TABLE 3.
Post‐diagnosis empirical dietary inflammatory pattern scores and 5‐year mortality in strata of cases based on the lymphocytic reaction in colorectal cancer in the Nurses’ Health Study (NHS) and the Health Professionals Follow‐up Study (HPFS) ¶
| 5‐year colorectal cancer‐specific mortality | 5‐year overall mortality | |||||||
|---|---|---|---|---|---|---|---|---|
| Strata of cases | Post‐diagnosis EDIP scores | No. of cases | No. of events |
Age‐adjusted HR * (95% CI) |
Multivariable |
No. of events |
Age‐adjusted HR * (95% CI) |
Multivariable |
| Tumor‐infiltrating lymphocytes | ||||||||
| Absent/low | Tertile 1 | 313 | 28 | 1 (reference) | 1 (reference) | 48 | 1 (reference) | 1 (reference) |
| Tertile 2 | 258 | 39 | 1.23 (0.74–2.02) | 1.32 (0.81–2.15) | 50 | 1.02 (0.67–1.55) | 1.09 (0.72–1.64) | |
| Tertile 3 | 289 | 50 | 1.56 (0.98–2.49) | 1.59 (1.01–2.53) | 76 | 1.55 (1.06–2.25) | 1.53 (1.05–2.23) | |
| Intermediate/high | Tertile 1 | 115 | 12 | 1 (reference) | 1 (reference) | 15 | 1 (reference) | 1 (reference) |
| Tertile 2 | 96 | 8 | 1.04 (0.40–2.70) | 0.83 (0.31–2.26) | 12 | 1.04 (0.46–2.34) | 0.88 (0.39–2.01) | |
| Tertile 3 | 118 | 6 | 0.62 (0.21–1.78) | 0.48 (0.16–1.48) | 15 | 1.07 (0.49–2.34) | 0.91 (0.41–2.02) | |
| P interaction ‡ | .012 | .002 | .15 | .10 | ||||
| Intratumoral periglandular reaction | ||||||||
| Absent/low | Tertile 1 | 39 | 3 | 1 (reference) | 1 (reference) | 8 | 1 (reference) | 1 (reference) |
| Tertile 2 | 40 | 10 | 4.38 (1.14–16.8) | 3.93 (1.07–14.5) | 10 | 1.78 (0.66–4.78) | 1.70 (0.66–4.39) | |
| Tertile 3 | 58 | 13 | 3.47 (0.93–12.9) | 2.96 (0.81–10.9) | 18 | 1.86 (0.75–4.62) | 1.54 (0.61–3.89) | |
| Intermediate/high | Tertile 1 | 389 | 37 | 1 (reference) | 1 (reference) | 55 | 1 (reference) | 1 (reference) |
| Tertile 2 | 314 | 37 | 0.97 (0.60–1.57) | 0.99 (0.62–1.59) | 52 | 0.94 (0.63–1.41) | 0.95 (0.64–1.42) | |
| Tertile 3 | 349 | 43 | 1.16 (0.74–1.82) | 1.16 (0.74–1.83) | 74 | 1.37 (0.95–1.98) | 1.34 (0.93–1.94) | |
| P interaction ‡ | .17 | .11 | .54 | .57 | ||||
| Peritumoral reaction | ||||||||
| Absent/low | Tertile 1 | 42 | 3 | 1 (reference) | 1 (reference) | 9 | 1 (reference) | 1 (reference) |
| Tertile 2 | 39 | 11 | 3.49 (0.93–13.0) | 3.34 (0.93–12.0) | 11 | 1.56 (0.62–3.90) | 1.53 (0.63–3.70) | |
| Tertile 3 | 57 | 13 | 2.44 (0.65–9.15) | 2.67 (0.75–9.49) | 17 | 1.35 (0.56–3.26) | 1.34 (0.57–3.16) | |
| Intermediate/high | Tertile 1 | 385 | 37 | 1 (reference) | 1 (reference) | 54 | 1 (reference) | 1 (reference) |
| Tertile 2 | 312 | 36 | 0.98 (0.61–1.59) | 1.00 (0.62–1.61) | 51 | 0.96 (0.64–1.44) | 0.97 (0.65–1.45) | |
| Tertile 3 | 349 | 42 | 1.17 (0.74–1.84) | 1.13 (0.71–1.80) | 74 | 1.41 (0.98–2.05) | 1.35 (0.93–1.96) | |
| P interaction ‡ | .39 | .13 | .94 | .76 | ||||
| Crohn's‐like reaction | ||||||||
| Absent/low | Tertile 1 | 272 | 26 | 1 (reference) | 1 (reference) | 44 | 1 (reference) | 1 (reference) |
| Tertile 2 | 203 | 36 | 1.42 (0.84–2.38) | 1.44 (0.87‐2.38) | 44 | 1.14 (0.74–1.77) | 1.20 (0.78–1.84) | |
| Tertile 3 | 241 | 39 | 1.38 (0.82–2.31) | 1.25 (0.74‐2.09) | 59 | 1.39 (0.92–2.09) | 1.29 (0.85–1.96) | |
| Intermediate/high | Tertile 1 | 92 | 9 | 1 (reference) | 1 (reference) | 12 | 1 (reference) | 1 (reference) |
| Tertile 2 | 80 | 6 | 0.81 (0.28–2.33) | 0.73 (0.25‐2.09) | 11 | 0.98 (0.42–2.28) | 0.88 (0.38–2.04) | |
| Tertile 3 | 93 | 5 | 0.63 (0.21–1.91) | 0.60 (0.21–1.72) | 13 | 1.02 (0.45–2.31) | 0.96 (0.43–2.14) | |
| P interaction ‡ | .37 | .36 | .59 | .61 | ||||
Abbreviations: CI, confidence interval; EDIP, empirical dietary inflammatory pattern; HR, hazard ratio.
Post‐diagnosis EDIP scores were estimated based on the first questionnaire returned between 6 and 48 months after diagnosis.
The inverse probability weighting method (for tumor tissue data availability) was integrated into the Cox proportional hazards regression models. All Cox regression models were stratified by age and disease stage and adjusted for age at diagnosis.
Multivariable Cox regression models originally included the following variables: year of diagnosis, tumor differentiation, tumor location, microsatellite instability, CpG islands methylator phenotype, BRAF mutation, KRAS mutation, PTGS2 (cyclooxygenase 2) expression, PIK3CA mutation, LINE‐1 hypomethylation, family history of colorectal cancer, pre‐diagnosis empirical dietary pattern scores, post‐diagnosis aspirin use, post‐diagnosis pack‐years of smoking, post‐diagnosis alcohol use, post‐diagnosis physical activity, post‐diagnosis body mass, and post‐diagnosis total energy intake. A backward stepwise selection was used to select the variables for the final models.
Pinteraction (two‐sided) was calculated by the Wald test for the cross‐product of post‐diagnosis EDIP scores (continuous with ceilings at 5th and 95th percentiles) and each lymphocytic reaction component (binary) in the Cox regression model.