Risk Factors for in Situ Cervical Cancer: Results ... - Cancer Research

4 downloads 292 Views 1MB Size Report
Health, University of Alabama, Birmingham, Alabama 35294 [H. F. L.J; Nassau County Department of Health, Minneola, New York I¡SOI[R. S. L.]; and Illinois.
[CANCER RESEARCH 50, 3657-3662, June 15, 1990]

Risk Factors for in Situ Cervical Cancer: Results from a Case-Control Study Carol J. Jones, Louise A. Brinimi,' Richard F. Hamman, Paul D. Stolley, Herman F. Lehman, Robert S. Levine,2 and Katherine Mailin Environmental Epidemiology Branch, Epidemiology and Bioslalistics Program, Division of Cancer Etiology, National Cancer Institute, Bethesda, Maryland 20892 [C. J. J., L. A. B.J; Section of Epidemiology and Public Health, Department of Preventive Medicine and Biometrics, University of Colorado School of Medicine, Denver, Colorado 80262 [R. F. H.J; Clinical Epidemiology Unit, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania 19104 [P. D. S.]; School of Public Health, University of Alabama, Birmingham, Alabama 35294 [H. F. L.J; Nassau County Department of Health, Minneola, New York I ¡SOI[R. S. L.]; and Illinois Cancer Council, Chicago, Illinois 60612 [K. M.J

ABSTRACT A case-control study of 293 patients with in situ cervical cancer and 801 community controls was conducted between 1982 and 1984 in Uve geographic areas in the United States. Relative risk (RR) was elevated among women reporting multiple sexual partners (RR for >5 partners = 5.0), a history of an abnormal Papanicolaou smear (RR = 5.0), interval since last Papanicolaou smear (RR for >10-year interval versus 0- to 2year interval = 4.1), use of oral contraceptives (RR for > 10 years use = 1.4), a history of nonspecific genital infection (RR = 2.6), and smoking (RR for current smokers = 1.9). Risk was low among diaphragm users (RR for >2 years use = 0.5). Neither age at first coitus nor number of births was predictive of risk of in situ disease. Comparisons between this analysis and risk factors previously identified for invasive cervical cancer in this same study indicate that the risk factors were quite similar.

INTRODUCTION Previous investigations have shown a relationship between in situ cervical cancer risk and measures of sexual activity (1, 2), Papanicolaou smear screening history (3, 4), smoking (1,5, 6), and oral contraceptive use (7, 8). Many of these variables, however, are highly correlated with each other, and investiga tors must carefully consider their separate effects. In the present case-control study, extensive information was collected con cerning sexual, menstrual, reproductive, and medical history, allowing us to investigate these relationships in detail. Similar analyses of the invasive cervical cancer case-control study per formed previously also allowed comparison of relationships between invasive and in situ disease (9-11). METHODS The methodology for the overall study has been described elsewhere (9, 10). The study included subjects from five areas reporting to the Comprehensive Cancer Patient Data System: Birmingham, Chicago, Denver, Miami, and Philadelphia. In each of these areas, incident cases of in situ and invasive cervical cancer occurring among women 20-74 years of age were ascertained between April 1982 and January 1984 from 24 participating hospitals. The present analysis focuses on the patients diagnosed with in situ disease during the study period. Community controls, ascertained through random digit dialing tech niques, were matched to each case of invasive cervical cancer (12, 13) on race and 5-year age group. After selection of a residential cluster matched on exchange for each case, telephone numbers were called and an enumeration of the female members of each household, ages 20-69 years, was taken. Of 23,494 telephone numbers sampled, an enumera tion of female members was obtained for 84.1%, and 13,561 (57.9%) had eligible household members for selection. A brief telephone inter view revealed that approximately 25% of the selected controls had had Received 5/22/89; revised 2/26/90. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1To whom requests for reprints should be addressed, at Environmental Epi demiology Branch, National Cancer Institute, Executive Plaza North, Rm. 443, Bethesda, MD 20892. 2 Present address: Community and Preventive Medicine Unit, Department of Medicine, Our Lady of Mercy Medical Center, Bronx, NY 10466.

a hysterectomy and they were replaced with other eligible controls. Home interviews by standardly trained interviewers were completed for 293 (75.7%) of 387 eligible in situ cases and for 801 (71.9%) of 1114 controls. Refusal (10.9% of cases, 9.7% of controls) was the major reason for nonresponse. Other reasons included location difficulties (4.1 versus 3.4%), death (0 versus 0.5%), illness (0.5 versus 1.1%), and other problems (0.8 versus 1.1%). Cases were interviewed a mean of 5.6 months following diagnosis. Interviews lasted approximately l h and elicited information on demographic factors, reproductive history, hy giene practices, sexual and contraceptive behavior, medical events, smoking, diet, marital history, and family history of cancer. The RR,3 as estimated by the odds ratio, was the measure of associ ation used for evaluating effects of exposures. All variables that showed evidence of univariate effects were entered into stepwise logistic models to determine the major predictors of risk. Unmatched logistic regression analyses were performed to obtain maximum likelihood estimates of the relative risks and 95% confidence intervals, while adjusting for confounding variables. Tests for trend in these analyses were obtained by categorizing the exposure variable, assigning the score i to the ith exposure level of the categorical variables and treating the scored variable as a continuous variable.

