of cases was then divided by the expected number to obtain standardized incidence ratios (SIR). Confidence intervals (CIs) of these SIRs were determined using the Poisson distribution for the observed cases (Breslow and Day, 1987).
The chi-square for trend was used to compare the frequency distribution according to age and calendar year at enrollment between HIV-positive IDU and heterosexual women, and between HIV-positive and HIV-negative former IDU women admitted to the San Patrignano Community.
Age and calendar year at enrollment of the women under study are listed in Table I. The 629 HIV-positive IDU and the 589 HIV-positive heterosexual women enrolled in the DMI-2 study group were of similar age (the median age was 28.5 years and 29.0 years, respectively). More heterosexuals (48%) than IDUs (22%) were enrolled in the study after 1992 (p < 0.001). In the ISS cohort, HIV-positive heterosexual women tended to be older and to be enrolled later than IDUs (p < 0.001).
Table I. DISTRIBUTION OF THE STUDY SUBJECTS, ACCORDING TO STUDY, HIV STATUS, HIV EXPOSURE CATEGORY, AGE AND CALENDAR YEAR AT ENROLLMENT, FRANCE (1987–1997) AND ITALY (1981–1997)1
| 15–19||9 (1.4)||22 (3.7)||25 (9.2)||12 (5.7)||117 (14.4)||29 (6.6)|
| 20–24||109 (17.3)||112 (19.0)||111 (40.8)||71 (33.5)||318 (39.2)||143 (32.6)|
| 25–29||268 (42.6)||188 (31.9)||82 (30.1)||73 (34.4)||261 (32.2)||183 (41.7)|
| 30–49||243 (38.6)||267 (45.3)2||54 (19.9)||56 (26.4)3||115 (14.2)||84 (19.1)4|
|Year at enrollment|
| ≤1989||223 (35.5)||74 (12.6)||166 (61.0)||105 (49.5)||206 (25.4)||191 (43.5)|
| 1990–1992||268 (42.6)||234 (39.7)||69 (25.4)||51 (24.1)||309 (38.1)||151 (34.4)|
| 1993–1997||138 (21.9)||281 (47.7)3||37 (13.6)||56 (26.4)3||296 (36.5)||97 (22.1)4|
|Person-years at risk||2,382.3||1,956.1||2,005.5||1,324.3||2,310.1||1,401.7|
Baseline CD4 cell count was available for 59% of women in the DMI-2 study group and for 29% of women in the ISS cohort. The median number of CD4 cells ranged from 284 for IDUs to 350 for heterosexuals in the DMI-2 study, and from 566 among heterosexuals to 697 for IDUs in the latter (data not shown).
At admission in the San Patrignano Community, 439 (35%) of the 1,250 former IDU women were HIV-positive and 811 were HIV-negative. HIV-positive former IDUs were older and were more frequently admitted in San Patrignano before 1993 than HIV-negative former IDU women (Table I).
Ten cases of ICC were diagnosed during the study period among HIV-positive women (8 cases among IDUs and 2 cases among heterosexuals), whereas 0.78 were expected (Table II). In comparison with women in the general population of Italy or France, the risk of ICC among HIV-positive women was 13-fold higher (SIR = 12.8, 95% CI: 6.6–22.4). The increase in ICC risk was particularly evident among HIV-infected IDU women (SIR = 16.7) (Table II). When stratifying by age group, younger IDU women (i.e., those under 30 years of age) had a higher SIR (56.6) than older IDU ones (SIR = 18.1), although the CIs of the 2 age groups widely overlapped (data not shown).
Table II. SIRs FOR ICC ACCORDING TO HIV STATUS AND HIV EXPOSURE CATEGORY, FRANCE (1987–1997) AND ITALY (1981–1997)
|Number of ICC observed cases||8||2||10||0|
|Number of ICC expected cases1||0.48||0.30||0.78||0.15|
Among HIV-positive heterosexual women, 2 cases of ICC were diagnosed (1 case in the DMI-2 study and 1 case in the ISS cohort), whereas 0.30 were expected. Such excess risk was not statistically significant (SIR = 6.7, 95% CI: 0.0–15.9). None of the HIV-negative former IDU women in the San Patrignano Community was diagnosed with ICC vs. 0.15 expected cases (Table II).
Risk factors for ICC are well known, particularly some sexual behaviors (e.g., an elevated number of sexual partners and an early age at first intercourse) that enhance the risk of infection with the human papillomavirus (HPV), by far the most important determinant of ICC development (International Agency for Research on Cancer, 1996). Since HPV and HIV share the sexual route of transmission, it is difficult to disentangle their respective contribution to the etiology of cervical cancer. Hence, the relationship between HIV infection and ICC could be explained either by the frequent occurrence of HPV infection in HIV-positive women and/or by a direct suppressive effect of HIV infection on the immune system which may promote the persistence of HPV infection (International Agency for Research on Cancer, 1996).
