Fig. 4

Hazard ratio for spouses grouped by IP survival in mutual exclusive groups. Spouse groups: group 1 = spouses of whom the IP belonged to the [≥0th & ≤1th percentile] of their birth cohort, group 2 = spouses of whom the IP belonged to the [≥1th & ≤5th percentile] of their birth cohort, group 3 = spouses of whom the IP belonged to the [≥5th & ≤10th percentile] of their birth cohort, group 4 = spouses of whom the IP belonged to the [≥10th & ≤15th percentile] of their birth cohort, group 5 = spouses of whom the IP belonged to the [≥15th & ≤20th percentile] of their birth cohort, group 6 = spouses of whom the IP belonged to the [≥20th & ≤100th percentile] of their birth cohort. The left column (panels a and b) shows the HRs of groups 1–5 compared to group 6. Groups were colored by the extremity of the HR. The darker the blue the stronger the survival benefit, the darker the red, the weaker the survival benefit and the effect was not significant with the red colors. The green lines represent the reference category, which is group 6. N_{green line} at the top-left = 8065, N_{green line} at the bottom-left = 7887. The right column (panels c and d) represents a post-hoc test of all groups and illustrates the p-values for the differences in HR between the spouse groups. p-Values are estimated with Cox regression. Blue color indicates a statistically significant effect after Bonferroni correction, red color indicates a non-statistically significant effect after Bonferroni correction. All estimates are adjusted for religion (UPDB only), sibship size, birth cohort, sex, socio-economic status, mother’s age at birth, birth order, birth intervals, and twin birth. Error bars represent confidence intervals