RESULTS The age and racial distributions of the cases and controls are presented in Table 1. The median age of the cases was 38.7 years, while that of controls was 44.0 years. Because of this disparity, all subsequent analyses were age adjusted, with age entered into the models as a continuous variable. Blacks were overrepresented among the controls and this was similarly taken into account in analyses. Education and income relation ships with risk are shown in Table 2. Although education initially appeared related to risk, after adjustment for other factors, the effect failed to persist. Income was not consistently related to risk. A major risk factor was the absence of regular prior Papani colaou smears, with those reporting not having had a Papani colaou smear in the 10 years prior to diagnosis being at a 4fold excess risk compared with those having been screened within 2 years of diagnosis. The trend of increasing risk with interval since last Papanicolaou smear was statistically signifi cant (P < 0.001). A limited number of women reported never having had a Papanicolaou smear, which was associated with a slight, nonsignificant elevation in risk. Among subjects reporting at least one prior Papanicolaou smear, a relatively large proportion reported having had an abnormal result at least 1 year prior to diagnosis (38.1% of cases versus 10.1% of controls, RR = 5.0, 95% CI = 3.4-7.5). Women having 3 or more abnormal test results experienced an 18-fold excess risk. Women whose first abnormal Papanicolaou smear was within 5 years prior to diagnosis had an 8-fold excess risk, while those whose first abnormal Papanicolaou smear was 5 years or more before diagnosis retained a 3-fold elevation in risk.

3657

3 The abbreviations used are: RR, relative risk; CI, confidence interval.

IN SITU CERVICAL

Table 1 Distribution of cases ana controls by age and race

CANCER

Table 2 Relative risk of in situ cervical cancer associated with selected variables

CasesAge P = test28979338325Trend

A comparison of the crude and adjusted risk associated with the lifetime number of sexual partners indicated that there was very little confounding of the association by any of the other exposures studied. A highly significant trend of increased risk with number of sexual partners was observed, those with more than 5 partners having a 5-fold excess risk compared with women reporting 1 partner. Age at first sexual intercourse was not a significant risk factor after its apparent crude association was adjusted for number of partners and other established risk factors. Joint effects of age at first sexual intercourse and number of sexual partners are shown in Table 3. Cross-tabula tion of the factors indicated a persistent relationship of risk with number of sexual partners in most categories of age at first intercourse. However, no clear relationship was observed for age at first intercourse taking the number of sexual partners into account. Effects of various reproductive measures are shown in Table 4. Although gravid women were at a 40% elevated risk, neither number of pregnancies nor age at first pregnancy was consist ently related to risk. When other reproductive measures were investigated, neither the number of livebirths, induced abor tions, nor vaginal deliveries was an important predictor of risk. In addition, neither stillbirths nor miscarriages were signifi cantly related with risk (data not shown). However, having had a cesarean section was associated with reduced risk, although after adjustment the effect was only marginally significant (RR = 0.6, 95% CI = 0.3-1.0). The risk associated with number of cesarean sections could not be evaluated since very few women had this procedure more than once. The risks of in situ cervical cancer associated with selected hygiene and contraceptive practices are presented in Table 5. Ever having used oral contraceptives was associated with in creased risk, with current users having a 1.8-fold elevation in risk compared with those who had never been users and former users having a 30% increased risk. When duration of use was investigated, a significant trend in risk was observed, with those reporting use for 6-10 years having a RR = 2.0, while use for >10 years resulted in a RR = 1.4 (for trend, P = 0.04). Neither years since first nor years since last use of oral contraceptives was predictive of risk after duration of use was taken into account. No significant trend in risk was observed according to age at which oral contraceptives were first used, although those who reported first use between the ages of 25 and 29 years were at an elevated risk (RR = 1.6, 95% CI = 1.0-3.0). Use of barrier methods of contraception (condom or dia phragm) was associated with reduced risk, with those who used them the longest having the lowest risk. After adjustment for other risk factors, these effects were lessened, but a significant trend in risk persisted, with those using barrier methods for >5 years showing a significant 40% reduction in risk. Barrier methods were further separated into condom and diaphragm use. Although a protective trend was observed for duration of

0.27*58578262429481]P2

3658

//V S/TU CERVICAL CANCER

Table 3 Relative risk afin situ cervical cancer according to age at first intercourse and number of sexual panners All risks are relative to women with 1-2 sexual partners and whose age at first intercourse was >20 years. Adjusted for age. race, interval since last Papanicolaou smear, number of abnormal smears, years of oral contraceptive use, years of diaphragm use, history of nonspecific genital infection or sore, and years of cigarette smoking. Virgins and unknowns are excluded. partners1-21.00(32)" of sexual

Age at first intercourse(yr)2:20

fortrend0.001 (7) (24) 18-19 0.001 2.3(15) 4.5 (22) 7.0 (26) 1.5(18) 16-17 6.0 (24) 2.9(19) 0.26 2.5 (25) 2.8(16) 10.7(22) 4.4(14) 0.02 102.7 0.24P P for trendNo. 0.163-47.1 0.155-95.3(15) " The numbers of women with in situ cervical cancer in each category are shown in parentheses.

Table 5 Relative risks of in situ cervical cancer associated with selected contraceptive variables ControlUse

Cases

CI1.0-3.40.9-2.00.7-1.81.1-3.4

contraceptivesNonuserCurrent of oral userFormer userUnknownYears oralcontraceptivesNonuser10Trend of use of

testYears

0.048744583319879378954168772863017, P =

oralcontraceptivesNonuser