Table 2.

Colorectal cancer mortality by postdiagnosis physical activity quartile

Postdiagnosis physical activity quartile (MET-h/wk)No.Colorectal cancer–specific mortalityOverall mortality
No. of eventsUnivariate HR (95% CI)Stage-stratified HR (95% CI)Multivariate stage-stratified HRa (95% CI)No. of eventsUnivariate HR (95% CI)Stage-stratified HR (95% CI)Multivariate stage-stratified HRa (95% CI)
Male
 Q1 (<6.4)77171 (referent)1 (referent)1 (referent)421 (referent)1 (referent)1 (referent)
 Q2 (6.4–18.4)7580.42 (0.18–0.98)0.43 (0.19–1.00)0.38 (0.16–0.90)390.87 (0.56–1.34)0.87 (0.56–1.36)0.80 (0.52–1.25)
 Q3 (18.6–46.5)78130.64 (0.31–1.31)0.61 (0.30–1.27)0.69 (0.33–1.44)310.58 (0.36–0.92)0.59 (0.37–0.94)0.66 (0.41–1.06)
 Q4 (≥47.1)7530.15 (0.04–0.51)0.16 (0.05–0.53)0.17 (0.05–0.57)300.60 (0.38–0.96)0.62 (0.38–0.99)0.63 (0.39–1.02)
Ptrendb0.00470.00510.00990.0350.0440.086
Female
 Q1 (<2.4)75201 (referent)1 (referent)1 (referent)361 (referent)1 (referent)1 (referent)
 Q2 (2.4–7.5)7190.44 (0.20–0.96)0.46 (0.21–1.01)0.43 (0.19–0.94)250.64 (0.38–1.06)0.63 (0.38–1.06)0.66 (0.39–1.10)
 Q3 (7.7–17.7)80100.43 (0.20–0.91)0.48 (0.22–1.04)0.48 (0.22–1.04)270.60 (0.36–0.99)0.64 (0.39–1.06)0.64 (0.39–1.06)
 Q4 (≥18.3)7490.41 (0.19–0.90)0.42 (0.19–0.93)0.40 (0.18–0.89)230.53 (0.31–0.89)0.52 (0.31–0.89)0.56 (0.33–0.96)
Ptrendb0.110.120.100.0640.0640.10
Combined
 Q1152371 (referent)1 (referent)1 (referent)781 (referent)1 (referent)1 (referent)
 Q2146170.43 (0.24–0.76)0.45 (0.25–0.79)0.42 (0.24–0.75)640.76 (0.55–1.06)0.77 (0.55–1.08)0.76 (0.54–1.06)
 Q3158230.52 (0.31–0.88)0.52 (0.31–0.88)0.54 (0.32–0.91)580.59 (0.42–0.83)0.59 (0.42–0.83)0.62 (0.44–0.88)
 Q4149120.29 (0.15–0.55)0.30 (0.16–0.57)0.29 (0.15–0.56)530.57 (0.40–0.80)0.57 (0.40–0.81)0.61 (0.43–0.87)
Ptrendb0.00110.00130.00060.0450.0570.022
Pinteractionc0.580.500.470.920.880.87
  • aThe multivariate, stage-stratified Cox regression model initially included sex, age, BMI, family history of colorectal cancer in any first-degree relative, year of diagnosis, postdiagnosis aspirin use, postdiagnosis smoking status, postdiagnosis alcohol consumption, tumor location, tumor differentiation, CpG island methylator phenotype, microsatellite instability, LINE-1 methylation, and BRAF and KRAS mutations. A backward elimination with threshold of P = 0.05 was used to select variables in the final models.

  • bTests for linear trend across categories were calculated by using the median value for each quartile of physical activity (MET-h/wk) as a continuous variable in a proportional hazards model.

  • cPinteraction between physical activity quartile and sex/cohort.