Page 143 - 2018_11-Haematologica-web

Dietary patterns and chronic lymphocytic leukemia
Figure 1. Association between adherence to dietary patterns and chron- ic lymphocytic leukemia in the multicase-control (MCC- Spain) study. OR: Odds Ratio; 95%CI: 95% confi- dence interval; Q: quartile; SD: standard deviation. Black squares indicate OR and horizontal lines repre- sent 95%CI. 1Logistic regression models adjusted for age, sex, education, energy intake (kcal/day) with province of residence as random effect.
Statistical analysis
As descriptive analyses, we compared anthropometric, socio- demographic and lifestyle characteristics between cases and con- trols. χ2 test was used to evaluate the level of significance of the differences observed in categorical variables, Student t-test for nor- mally distributed continuous variables, and Wilcoxon rank-sum test for non-normally distributed continuous variables. In addi- tion, we analyzed the distribution of each dietary pattern (contin- uous) across categories of descriptive variables. Student t-test was used to assess differences observed in variables with two cate- gories and ANOVA for those with more than two categories.
The association between the dietary patterns and CLL was eval- uated using mixed logistic regression models with random province-specific intercepts. The exposure variables (adherence to Western, Prudent or Mediterranean patterns) were included in the model both as continuous variables [1-standard deviation (SD) increase in the controls’ scores] and as categorical variables (according to the quartile distribution in all controls). All models were adjusted for age (years, continuous), sex, education (no for- mal education, primary school, secondary school, university), and energy intake (kcal/day, continuous) as fixed effects and province of residence as a random effect term. Height (cm, continuous), waist-to-hip ratio (continuous), Body Mass Index (BMI in kg/m2, continuous), experience working on a farm (yes, no), family histo- ry of hematologic malignancies (yes, no), alcohol consumption (g/day, continuous), smoking status (never, past, current), and physical activity [in the last 10 years, measured in Metabolic Equivalent of Task (METs)/week: inactive (0), low (0.1-8), moder- ate (8-15.9) and very active (≥16)] were examined as potential con- founders, but were not included in the final models as they were not found alone, or in combination, to affect the estimates. Interaction terms were modeled between each of these separate variables and the dietary score (continuous), and tested using log- likelihood ratio tests. A possible effect modification of sex, BMI, energy intake, tobacco, physical activity, working on a farm, and family history of hematologic malignancies was tested including an interaction term between each of the patterns and such vari-
ables. The estimation of the effects according to Rai stage (0 vs. I-IV) was calculated with multinomial logistic regression models adjusted by the set of variables described above plus province of residence as random effect term. Finally, sensitivity analyses were performed to examine how the inclusion of: i) cases with longer period of time from diagnosis to recruitment (<1 year vs. ≥ 1 year); and ii) cases treated before the interview affected the overall esti- mates. Odds Ratios (OR) and 95% confidence intervals (CI) were also obtained with multinomial logistic regression models. The P-value for heterogeneity of effects across Rai stage and for sensi- tivity analyses was obtained with the Wald test. All analyses were performed using STATA/MP (v.14.1, 2015, StataCorp LP) and sta- tistical significance was set at two-sided P<0.05.
Results
Distribution of baseline characteristics between cases and controls is shown in Table 1. Compared with controls, cases were more adherent to the Western pattern, while no differences in level of adherence to the Prudent and Mediterranean patterns were observed in bivariate analy- ses. CLL cases were also slightly older, had a higher waist- to-hip ratio, and were more likely to have a family history of hematologic malignancy and to have worked on a farm. No other differences were observed for any of the other pre-selected variables.
The distribution of key characteristics of controls according to level of adherence to each dietary pattern is shown in Online Supplementary Figure S1. Controls with greater adherence to a Western pattern were more likely to be men, younger, taller, current smokers, less prone to have worked in farming or agriculture, had a lower BMI and waist-to-hip ratio, and a higher level of education, energy and alcohol intake. Those with a higher adherence to a Prudent pattern were more likely to be women, younger, taller, physically active, never/former smokers,
haematologica | 2018; 103(11)
1883