# Table 4

Hazard ratios of cancer mortality according to the frequency of nut consumption (including and not including walnuts)
Cancer mortality Never 1 to 3 servings/week >3 servings/week P for trend
Frequency of total nut consumption: n = 2,118 n = 2,803 n = 2,295
Cancer death, % (n) 2.1 (44) 1.9 (52) 1.5 (34)
Person-years, n 8,724 12,168 10,185
Crude model 1 (Reference) 0.82 (0.55 to 1.23) 0.64 (0.41 to 1.00) 0.070
Multivariable model 1 1 (Reference) 0.77 (0.51 to 1.16) 0.54 (0.34 to 0.86) 0.015
Multivariable model 2 1 (Reference) 0.79 (0.52 to 1.20) 0.60 (0.37 to 0.96) 0.052
Multivariable model 3 1 (Reference) 0.79 (0.52 to 1.20) 0.60 (0.37 to 0.98) 0.064
Frequency of walnut consumption: n = 2,916 n = 2,547 n = 1,753
Cancer death, % (n) 2.1 (62) 1.9 (48) 1.1 (20)
Person-years, n 12,124 11,122 7,825
Crude model 1 (Reference) 0.82 (0.56 to 1.20) 0.48 (0.29 to 0.80) 0.005
Multivariable model 1 1 (Reference) 0.76 (0.52 to 1.12) 0.41 (0.25 to 0.69) 0.001
Multivariable model 2 1 (Reference) 0.77 (0.52 to 1.14) 0.46 (0.27 to 0.77) 0.003
Multivariable model 3 1 (Reference) 0.76 (0.51 to 1.12) 0.46 (0.27 to 0.79) 0.005
Frequency of consumption of other nuts (excluding walnuts): n = 3,308 n = 2,643 n = 1,265
Cancer death, % (n) 2.0 (66) 1.6 (43) 1.7 (21)
Person-years, n 13,936 11,573 5,566
Crude model 1 (Reference) 0.77 (0.52 to 1.13) 0.79 (0.48 to 1.29) 0.439
Multivariable model 1 1 (Reference) 0.74 (0.50 to 1.10) 0.68 (0.41 to 1.14) 0.213
Multivariable model 2 1 (Reference) 0.79 (0.53 to 1.18) 0.73 (0.43 to 1.23) 0.318
Multivariable model 3 1 (Reference) 0.79 (0.53 to 1.18) 0.75 (0.44 to 1.27) 0.369

One serving of nuts equals 28 g. Cox regression models were used to assess the risk of cancer mortality by frequency of nut consumption. Multivariable model 1 was adjusted for age (years), sex, and intervention group. Model 2 was additionally adjusted for BMI (kg/m2), smoking status (never, former, current smoker), educational level (illiterate/primary education, secondary education, academic/graduate), leisure time physical activity (MET-min/day), history of diabetes (yes/no), history of hypercholesterolemia (yes/no), use of oral antidiabetic medication (yes/no), use of antihypertensive medication (yes/no), use of statins (yes/no), and total energy intake (kcal/day). Model 3 was also adjusted for dietary variables in quintiles (vegetables, fruits, red meat, eggs, and fish), alcohol intake (continuous, adding a quadratic term) and Mediterranean diet adherence (13-point score). All models were stratified by recruitment centre. Extremes of total energy intake were excluded.