# K10 control – nothlings2009 (Nöthlings et al., Cancer Epidemiol Biomarkers Prev 2009;18:2098-106, MEC nested case-control)

Source: historisk/nothlings2009.txt (PMC XML-derived, complete). Pages per table are not available, so the source_page range 2098-2106 is acceptable.
Rows checked: 41.

## Discrepancies
| estimate_id | field | in database | in paper | severity |
|---|---|---|---|---|
| _012, _016 | category_label | "110.+ (sic; = 11.0+)" | printed "110.+" (a typo in the paper). Coding dose_low = 11.0 is correct. | – |
| studies.csv | cup_role | cup2010_superseded | not in the cup_role vocabulary (dose_response, hvl_only, excluded, post_2016; pre2010 is used elsewhere) | minor |
| cohorts.csv mec | notes | lists Ollberding 2012 only | nothlings2009 also has cohort_id mec. Pooling exclusion works through cohort_id, but the cohort note is incomplete. | minor |

No numeric discrepancies. Table 2 (red and processed meat, both models: cases/controls, ORs, CIs, P trend 0.34/0.67 and 0.05/0.46), Table 3 (genotype strata, age-sex-ethnicity adjusted, cross-stratum reference) and the text estimate _041, 1.35 (0.96–1.91), all match.

## Re-derivations
- Table 2 red meat cases 238+249+282+240 = 1,009 and controls 380+381+381+380 = 1,522, equal to the totals. Processed: 1,009 / 1,522.
- Table 3 NAT2 cases: 244+260+246+83+72+87 = 992, equal to the NAT2 analysis n. NAT1: 120+122+120+162+152+168 = 844, equal to the NAT1 n.
- The genotype-rapid rows (020–022, 026–028, 032–034, 038–040) are relative to the slow/no-*10 lowest tertile (cross-stratum). They correctly have no category_rank and are non-main.
- Multivariate covariates ("sex, age at blood draw, ethnicity, family history, BMI, fibre, calcium, vitamin D, folic acid, ethanol, physical activity, smoking ... or meat intake, as appropriate") match m2. "Meat intake, as appropriate" leaves mutual adjustment for processed unclear, so `unclear` on 005–008 is correct.

## Missing estimates (optional, out of R1 scope)
Not extracted:
- Table 2 HCA rows: total HCA, DiMeIQx, MeIQx, PhIP tertiles (398 cases / 1,444 controls), both models.
- Doneness preference: well-done vs medium/rare, 1.03 (0.87–1.22) and 1.07 (0.90–1.28).
- The "meat and fat" dietary pattern quartiles.

The vocabulary has hca_* codes, but these are never pooled, so this is optional. Doneness and the dietary pattern are not meat-amount exposures.

## Cross-study point: Nöthlings 2009 vs Ollberding 2012 (same cohort)
Both studies have `cohort_id = mec`. Nöthlings is a nested case-control drawn from MEC, with blood mostly collected after diagnosis, cases from 1995 onward, and frequency-matched controls. Ollberding 2012 is the full-cohort analysis (165,717 participants, 3,404 CRC cases) using the same baseline QFFQ. All Nöthlings cases are MEC cases, so the two must never be pooled together; the studies.csv note says this explicitly. Nöthlings is correctly non-CUP 2016 and should not replace Ollberding. Nöthlings' red meat is undefined; Ollberding separates total red meat from red meat excluding processed.

## Confirmed
All 41 rows numerically. design nested_cc and effect_measure OR are correct, as are the adjustment sets and studies.csv counts (1,009 / 1,522).

## Verdict
Fit for use as a secondary MEC source only. Use Ollberding 2012 for MEC in pooling.