Evidence of an increased incidence of ICC as a consequence of the HIV epidemic is not consistent. A linkage study conducted, in Italy, between the National AIDS Registry and the population-based cancer registries showed a nearly 20-fold excess risk for ICC for women with AIDS (Franceschi et al.,1998a), but such an increased risk was not clearly found by a similar, much larger, study conducted in the United States (Goedert et al.,1998). However, in the United States, a 5.5-fold increased risk of ICC-related deaths was estimated between 1990 and 1995 for HIV-infected women under 45 years of age (Selik and Rabkin, 1998).
We have investigated whether HIV-positive or HIV-negative IDU women and HIV-positive heterosexual women were at increased risk of ICC in comparison with women of the general population. Our main finding was an overall increased ICC risk among HIV-infected women. Higher ICC rates among HIV-infected women have important implications both at the individual and at the public health level, since marked decreases in the incidence of ICC have occurred in the general population following the development of an effective approach to cervical cancer control (International Agency for Research on Cancer, 1996).
The investigation of ICC risk according to HIV transmission category was limited by the small number of cases of ICC diagnosed in the heterosexual group. However, the high ICC risk registered among IDU women in our study is in agreement with previous reports suggesting that HIV-positive IDU women are at a high risk of developing HPV-related diseases compared with HIV-negative IDUs (Conti et al.,1993), or with seronegative partners of HIV-positive IDU men (Johnstone et al.,1994), or with HIV-negative heterosexual women (Rezza et al.,1998). In the absence of data showing increased risk of ICC with increased HIV-related immunosuppression (Goedert et al.,1998), these findings could reflect greater exposure to HPV among HIV-infected IDUs than among heterosexual women. On the other hand, HIV-positive IDUs appear to have a reduced compliance to cervical screening programs than other women (Zanetta et al.,1995), and this may further contribute to enhance their ICC risk.
It has been hypothesized that IDU-related immune suppression may play a role in the elevated incidence of ICC among IDU women (Korn and Landers, 1995). This was not confirmed by the results of our study, since none of the HIV-negative IDU women developed ICC during the observational period. Since person-years of observation among HIV-negative IDUs were small, this investigation lacked statistical power to detect a potential increased ICC risk in such at-risk group. It appears, however, that HIV infection, and not the use of intravenous drugs or of related lifestyle factors, may account for the significant association between HIV and cervical cancer incidence in the women studied in our longitudinal investigation.
Lack of statistical power also hampered, in the present investigation, the assessment of ICC risk among HIV-positive heterosexual women. In addition to a reduced exposure to sexually transmitted diseases, we assume that heterosexual women may have been, since now, at a lower risk of ICC because of the later spread of the HIV epidemic among them, compared with IDU women.
We are aware that biases that limit the interpretation of study results in HIV seroprevalent or sero-incident cohorts (Alcabes et al.,1997) may apply as well to the present investigation. Because of selection bias, the IDU women admitted to the San Patrignano Community, or enrolled in the ISS or in the DMI-2 studies, cannot be considered representative of IDU women of Italy or France. It is difficult, however, to establish whether some characteristics of the IDUs included in our study (e.g., sexual lifestyles or adherence to Pap test programs) may explain, at least partially, the reported high risk of ICC. Second, some HIV-infected women may have died of ICC before enrollment, thus making ICC risk potentially underestimated. Although we have used 3 different data sources to increase the power of the study, lack of statistical power limited the assessment of ICC risk in HIV-positive heterosexual women and among HIV-negative IDU women. The results were consistent across the 3 different data sources. Finally, it was not possible to use cancer registry data from the same regions of the members of the cohorts. Women enrolled in the 3 studies were resident, or lived in different areas of Italy or France, and some of these areas were not covered by population-based cancer registration schemes. While it is unlikely that the use of the ICC national average incidence rates may have substantially distorted our results, these rates are representative of ICC incidence in the general population of Italian and French women (Parkin et al.,1997).
In conclusion, our results indicate that HIV-positive women under 50 years of age are at increased risk for ICC than women in the general population. This higher risk, which appears to be particularly high among younger IDU women, has important public health implications. It emphasizes the well-known needs both to reduce at-risk sexual behaviors and to strengthen effective preventive programs among HIV-positive women.
We thank Ms. N. Oran and Ms. C. Gisbert for their invaluable support in the management of the DMI-2 database at the CISIH–Nice University Hospital, France; the staff of Centro Operativo AIDS, Istituto Superiore di Sanità, Rome; the staff of Centro Medico, Comunità di San Patrignano, Rimini, Italy; and Ms. I. Calderan, Epidemiology Unit, Centro di Riferimento Oncologico, Aviano, Italy, for editorial assistance.