Relation Of Alcohol Intake To Kidney Function And Mortality Observational, Population-Based, Cohort Study
Jun 28, 2024
3. Results
3.1. Descriptive Statistics
Figure 1 show the number of participants with age ≥ 18 years and complete data at Exam-1, examinees lost to follow-up after Exam-1, examinees who died before undergoing Exam-2, examinees participating in Exam-2, examinees lost to follow-up after Exam-2, examinees who died before undergoing Exam-3, and examinees participating in Exam-3. The mortality-corrected response rate was 75.7% at Exam-2 and 82.5% at Exam-3. Of the 2075 participants in all exams, 6 were excluded due to missing data at Exam-2 and/or Exam-3. The 2069 examinees with complete data at all exams made up the study cohort for the analyses on lab markers of alcohol intake and eGFR, while the 4524 examinees with age ≥ 18 years and complete data at Exam-1 made up the study cohort for the analysis on mortality
Figure 1. Examinees participating in each exam, examinees lost to follow-up, and dead examinees. Table 1 report descriptive statistics on sex, age, alcohol intake, eGFR, and covariates in the 2069 examinees with complete data at the three exams. Urinary albumin/creatinine ratio was measured only at Exam-2 in examinees with age 45–64 years (mg/g: median = 6.0, IQR = 3.4/11.8). Age and eGFR data were not skewed (skewness < 0.6). Alcohol intake was positively skewed at all exams (Figure 2). The median alcohol intake was similar in the three exams, while the IQR range was larger at Exam-2 (Table 1).
Table 1. Descriptive statistics in examinees with complete data at Exam-1, Exam-2, and Exam-3:prevalence for categorical variable, mean ± SD for non-skewed variables, and median (IQR) for skewed variables.
Figure 2. Skewness and frequency distribution of alcohol intake (g/d = g/day) at Exam-1 (black bars), Exam-2 (grey bars), and Exam-3 (white bars) in the 2069 examinees with complete data at all exams.
The Pearson correlation coefficient of alcohol intake was 0.686 between Exam-1 and Exam-2 and was 0.664 between Exam-2 and Exam-3. In alcohol-drinkers, the intake of alcohol in the form of wine accounted for 97.1% of total alcohol intake at Exam-1, for 98.8% at Exam-2, and for 94.7% at Exam-3. At all exams, there were robust trends of the relationship between reported alcohol intake and erythrocytic mean corpuscular volume (Figure 3). A similar trend was found with gamma-glutamyl transferase at Exam-2 (Figure 4).
Figure 3. Mean and 95%CI of erythrocytic mean corpuscular volume by stratum of alcohol intake (g/d = g/day) at Exam-1, Exam-2, and Exam-3. Number of examinees per stratum is reported in Table 1. p-values are from non-adjusted ANOVA.
Figure 4. Mean and 95%CI of serum gamma-glutamyl transpeptidase by stratum of alcohol intake (g/d = g/day) at Exam-2. Number of examinees per stratum is reported in Table 1. p-value is from non-adjusted ANOVA.
At all exams, male sex, age, urinary creatinine, body mass index, and systolic pressure correlated positively with alcohol intake (Supplementary Table S2). Alcohol intake was associated positively with systolic pressure but not with antihypertensive drug treatment at all exams (Table S3). Descriptive statistics in Supplementary Table S4 report data at Exam-1 in examinees participating in all exams and examinees with missing Exam-2 or Exam-3 because of death during follow-up or loss to follow-up. Regarding alcohol intake, the differences between the group of examinees with all exams and the group of examinees with missing exams were ≤2.0% and inconsistent: examinees with all exams had a 2.0% higher prevalence of no alcohol intake (29.0% and 27.0%), 2.0% lower prevalence of alcohol intake 1–24 g/day (43.5% and 45.5%), 1.9% lower prevalence of alcohol intake of 25–48 g/day (11.1% and 13.0%), but 1.9% higher prevalence of alcohol intake > 48 g/day (16.4% and 14.5%). The means of annualized eGFR change and eGFR slope were negative. Annualized eGFR change from Exam-2 to Exam-3 was 1.67-time greater in comparison to annualized eGFR change from Exam-1 to Exam-2 (Table 2).
Table 2. Descriptive statistics in 2069 examinees with complete data at all exams for follow-up duration and eGFR data (mean ± SD).
