Six hundred seventy-eight women were consented and completed a baseline questionnaire. Fifty-six women were excluded from the analysis based on either missing a cytology or biopsy diagnosis from the centralized laboratory (see ). There were 43 women who had normal colposcopy and no biopsy or ECC were indicated. Of these, 19 had a negative cytology interpretation at the screening visit and were included for analysis as “benign”. Those with abnormal Pap test results (ASC-US or worse) from the screening visit were excluded since a histologic diagnosis was not available in the face of persistent abnormal cytology. Demographic characteristics and selected behaviors of the 622 women by CIN status are given in . Overall, the group was racially/ethnically diverse; 32.5% were white, 17.5% black, 15% Latino, 6.9% Asian, and the remaining mixed or other. The cohort appeared to have relative high-risk sexual behaviors compared to national data (4
) defined by a high number of previous pregnancies (37.9%, 95% CI = [34.1-41.8%]), C. trachomatis
infections (25.7%, 95% CI = [22.3-29.1%]), total number of lifetime sexual partners (mean = 8 [standard deviation (SD) = 8.6]), and history of anal intercourse (38.9%, 95% CI = [34.1-41.7%]). Comparison between the women who were included vs excluded (n = 56) for analysis showed that the excluded group had a fewer number of C. trachomatis
infections (12.5%; 95% CI = [3.8-21.2%]; p = 0.03) and were less likely to report a history of anal intercourse (25.0%; 95% CI = [21.6-28.4%]; p = 0.04) than those included. No other differences were found for any of the variables listed in .
Of the 622 women, 41 (6.6%; 95% CI = [4.6 - 8.6%] ) had CIN 3, 81 (13%; 95% CI = [10.4-15.6%]) had CIN 2. 157 (25.2%; 95% CI = [21.8-28.6%]) had CIN 1 and 343 (55.1%; 95% CI = [51.2 - 59.0%]) were considered benign. Of the 622 women, 36 had missing referral diagnosis from KPNC. compares the cytologic referral diagnosis from KPNC to the centralized histologic diagnosis. The majority of CIN 2 and 3 cases (95%; 95% CI = [91.1-98.8%]) were diagnosed from ASC-US or LSIL referral diagnosis. HSIL diagnoses were rare (1.9%; 95% CI = [0.1-2.9%]) and had a similar rate as reported by the cytology laboratories atKPNC for the year 1999-2000, near the time the study was initiated. Only 4 of the 11 HSIL diagnoses were confirmed by biopsy. The number of biopsies taken influenced final diagnosis. From the 622 women, there were 1,651 biopsies. Those with benign diagnosis including only those with biopsies had the least average number of biopsies (mean 2.3 [S.D. = 1.2] biopsies per women). This was significantly lower than those with CIN 1, CIN 2, and CIN 3 (mean number of biopsies per women was 3.1 [S.D. = 1.2], 2.9 [S.D. = 1.1], and 3.5 [S.D. = 1.1], respectively) with all the p-values < 0.001. In addition, the average number of biopsies of those with CIN 3 was also significantly different from those with CIN 2 (p < 0.01), but not with CIN 1.
Twelve HPV tests were considered inadequate or missing. shows the number of HPV positive tests within each histologic category. HPV 16/18 was strongly associated with grade of CIN, with the observed percentage of women with HPV 16/18 gradually increasing with increasing severity of lesion. Women with benign examinations had the lowest rate of HPV 16/18 detection (20.5%, 95% CI = [16.2-24.8%]) and those with CIN 3 had the highest (65.9%, 95% CI = [51.2-80.1%]). Rates of HPV 16/18 were significantly higher in women with CIN 3 than in those with either CIN 1 (24.4%; 95% CI = [17.6-31.2%])or benign diagnoses (p<0.001 for both), and marginally higher than in those with CIN 2 (48.2%; 95% CI = [37.3-59.2%], p=0.06). Rates of HPV 16/18 in women with CIN 2 were also higher than in those with benign or CIN 1 diagnoses (p<0.001 for both). No differences for HPV 16/18 detection were seen between women with benign and CIN 1 outcomes. Only 2 women with CIN 3 were HPV negative and none had low risk HPV only.
Prevalence of any high-risk HPV was also higher in women with CIN 2 (87.7%, 95% CI = [80.5-94.9%]) or CIN 3 (95.1%, 95% CI = [88.4-100%]) than in those with CIN 1 (65.8%, 95% CI=[58.3-73.3%]) or benign diagnoses (54.5%, 95% CI = [49.1-59.9%], p <0.003 and p < 0.01, respectively).
Separate multinomial logistic regression models were fitted for each candidate predictor variable, with results summarized using p-values from tests of association for each of the following outcome comparisons: CIN 3 vs. CIN 1 or less, CIN 3 vs. CIN 2, and CIN 2 vs. CIN 1 or less. These results are summarized in . The first comparison (CIN 3 vs. CIN 1 and less) showed that the following variables associated with risk for CIN 3 at p < 0.1 level: total months of OCP use, total months on medroxyprogesterone, number of sex partners in last 2 months, and number of cigarettes smoked in the last 24 hours. The second comparison (CIN 3 vs CIN 2) found only number of sex partners in last 2 months significant at p < 0.1 level. The third comparison (CIN 2 vs. CIN 1 or less) showed that the risk of CIN 2 is associated with younger age, history of Chlamydia, and total months on medroxyprogesterone at the p < 0.1 level.
Predictor variables with associations significant at p< 0.1 level in were included together in a final multinomial logistic model, with results summarized as odds ratios (and 95% confidence intervals) for the outcome comparisons introduced above. The model also adjusted for years of sexual activity and clinic site. The first comparison for CIN 3 vs. CIN 1 or less found total years of OCP use increased risk of CIN 3. For each one additional year on OCPs, the odds of having CIN 3 increases by 36% on average. Although not statistically significant, having 2 or more recent sex partners had an increased odds of CIN 3 (p=0.08)
The second comparison for CIN 2 vs. CIN 1 or less found that for each additional year on medroxyprogesterone increases the odds of CIN 2 by 46%. Both analyses found HPV 16 and/or 18 and other high-risk HPV detection significant: the former increases the odds of CIN 3 by almost 3000%, and CIN 2 by 600%; and the latter increases the odds of CIN 3 by 500%, and CIN 2 by 260%, comparing with those negative for HPV or low-risk HPV. No statistically significant differences were found for the comparison between CIN 3 and 2. The results are presented in .
Multivariable logistic regression model* for risk of CIN 3 and 2