# K9b – tammi2024 (opus-S2 rows 034–059) vs Online Resources 1–5

Source: /home/claude/drive/supplements/tammi2024_OnlineResources.txt (OR4 p. 4–5, OR5 p. 6). All 26 rows read line by line against OR4/OR5.

## Discrepancies
| estimate_id | field | in database | in paper | severity |
|---|---|---|---|---|
| tammi2024_037, _047 | p_value | empty | OR4 last column P = 0.009 (red, M1) and 0.1 (processed, M2) | minor |
| tammi2024_036, _043 | notes | "P heterogeneity between cohorts (model 2) 0.06 / 0.99" on Model 1 rows | OR4 prints a Pheta on each model row (0.06 on M1, 0.11 on M2); footnote a says Q-test is for model 2. Paper-internal ambiguity, note wording misleading | minor |
| tammi2024_dc03, _dc04 (dose_conversions, opus-G) | notes | "Quintile bounds/medians in Online Resource 4 (not available)" | OR4 now available; medians coded in dc05–dc08 | minor (stale note) |

## Checked and confirmed
- Ranks 1/3/5 = quintile numbers (OR4 prints only Q1, Q3, Q5): correct. Q2/Q4 absent from paper, not missing.
- Units: dose_point in g_week (234/467/860 red; 107/359/916 processed) correct; continuous per_100g_week (red) and per_50g_week (processed) correct per OR4 footnote b.
- Cases per quintile (195/231/238; 161/240/263) correct; only cases printed, no PY.
- All HR/CI and P-trend values (0.033, 0.041, 0.011, 0.026) correct.
- OR5 ATBC (men, 048–052, 054, 056, 058) and remaining-cohorts processed rows (053, 055, 057, 059): all 12 HR/CI correct; ATBC sex=men correct (male smokers).
- Exposure coding: red meat = beef, pork, lamb, processed separate (OR2) → red_unprocessed, includes_processed=no correct; processed includes beef/pork/lamb products → yes.
- Adjustment sets m1/m2/m4 match OR4 footnotes c/d and OR5 footnote b (sex, age, energy; + education, smoking, height, BMI, LTPA, HRT, alcohol, dairy). adj_family_history=unclear is reasonable (not in supplement footnotes).
- dose_conversions dc05–dc10: medians/7 (33.4, 66.7, 122.9; 15.3, 51.3, 130.9) and 100/7, 50/7 arithmetic correct.
- is_main_model: M2 quintile rows and 047 = yes, M1 rows = no; consistent with existing 001–004.

All 26 rows confirmed numerically.

## Missing estimates
None. OR5 red-meat "remaining cohorts" substitutions exist as tammi2024_021–024 (opus-G); red meat M2 per 100 g/week exists as 003; processed M1 per 50 g/week as 004.

## Paper-internal inconsistencies
- OR4 footnote a says heterogeneity was tested on model 2, yet a Pheta is printed on both model rows.