br We used the statistical software package STATA
We used the statistical software package STATA version 14.0 (STATA, College Station, TX) for analysis. First, we performed uni-variate analysis to evaluate whether median levels of serum CRP dif-fered according to selected characteristics (Table 1). Nonparametric Wilcoxon rank-sum test was applied. To further test for trend across ordered groups, nonparametric trend test was applied. Then, using median levels of serum CRP among all study participants as the cutoff point (CRP = 3.20), we stratified the study participants into two groups, low and high levels of CRP. We then conducted multivariable logistic regression analyses to assess the association between selected characteristics and high levels of CRP (Table 2). Age category, sex, marital status, education, place of birth, language acculturation, ci-garette smoking status, alcohol drinking status, BMI category, and physical activity were included in the model. Adjusted odds ratios (ORs) and 95% confidence intervals (CIs) were estimated. Then, among those born in the Mexico, we explored whether years of living in U.S. might affect median CRP levels and risk of high CRP levels (Table 3). Finally, the association between serum CRP levels and incidents of cancer was assessed using Cox proportional hazards regression models, controlling for variables significantly associated with CRP levels, in-cluding age group, sex, alcohol consumption status, BMI category, and physical activity levels (Table 4). Study participants were followed up from their initial enrollment to the end of 2016. The median follow-up time 127 months. Those who were still cancer-free at the end of 2016 were censored in the Cox regression analysis. For those without cancer diagnosed but died during the follow-up, they Adriamycin were censored at the time of their death. Death was identified during the annual follow-up and further confirmed with the death report. The proportional hazards as-sumption was tested using the Schoenfeld residuals. Serum CRP levels were analyzed as the categorical variable. The cutoff points were set at
1) the median and quartile values in the study participants and 2) clinical values. Adjusted hazard ratios (HRs) and 95% CIs were Cancer Epidemiology 60 (2019) 1–7
estimated. I further stratified analysis, we assess whether the cancer risk differed by the follow-up time, 1 to 5 years vs at least 5 years. All sta-tistical tests were 2-sided, and P values of less than 0.05 were con-sidered statistically significant.
For basic demographic characteristics, serum CRP levels differed by sex, age category, marital status, and education levels, but not by place of birth and language acculturation (Table 1). Women had significantly higher median levels of CRP than men did (3.75 vs 2.00, P < 0.001). A trend of increasing median levels of serum CRP was observed with the increase of age category from < 30, 30 to 39, 40 to 49, 50 to 59, to ≥60 years old (P for trend < 0.001). Those who married had significantly higher median levels of CRP than their counterparts did (3.31 vs 3.01, P = 0.009). And those who had at least high school degree had sig-nificantly lower levels median levels of CRP than their counterparts did (3.00 vs 3.40, P = 0.018). However, when stratified by sex, the sig-nificant associations were only observed for age category for both women and men, and marital status for women only. Interestingly, men cerebral cortex were born in U.S. had significantly higher median levels of CRP than those born in Mexico (2.25 vs 1.80, P = 0.029). Similar difference was also observed for women, but the P value didn’t reach statistical significance (P = 0.076).
For healthy behavior variables, serum median levels of CRP differed by cigarette smoking status, alcohol drinking status, BMI category, and physical activity level (Table 1). Specifically, with the increase of BMI category from normal weight (BMI < 25), overweight (BMI: 25 to 30), class I obesity (BMI: 30 to 35), class II obesity (BMI: 35 to 40), to class
III obesity (BMI: ≥40), median levels of CRP increased steadily and significantly (P for trend < 0.001). On the other hand, a trend of de-creasing median CRP levels was noted for cigarette smoking and al-cohol drinking status from never, former, to current, and physical ac-
tivity levels from low, medium, to high (P for trend = 0.002, < 0.001, < 0.001, respectively). However, in the stra-tified analysis by sex, the significant trend of decreasing median CRP levels was no longer existed for cigarette smoking and alcohol drinking status in both men and women. The significant trend was remained for BMI category and physical activity levels in both men and women.
Next, we performed multivariable logistic regression analysis to identify demographic characteristics and health behaviors associated with high levels of CRP, defined as ≥ median levels among all study participants (Table 2). The included variables were age category, sex, marital status, education, place of birth, language acculturation, ci-garette smoking status, alcohol drinking status, BMI category, and physical activity. The variables significantly associated with high levels of CRP included sex, age group, alcohol drinking status, BMI category, and physical activity levels. Specifically, compared to men, women had 1.88-fold increased likelihood of having high levels of CRP (OR = 1.88, 95%CI: 1.48, 2.39). With the increasing of age, the likelihood of having high levels of CRP was significantly increased (P for trend = 0.018). Compared to those with normal weight, those with overweight, class I obesity, class II obesity, to class III obesity had significantly increased likelihood of having high levels of CRP. And a dose-response trend of increasing likelihoods was observed across the BMI category (P for trend < 0.001). On the other hand, compared to those who never consumed alcohol drinking, those who currently consumed had 21% decreased likelihood of having high levels of CRP (OR = 0.79, 95%CI: 0.60, 0.97). And compared to those with low levels of physical activity, those with medium levels of physical activity had 19% decreased risks of having high levels of CRP (OR = 0.81, 95%CI: 0.66, 0.99). When stratified by sex, BMI category was the only variable left significantly associated with high levels of CRP among women. In addition, a sig-nificant dose-response trend was observed between levels of physical activity and high levels of CRP (P = 0.045). Among men, both age category and BMI category were still significantly associated with high