| |
Low incidence of invasive cervical cancer among HIV-infected US women in a prevention program in WIHS Study
|
| |
| |
AIDS: Volume 18(1) 2 January 2004 pp 109-113
Massad, L Stewart; Seaberg, Eric Ca; Watts, D Heatherb; Hessol, Nancy Ac; Melnick, Sandrad; Bitterman, Pincase; Anastos, Kathrynf; Silver, Sylviag; Levine, Alexandra Mh; Minkoff, Howardi
From the Southern Illinois University School of Medicine, Springfield, Ilinois, the aJohns Hopkins Bloomberg School of Public Health, Baltimore, the bNational Institute of Child Health and Human Development, Bethesda, Maryland, the cUniversity of California, San Francisco, California, the dNational Cancer Institute, Bethesda, Maryland, eRush Medical College, Chicago, Ilinois, fMontefiore Medical Center and Lincoln Medical and Mental Health Center, Bronx, New York, the gGeorge Washington University Medical Center, Washington, DC, hKeck School of Medicine, University of Southern California, Los Angeles, California, and the iMaimonides Medical Center, State University of New York, Brooklyn, New York, USA.
BACKGROUND
Invasive cervical cancer (ICC) is a criterion for a diagnosis of AIDS. The incidence of cervical cancer precursors is high among women infected by HIV, and recurrences are common despite therapy. While large studies using HIV and cancer registry and chart review data suggest that women with HIV face an increased risk of ICC, the incidence of ICC is unknown.
In Africa, where HIV infection is widespread, ICC has emerged as a significant complication of HIV disease. However, in several countries cervical cancer incidence has remained stable as HIV infection rates have exploded suggesting either that HIV does not predispose to ICC or that women die of other causes before ICC becomes symptomatic. These studies have been limited by lack of central tumor registries and incomplete HIV testing among ICC patients and controls, and studies have shown an increased risk of ICC among HIV infected African women. Cervical cancer is a leading AIDS-defining condition in US women. Women with HIV express oncogenic human papillomavirus (HPV) at high rates, and persistent HPV infection has been linked to ICC risk. Longer survival with highly active antiretroviral therapy (HAART) may allow ICC to emerge in women who might otherwise have died as a result of opportunistic infections. In an effort to determine the magnitude of ICC risk for women receiving regular medical care and standard cancer prevention services, we describe cervical cancer incidence in a national multicenter cohort study of women with and at risk for HIV.
ABSTRACT/SUMMARY
Objective: To measure the incidence of invasive cervical cancer (ICC) in US women infected with HIV.
Design: Multicenter prospective cohort study, conducted between October 1994, and September 2001.
Setting: HIV research centers operating as six urban consortia in the Women's Interagency HIV Study.
Subjects: A total of 2131 women (462 HIV seronegative, 1661 HIV seropositive, and eight seroconverters). Women with a history of hysterectomy or of cervical cancer at baseline evaluation were excluded.
Intervention: Cervical cytology obtained at 6-month intervals, with a colposcopy referral threshold of atypia, followed by individualized treatment.
Main outcome measure: ICC diagnoses obtained from study databases and regional cancer registries and confirmed by a gynecologic pathologist.
Results: No incident ICC were observed in HIV seronegative women during 2375 woman-years of observation. During 8260 woman-years of observation, eight putative incident cases of cervical cancer were identified in HIV seropositive women, but only one was confirmed, yielding an incidence rate of 1.2/10 000 woman-years (95% confidence interval, 0.3-6.7/10 000 woman-years). The difference in incidence between HIV seropositive and seronegative women was not significant (P = 1.0).
Conclusion: ICC is uncommon in HIV-infected US women participating in a regular prevention program.
RESULTS
At enrollment, the median age was 36 years for the 1661 HIV seropositive women and 34 years for the 470 seronegatives eligible for analysis. At entry, HPV was detected by cervicovaginal lavage in 63% of HIV seropositive and 32% of seronegative women, with high-risk oncogenic types in 14% and 4%, respectively. Cytology results were abnormal at baseline in 38% of HIV seropositive women and 17% of seronegative women, but high grade abnormalities or cancer were found in only 2% of HIV seropositives and 1% of seronegatives.
