# K1 control – lee2009 (SWHS, Nutr Cancer 2009; NIHPA manuscript)

Controller: K1, 2026-10-30. Source read: `lee2009.txt` (PMC2810117 manuscript; no journal page numbers).

## Method
- **Table 2 (105 rows).** A script parsed it and compared intake bounds/labels, person-years, cases for 3 outcomes, point, Q5 CI and p_trend: 840 field comparisons, **0 mismatches**.
- **Table 4 (21 rows) and the Discussion cut-point rows (3).** These were checked by hand against the text: all match.
- Case sums were re-derived. Every Table 2 set sums to 394 / 236 / 158, with one exception: marine fish rectum sums to 157. The paper prints 37+30+28+30+32 = 157, and the database records it as printed.
- Person-years per exposure sum to 540,153–540,156 (paper total 540,156), which is rounding. Salted fish (Table 4) sums to 540,148.

## Discrepancies

| estimate_id | field | in database | in paper | severity |
|---|---|---|---|---|
| – | – | – | – | – |

No discrepancies found.

## Key points verified
- **Red meat** → `red_undefined`, `exposure_includes_processed = unclear`. Correct.
  - The paper never defines red meat.
  - The Appendix 1 FFQ items (pork cuts, organ meats, beef/lamb) are all fresh.
  - Smoked meat/bacon and salted meat are asked separately as preserved foods (Methods).
  - "Probably unprocessed, incl. organ meat" is a fair reading, but it is not stated; `unclear` is the honest code.
- **"White meat" coded `poultry`.** Acceptable. The Results text calls the same estimate "poultry intake (Ptrend =0.23)", and the FFQ items are chicken, duck and goose.
- **is_main_model:** `yes` on the Table 2 sets for total meat, red meat (016–030), poultry and total fish. Marine fish, freshwater fish, eel, the Table 4 rows and the Discussion cut-point rows are `no`. Correct; there is one multivariable model.
- **CUP-relevant rows.** CUP's contrast ">=67 vs <=23.9 g/day" matches the red-meat quintiles. CUP used colorectal **016–020**, colon **021–025** and rectum **026–030**. These are also the rule-conform rows, since there is no alternative red-meat definition.
- **Doses:** only upper bounds are printed. `dose_low` = the previous bound, and `dose_point` is empty (no medians). Correct.
- **Q2–Q4 CIs** are not printed; confirmed.
- **adjustments:**
  - m1 matches the Table 2 footnote (age, education, income, season, tea, NSAID, energy, fiber).
  - m3 matches the Table 4 footnote, which has no NSAID and no fiber.
  - Smoking, alcohol and HRT were deliberately not adjusted (Results). The `no` flags are correct.
- **studies.csv:** n 73,224; mean follow-up 7.4 y; 394/236/158 cases; r = 0.52; cumulative-average diet (`diet_repeated = yes`); raw weight basis ("raw red meat" in the Discussion). All correct.

## Missing estimates
- None required.
- **Scope decision only:** shrimp and shellfish (Table 2). If the project's "fish" negative control follows CUP's fish/seafood grouping, these would be candidates. The extractor excluded them as crustaceans/molluscs and documented this.
- The cooking-method rows (deep-fried, stir-fried, roasted) apply to meat or fish, so their exclusion is reasonable.

## Rows confirmed
All 129 rows (lee2009_001–129) are confirmed correct.

## Internal inconsistencies in the paper
- Marine fish rectal cases sum to 157 vs 158.
- The Table 2/3 footnote says "Range of fat intake (g/day)" for the food rows. The label is wrong; the values are food intakes.
- Table 1 vegetable/fruit Q3 colon "1.2 (0.9-1.0)" is an impossible CI (out of scope).
- Table 4 salted meat <1/month rectal "0.7 (0.6-1.3)" is strongly asymmetric, a probable typo. Row 122 records it as printed.
- Shellfish Q3 prints only a CI.
- The Discussion cut-point analysis gives no direction, covariates or case counts. The rows are correctly `hvl`, main = `no`, model m2 = unclear.

## Verdict
Fit for use. No errors found. All red-meat rows CUP relied on are exact.