# K9b – dankner2024 (opus-S2 rows 044–069) vs supplement

Sources: /home/claude/drive/supplements/dankner2024_suppl.txt (S2 p. 5, S4 p. 7); S5a rendered from /home/claude/suppl_in/dankner2024_suppl.pdf pp. 8–9 (pdftoppm) and read visually.

## Discrepancies
| estimate_id | field | in database | in paper | severity |
|---|---|---|---|---|
| dankner2024_044–049 | dose_unit | per_1g (new value) | S2 prints no increment; per_1g is an inference and is not in the KODEBOK vocabulary/tilleggsverdier | minor |
| dankner2024_043 (opus-G, not S2) | ci_low/ci_high | 1.02/1.78 (disputed) | S4: beef, 24-h recall, CRC 1.59 (1.11, 2.28). 1.02–1.78 is the S4 GI-cancer red meat FFQ CI (1.35 (1.02, 1.78)) | critical for 043 if used; already marked disputed. Recommend correct to 1.11–2.28 or retire 043 in favour of 059 (same estimate; never pool both) |
| studies.csv dankner2024 | notes | "Supplementary Tables S1–S6 ... not available" (older text, later appended) | supplement now available | minor (stale) |

## Checked and confirmed
- S4 (10 rows, 050–059): all HR/CI, I² and cases/participants 746/24,232 (FFQ) and 67/2,657 (24-h recall) correct. Cross-check: FFQ = NICCCS 122 + HDS 14 + Ovary 53 + GOH 20 + IIHD 537 = 746; 24-h = Mabat 41 + Negev 26 = 67; total 813/26,889 = Table 4a. Cohort assignment by S3 questionnaire type correct.
- **Row 059: beef (S4 "Unprocessed Beef"), 24-h recall, 1.59 (1.11–2.28), I² 0% — confirmed.**
- S2 (6 rows): NICCCS 1 (0.996,1.004) OR and HR; Mabat 1.001 (0.999,1.004) OR/HR; IIHD unimputed OR 1.000 (0.993,1.008), HR 1.001 (0.993,1.008) — correct. Adjustment m2 (adds physical activity, BMI) matches S2 footnote a.
- S5a (10 rows, 060–069) read from the rendered image: IIHD unimputed 50/1002: total 1.16 [0.57;2.34] w 5.4%; red 0.89 [0.30;2.64] 6.4%; beef 3.35 [0.92;12.21] 6.2%; processed 0.35 [0.11;1.05] 2.9%; poultry 1.32 [0.61;2.83] 5.4%. Pooled 326/18,034: total 1.09 [0.92;1.28] p=0.64; red 1.12 [0.85;1.48] p=0.96; beef 1.61 [1.17;2.22] p=0.85; processed 0.94 [0.77;1.14] τ²<0.0001 p=0.47; poultry 1.09 [0.91;1.30] p=0.76. All correct.
- Claim that the other six cohort rows in S5a equal Table 4a: verified against rows 001–040 (all 30 values identical).
- Exposure: red meat = mammalian muscle meat, processed separate → red_unprocessed correct; beef no; poultry unclear acceptable.

All 26 rows confirmed numerically.

## Missing estimates
None for colorectal outcomes. S4/S5b gastrointestinal-cancer estimates and S6 PAFs are out of scope (R1).

## Paper-internal inconsistencies
- Main-text Results print the 24-h-recall beef estimate as 1.59 (1.02, 1.78); S4 prints 1.59 (1.11, 2.28). The text CI equals the GI red-meat FFQ CI in S4.
- S4 shows a significant beef association in 24-h recall cohorts, contradicting the Discussion's statement that no association was found in the 24-h recall studies.