No cases of ICC were observed in HIV seronegative women during 2375 years of observation, yielding an incidence rate of 0.0/10 000 woman-years [95% confidence interval (CI), 0.0-15.5/10 000 woman-years].
Eight cases of incident ICC were preliminarily identified in HIV seropositive women through WIHS databases. Upon subsequent review by cancer registry staff, three were found to have been erroneously identified as ICC in registry reports; all were from New York. Histologic re-evaluation of the remaining five cases showed that three were pre-invasive squamous lesions (two grade 2 cervical intraepithelial neoplasia, one carcinoma in situ) and one was an in situ adenocarcinoma. The fifth case was a confirmed ICC, observed during 8260 woman-years of observation, yielding an incidence rate of 1.2/10 000 woman-years (95% CI, 0.3-6.7/10 000 woman-years). The difference in incidence between HIV seropositive and seronegative women was not statistically significant (P = 1.0).
Further information was obtained about the incident ICC case. Cytology at enrollment in 1995 and 6 months later was high grade squamous intraepithelial lesion/atypical glandular cells of uncertain significance. The participant twice refused recommended colposcopy, then was incarcerated for 3 years. Cytology obtained in 1998 showed squamous cell carcinoma, which biopsy confirmed. At the WIHS data collection visit 3 cell/l weeks prior to diagnosis, her CD4 lymphocyte count was 207 x 106 cell/l and her HIV RNA level was 20,000/ml. She was on no antiretroviral therapy, though she had been on monotherapy at some prior visits. Loop conization revealed 2.5 mm invasion with lymph-vascular space involvement by tumor (Stage IA1). She achieved remission with radiation therapy and remained free of cancer as of 12 September 2002.
Thus, we observed one incident ICC among 2131 women with a total of 10 635 woman-years of follow-up. Using the sex-, age-, and ethnicity-specific ICC incidence rates from SEER, 1.59 incident ICC were expected.
DISCUSSION by authors
Among US women with HIV enrolled in a regular cervical cancer prevention program, the incidence of ICC is only 1.2/10 000 woman-years, a low risk statistically indistinguishable from that in HIV seronegative women and similar to that reported among age- and race-matched women in the general population. This low rate was observed in spite of previously reported ICC risk factors in our cohort: a high rate of infection with HPV [21], a substantial prevalence and incidence of abnormal cytology, and non-compliance with recommended follow-up for abnormal cervical cytology results.
This low ICC incidence cannot be generalized to women not in a regular program of ICC screening and prevention. It may not be generalizable to women not receiving HAART. As an observational study, WIHS cannot determine how screening and HAART use might have contributed to our low ICC rate, but ICC rates may be higher in unscreened women and those not receiving optimal antiretroviral therapy.
Screening HIV infected women for ICC is cost-effective. Clinicians and participants should be aware of standard recommendations for surveillance and referral, including two cytologic assessments in the first year after HIV diagnosis and annually thereafter, with referral for colposcopy after any smear showing atypical glandular cells of undetermined significance or a more severe lesion.
The single case identified during our analysis period suggests that some incident ICC may occur because individual and social factors prevent eradication of precursors rather than because screening fails. Further decreases in ICC incidence among HIV infected women may require intensive efforts to monitor compliance, reinforce the value of intervention, and address social and emotional barriers to care.
Our findings are consistent with a 2000 study that reported no ICC cases in Scottish registries after 1993. However, in contrast to our findings, other studies suggest higher ICC rates among HIV seropositive women than in the general population. Several reasons may underlie this discrepancy. First, WIHS may have inadequate statistical power to detect a true difference, despite being the largest cohort study of HIV seropositive women in the USA. Second, most studies that showed a difference in ICC incidence relied on cancer registry data or chart review without pathology confirmation. Our experience suggests that these may contain substantial numbers of women who have cervical dysplasia miscoded as ICC. Third, other studies may not have been as successful in interrupting oncogenesis through treatment of precursors. Fourth, by improving cell mediated immunity, HAART use in our cohort may have altered the natural history of precursor lesions; studies on the effect of HAART on cervical cancer risk are conflicting. Finally, our participants may not have been observed for a sufficient period of time to allow differences to emerge. Surveillance is ongoing.
