Table 2.
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*† | |
Colonoscopy | ||||
All late-stage colorectal cancers | ||||
Screening colonoscopy | 13 (2.8) | 46 (9.0) | 0.32 (0.17–0.61) | 0.30 (0.15–0.59) |
No screening colonoscopy | 458 (97.2) | 463 (91.0) | — | — |
Right colon late-stage cancers | ||||
Screening colonoscopy | 10 (4.0) | 29 (10.6) | 0.40 (0.19–0.86) | 0.37 (0.16–0.82) |
No screening colonoscopy | 240 (96.0) | 244 (89.4) | — | — |
Left colon/rectum late-stage cancers | ||||
Screening colonoscopy | 3 (1.5) | 14 (6.4) | 0.33 (0.09–1.22) | 0.26 (0.06–1.11) |
No screening colonoscopy | 201 (98.5) | 204 (93.6) | — | — |
Sigmoidoscopy | ||||
All late-stage colorectal cancers | ||||
Screening sigmoidoscopy | 92 (19.5) | 173 (34.0) | 0.46 (0.33–0.63) | 0.51 (0.36–0.71) |
No screening sigmoidoscopy | 379 (81.5) | 336 (66.0) | — | — |
Right colon late-stage cancers | ||||
Screening sigmoidoscopy | 68 (27.2) | 89 (32.6) | 0.72 (0.48–1.08) | 0.80 (0.52–1.25) |
No screening sigmoidoscopy | 182 (72.8) | 184 (67.4) | — | — |
Left colon/rectum late-stage cancers | ||||
Screening sigmoidoscopy | 23 (11.3) | 78 (35.8) | 0.24 (0.13–0.42) | 0.26 (0.14–0.49) |
No screening sigmoidoscopy | 181 (88.7) | 140 (64.2) | — | — |
Note: Screening was defined as exposure to a ‘definite or probable’ screening test. Analyses were performed on matched case-controls sets using conditional logistic regression. Twelve subjects had screening by both colonoscopy and sigmoidoscopy; 16 had ‘definite’ screening by barium enema and 191 by fecal occult blood test (FOBT); 18 subjects had both FOBT and colonoscopy, and 73 had both FOBT and sigmoidoscopy. Seventeen cases and 18 controls had an unknown location of cancer.
Model I was estimated with indicator variables for colonoscopy and sigmoidoscopy and receipt of ‘definite’ screening barium enema and FOBT.
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.