# K10 control – flood2003 (Flood et al., Am J Epidemiol 2003;158:59-68, BCDDP follow-up cohort)

Sources: historisk/flood2003.txt and a pdftotext extraction of flood2003.pdf (native text layer). Pages come from the running heads: Table 4 p.64, Table 6 p.65, Table 7 p.66, Table 2 cut-points p.62.
Rows checked: 49.

## Discrepancies
| estimate_id | field | in database | in paper | severity |
|---|---|---|---|---|
| _044–_049 | category_rank | empty (while _041–_043 have 1–3) | Table 7 is a 3×3 grid with a single reference (meat Q1 × fat Q1). Ranks are inconsistent across the fat strata. These are not dose-response sets; leave them out of GL. | minor |
| adjustments flood2003_m4 | adj_energy | unclear | not stated in the Table 7 fns. Correct as coded. | – |

No numeric discrepancies. All RRs, CIs, medians, trend Ps (0.28, 0.39, 0.73, 0.47, 0.73, 0.22, 0.35; Table 6 total meat 0.28 with Q5 CI 0.79–1.38, which differs from Table 4's 0.80 and is correctly kept), Table 2 cut-points for total meat (<26.5, 26.5–39.3, 39.4–53.7, 53.8–77.7, ≥77.8), and Table 7 counts and cases match.

## Re-derivations
- Table 7 cases: 31+56+8+61+164+72+20+55+20 = 487, equal to the total cases. Table 7 n: 2,903+…+2,262 = 45,496, equal to the cohort.
- Table 4 prints no cases per quintile, so empty n_cases is correct.
- The white-meat trend P (0.47) is printed on the CI line in the PDF; its placement on rank 5 is correct.
- Abstract red meat 1.10 (0.83, 1.45) = row _015.

## Exposure coding
- Red meat = bacon, beef, hamburger, ham/lunch meat, hot dogs, liver, pork, sausage and meat in mixed dishes (Methods p.61). red_total with includes_processed = yes is correct.
- Processed = bacon, ham/lunch meat, hot dogs, sausage. Correct.
- White meat = fried chicken, fried fish, other chicken, other fish. white_meat is correct.
- Total meat includes processed: yes, correct.
- Units g/1,000 kcal (nutrient density). Correct.

## Missing estimates
None. The "well-done meat" proxy named in the abstract is reported only as "data not shown". The meta-analytic numbers in the Discussion (Sandhu, Norat) are not estimates from this study.

## Confirmed
All 49 rows; adjustment sets m1–m4; studies.csv: 45,496 women, 62-item Block FFQ 1987–89, 386,716 person-years, 487 CRC, no subsite analyses.

## Verdict
Fit for use. Note that the doses are energy densities (g/1,000 kcal); conversion to g/day needs mean energy (1,279 kcal is printed in the Discussion).