In summary, we have found that HIV-infected women, despite having an amalgam of risks for cervical cancer and despite previous assertions of substantive increases in ICC prevalence, have reassuringly low ICC incidence rates when in programs that provide careful monitoring for cervical disease.
Methods
This study was conducted as part of the Women's Interagency HIV Study (WIHS), a multicenter prospective cohort study of the natural history of HIV infection and related health conditions among HIV seropositive women and at-risk HIV-uninfected women. The protocols, procedures, and baseline information of the WIHS have been described previously; WIHS participants are representative of US women with HIV. Briefly, the WIHS began enrollment in 1994 at six study sites enrolling 2628 women. Written informed consent was obtained after approval by local human subjects committees. HAART was rapidly adopted in the cohort after commercial availability in 1996, with 51% having initiated HAART by 1998. Although follow-up is ongoing, this analysis included all follow-up information obtained before 1 October 2001 to allow time for ICC ascertainment.
All participating women were screened with conventional cervical cytology every 6 months. Smears were interpreted at Dianon Systems (New York, New York, USA; formerly Kyto or Kyto Meridien) using the 1991 criteria of the Bethesda System for cervicovaginal cytologic diagnosis [24]. At baseline, cervicovaginal lavage with saline was used to obtain specimens for PCR testing and typing for HPV [21], but results were not used in management.
A protocol for evaluation of abnormal cytology prescribed colposcopy for all grades of epithelial abnormality, including atypia. Colposcopy compliance was tracked, missed appointments rescheduled, and participants contacted by dedicated WIHS staff. Local and national community advisory boards also actively promoted compliance. Treatment decisions were not specified by study protocols. Histologic findings were entered into a central database.
To identify ICC cases, the WIHS database was reviewed. This included information from local cancer registries, medical record review prompted by participant report, and WIHS gynecologic examinations. ICC identified prior to WIHS enrollment (nine cases) or diagnosed as a result of evaluation prompted by the enrollment Papanicolaou smear (one case) were considered to be prevalent and so were excluded from study. Women also were excluded if they had a history of hysterectomy (218), if hysterectomy information was missing (three), or if no follow-up was available (266), leaving 2131 women for analysis: 462 HIV seronegative, 1661 HIV seropositive, and eight who seroconverted during follow-up. Seroconverting women were classified as seronegative prior to the date of seroconversion and seropositive thereafter.
Original pathology diagnoses were sought for all women identified as having incident ICC. Participants were reclassified as not having ICC when registries reported a coding error or review of original pathology reports revealed either a pre-invasive lesion or an ICC diagnosis based on cytology alone. Slides for the remaining participants with presumed ICC were retrieved and forwarded for review to a gynecologic pathologist (P.B.).
ICC incidence rates were computed as the number of observed incident ICC divided by the total number of person-years of observed follow-up. The follow-up time available for any woman reflected the number of years from her baseline visit until the diagnosis of ICC, death, loss to follow-up, or 30 September 2001, whichever occurred first. ICC incidence rates for HIV-negative and HIV-positive women were compared assuming the Poisson distribution using a two-sided test. All analyses were performed using exact statistical methods available in StatXact version 5 (Cytel Software Corp., Cambridge, Massachusetts, USA).
To place the results into a more general context, we also estimated the expected number of ICC for this cohort using data from the United States Surveillance, Epidemiology, and End Results (SEER) Program [25]. The yearly sex-, age- (in 5-year categories), and race- (White, Latino/Hispanic, and all others) specific incidence rates from SEER for the period 1994-1999 were multiplied by the follow-up time accrued by WIHS participants for the corresponding strata and then summed to estimate the number of ICC expected among the analyzed cohort. A formal statistical comparison of the observed and expected number of ICC was not appropriate, as the case definition used for this study specifically excluded registry documented cancers not confirmed by the study pathologist, while SEER does not provide for histologic confirmation.
|
|
| |
| |
|
|
|
