Fig 3. Relative risk of death by CRP concentration.
Relative risk of death was estimated as hazard ratios (HRs) using Cox proportional hazard models. Unadjusted HRs of all deaths correlated with log CRP concentrations (A). To exclude possible confounders, HRs of all deaths were adjusted for age, sex, PD disease duration, MMSE (≤24 vs. >24), mH-Y (1–3 vs. 4–5), and serum albumin levels, and the analysis showed a log-linear association with CRP concentrations (B). Similarly unadjusted and adjusted HRs of death, excluding PD-unrelated deaths, correlated with CRP concentrations (C, D). Unadjusted and age- and sex-adjusted HRs of deaths from pneumonia correlated with CRP concentrations (E, F). Unadjusted and age- and sex-adjusted HRs of sudden deaths correlated with CRP concentrations, although there was no significance (G, H). Unadjusted and age- and sex-adjusted HRs of deaths from cancer correlated with CRP concentration, although there was no significance (I, J).
