Table 3.
Receipt of screening colonoscopy or sigmoidoscopy according to colon location | Sample size (n) and % by cases and controls | Odds ratios and 95% confidence intervals | ||
---|---|---|---|---|
| ||||
Cases | Controls* | Model I† | Model II‡ | |
All late-stage colorectal cancers | ||||
Screening colonoscopy | 7 (1.9) | 16(4.8) | 0.36 (0.14–0.91) | 0.36 (0.14–0.95) |
Screening sigmoidoscopy | 62 (17.0) | 96 (28.6) | 0.47 (0.33–0.68) | 0.51 (0.35–0.75) |
No screening | 296 (81.1) | 224 (66.7) | — | — |
Right colon late-stage cancers | ||||
Screening colonoscopy | 6 (3.3) | 16 (4.8) | 0.75 (0.28–2.04) | 0.72 (0.25–2.05) |
Screening sigmoidoscopy | 48 (26.4) | 96 (28.6) | 0.89 (0.58–1.35) | 0.94 (0.60–1.45) |
No screening | 128 (70.3) | 224 (66.7) | — | — |
Left colon/rectum late-stage cancers | ||||
Screening colonoscopy | 1 (0.6) | 16 (4.8) | 0.08 (0.01–0.65) | 0.09 (0.01–0.72) |
Screening sigmoidoscopy | 13 (7.7) | 96 (28.6) | 0.18 (0.10–0.33) | 0.20 (0.11–0.38) |
No screening | 155 (91.7) | 224 (66.7) | — | — |
Note: Screening was defined as exposure to a ‘definitely’ screening test only, and excluded cases and controls who had: screening by both colonoscopy and sigmoidoscopy; or screening by barium enema and fecal occult blood test (FOBT); surveillance colonoscopy or sigmoidoscopy; diagnostic colonoscopy for positive FOBT; or unknown colonoscopy or sigmoidoscopy indications. Fourteen cases with unknown cancer location were not included in the right or left-sided analyses.
The analyses in this table used the same sample of controls; we used unconditional logistic regressions in order to retain all eligible subjects, including those whose matched controls or case had been excluded.
In Model I, odds ratios and confidence intervals were obtained using unconditional logistic regression that adjusted for matching variables (study site, age, sex, and health plan enrollment duration).
Model II was further adjusted for census block-group poverty levels (as a continuous variable), number of preventive health care visits, family history of colorectal cancer, and comorbidity index at baseline. Missing values of poverty level were imputed using predictive mean matching.