3.2. Cross-Sectional Analyses on eGFR
At all exams, eGFR differed among strata of alcohol intake (Figure 5, ANOVA without adjustment and with adjustment for covariates). A positive linear trend of eGFR along alcohol strata was significant in multivariable regression at all exams (Exam-1: B= 1.70, 95%CI = 1.00/2.40, p< 0.001; Exam-2: B = 1.03, 95%CI = 0.59/1.48, p < 0.001; Exam-3: B= 0.55, 95%CI = 0.13/0.98, p = 0.010). At Exam-2, the trend was identical also when controlling for log-transformed urinary/albumin ratio in the subgroup with measured urinary albumin (n = 956, age = 45–64 years, B = 1.02, 95%CI = 0.38/1.66, p= 0.002). Compared to no intake, alcohol intake in the range 25–48 g/day was found to be related to higher eGFR both at Exam-1 and at Exam-2, while alcohol intake > 48 g/day related to higher eGFR at all exams (Table 3). Findings were similar when the multivariable model for Exam-1 was analyzed in the examinees who did not participate in follow-up exams (Supplementary Table S5). Covariates independently associated with eGFR at all exams were sex, age, urinary creatinine, body mass index, and serum total cholesterol (Supplementary Table S6).
Figure 5. Cross-sectional analyses: mean and 95%CI of eGFR by stratum of alcohol intake (g/d = g/day) at Exam-1, Exam-2, and Exam-3 in non-adjusted ANOVA (black lines) and ANOVA adjusted for covariates (grey lines). Number of examinees per stratum is in Table 1. p-values are from ANOVA. Covariates in adjusted ANOVA for Exam-1 data: gender and data at Exam-1 for age, education, log-transformed urinary sodium/creatinine ratio, log-transformed urinary potasium/creatinine ratio, urinary creatinine, body mass index, systolic pressure, diastolic pressure, anti-hypertensive drug treatment, serum total cholesterol, smoking, and diabetes. Covariates in adjusted ANOVA for Exam-2 data: gender and data at Exam-2 for age, education, log-transformed urinary sodium/creatinine ratio, log-transformed urinary potassium/creatinine ratio, log-transformed urinary urea nitrogen/creatinine ratio, urinary creatinine, body mass index, systolic pressure, diastolic pressure, antihypertensive drug treatment, serum total cholesterol, smoking, and diabetes. Covariates in adjusted ANOVA for Exam-3 data: gender and data at Exam-3 for age, education, urinary creatinine, body mass index, systolic pressure, diastolic pressure, antihypertensive drug treatment, serum total cholesterol, smoking, and diabetes.
Table 3. Cross-sectional analysis: multi-variable linear regression models for data of Exam-1, Exam-2,and Exam-3 with eGFR regressed over stratum of alcohol intake.
Regression coefficient (B), 95% confidence interval (italic), and p-value. Covariates included in Model 1: gender and data at Exam-1 for age, education, log-transformed urinary sodium/creatinine ratio, log-transformed urinary potassium/creatinine ratio, urinary creatinine, body mass index, systolic pressure, diastolic pressure, antihypertensive drug treatment, serum total cholesterol, smoking, and diabetes. Covariates included in Model 2: gender and data at Exam-2 for age, education, log-transformed urinary sodium/creatinine ratio, log-transformed urinary potassium/creatinine ratio, log-transformed urinary urea nitrogen/creatinine ratio, urinary creatinine, body mass index, systolic pressure, diastolic pressure, antihypertensive drug treatment, serum total cholesterol, smoking, and diabetes. Covariates included in Model 3: gender and data at Exam-3 for age, education, urinary creatinine, body mass index, systolic pressure, diastolic pressure, antihypertensive drug treatment, serum total cholesterol, smoking, and diabetes. Number of examinees per stratum of alcohol intake is the same as shown in Table 1.
3.3. Longitudinal Analyses on eGFR
Annualized eGFR changes differed among strata of alcohol intake either in the follow-up from Exam-1 to Exam-2 and in the follow-up from Exam-2 to Exam-3 in ANOVA without adjustment and with adjustment for covariates (Figure 6, upper and intermediate panels). A positive linear trend of eGFR change along alcohol strata was significant in the multivariable regression for follow-up from Exam-1 to Exam-2 (B = 0.133, 95%CI = 0.049/0.216, p = 0.002) and for follow-up from Exam-2 to Exam-3 (B = 0.065, 95%CI = 0.024/0.111, p = 0.004). The trend for follow-up from Exam-2 to Exam-3 was similar when also controlling for log-transformed urinary/albumin ratio in the subgroup with measured urinary albumin (n = 956, age = 45–64 years, B = 0.063, 95%CI = 0.013/0.132, p = 0.035).
