A total of 7,381 persons died during follow-up time, of whom 2,245 (30.4 %) died of cancer and 3,276 (44.4%) of circulatory disease. All the population characteristics are shown in . A higher CCI was observed for those who died during follow-up than for those who were alive at the end of follow-up (4.73% versus 0.68% with CCI=4+) ().
Descriptive statistics of study population by vital status.
A multivariate Cox proportional hazards model including continuous variables of CRP, albumin, GGT, and HDL (adjusted for age, gender, SES, and CCI) showed that each variable was statistically significantly associated with all-cause mortality: HR or one unit increase: 1.08 (95%CI: 1.06-1.11), 0.94 (0.93-0.94), 1.37 (1.33-1.41), and 0.78 (95%CI: 0.73-0.82), respectively. Note that CRP and GGT were log-transformed due to their skewed distributions.
When using the summary mortality score, a clear statistically significant positive trend was observed between the score and all-cause mortality as well as cancer and circulatory disease- specific death (e.g. HR for all-cause mortality: 1.39 (95%CI: 1.32-1.46), 2.04 (95%CI: 1.89-2.21), and 3.36 (95%CI: 2.87-3.93) for score=1, 2, and 3 compared to score=0). Stratification by gender showed similar patterns for men and women ().
Hazard Ratio (HR) and 95% Confidence Intervals (CI) for risk of all-cause, cancer-specific, and circulatory disease death. All models were adjusted for age, gender, socio-economic status, and Charlson Comorbidity Index.
To identify whether these associations were affected by age, a stratified analysis by age groups was conducted (). The same pattern as seen in was observed in . However, the hazard ratios were slightly higher among those in the youngest age group (e.g. HR for overall death when score=3: 4.79 (95%CI: 3.68-6.22), 3.28 (95%CI: 2.45-4.38), and 2.55 (95%CI: 1.95-3.33) for persons aged 50-65, 65-74, and 75+, respectively).
Age-group specific Analysis: Hazard Ratio (HR) and 95% Confidence Intervals (CI) for risk of all cause, cancer-specific, and circulatory disease death. All models were adjusted for age, gender, socio-economic status, and Charlson Comorbidity Index.
In order to compare our mortality score with CCI, we also calculated the HRs for the association between CCI and all-cause mortality, which resulted in similar risks (HR: 1.86 (95%CI: 1.74- 1.98), 2.10 (95%CI: 1.96-2.25), 2.58 (95%CI: 2.31-2.89), and 3.59 (95%CI: 3.57-4.46) for CCI=1, 2, 3, and 4+ compared to CCI=0). The effect of co-morbidity was then assessed with a stratified analysis by values of the CCI (). The patterns observed in were seen in each stratum of CCI, even among those with CCI=0. For instance among those with CCI=2 the risk of cancer-specific death was 1.08 (95%CI: 0.86-1.36), 2.16 (95%CI: 1.53-3.04), and 5.10 (95%CI: 2.83-9.19) for score=1, 2, and 3, compared to score=0 (). The univariate association between our mortality score and the CCI indicated that there was not a strong correlation between both measurements (correlation coefficient: 0.11; p<0.0001 and kappa’s coefficient of agreement: 0.05; p: 0.003).
Table 4 Hazard Ratio (HR) and 95% Confidence Intervals (CI) for risk of all-cause, cancer-specific, and cardiovascular death, stratified by Charlson comorbidity index. All models were adjusted for age, gender, socio-economic status. *Also adjusted for history (more ...)
The sensitivity analysis in which those with follow- up of < 1 year were excluded showed similar, but attenuated, patterns to those observed in , however the results were still statistically significant (results not shown). For instance, the hazard ratios for all-cause death increased with the values of our mortality score: 1.23 (95%CI: 1. 21-1.35), 1.66 (95%CI: 1.43-1.94), and 2.05 (95%CI: 1.48-2.85), for score=1, 2, and 3 compared to score=0.
To assess and illustrate clinical relevance of our mortality score, we calculated the hazard ratios for all-cause death among cancer patients with no other co-morbidities (n=1,955). We found a clear statistically significant positive trend between the score and all-cause mortality (e.g. HR: 1.24 (95%CI: 1.0.-1.49), 2.38 (95%CI: 1.76-3.22), and 5.47 (95%CI: 2.98-10.03) for score=1, 2, and 3 compared to score=0). When assessing death after one year of follow-up, this resulted in a specificity of 0.68 and a sensitivity of 0.59 for a mortality score with a cut-off of 0 versus ≥ 1. The area under the curve (AUC) was 0.65 when using the mortality score as a predictor of death within one year of follow-up ().
ROC curve for the mortality score after 1, 2, and 3 years of follow-up among cancer patients with no other co-morbidities.