Significance statement
The prognostic significance of variabilities in body mass index (BMI) or metabolic parameters in patients with CKD is uncertain. In this observational cohort study of 84,636 patients with predialysis CKD in South Korea, the authors analyzed the association between variability of BMI or various metabolic parameters and risks of all-cause mortality and incident myocardial infarction, stroke, and requirement for KRT. They found that elevated variability in BMI or certain metabolic parameters was associated with higher risks of adverse outcomes, independent of baseline metabolic status. These findings may encourage clinicians in the nephrology field to carefully assess not only baseline BMI or metabolic status in patients with CKD, but also the fluctuating status of metabolic parameters, due to their potential prognostic significance in such patients.
Keywords: body mass index, epidemiology, chronic kidney disease
Visual Abstract
Abstract
Background
The association between variabilities in body mass index (BMI) or metabolic parameters and prognosis of patients with CKD has rarely been studied.
Methods
In this retrospective observational study on the basis of South Korea’s national health screening database, we identified individuals who received ≥3 health screenings, including those with persistent predialysis CKD (eGFR <60 ml/min per 1.73 m2 or dipstick albuminuria ≥1). The study exposure was variability in BMI or metabolic parameters until baseline assessment, calculated as the variation independent of the mean and stratified into quartiles (with Q4 the highest quartile and Q1 the lowest). We used Cox regression adjusted for various clinical characteristics to analyze risks of all-cause mortality and incident myocardial infarction, stroke, and KRT.
Results
The study included 84,636 patients with predialysis CKD. Comparing Q4 versus Q1, higher BMI variability was significantly associated with higher risks of all-cause mortality (hazard ratio [HR], 1.66; 95% confidence interval [95% CI], 1.53 to 1.81), P [for trend] <0.001), KRT (HR, 1.20; 95% CI, 1.09 to 1.33; P<0.001), myocardial infarction (HR, 1.19; 95% CI, 1.05 to 1.36, P=0.003), and stroke (HR, 1.19; 95% CI, 1.07 to 1.33, P=0.01). The results were similar in the subgroups divided according to positive or negative trends in BMI during the exposure assessment period. Variabilities in certain metabolic syndrome components (e.g., fasting blood glucose) also were significantly associated with prognosis of patients with predialysis CKD. Those with a higher number of metabolic syndrome components with high variability had a worse prognosis.
Conclusions
Higher variabilities in BMI and certain metabolic syndrome components are significantly associated with a worse prognosis in patients with predialysis CKD.
CKD is a major morbidity that is associated with a substantial socioeconomic burden.1 Patients with CKD suffer from high risks of cardiovascular disease and mortality. The prevalence of CKD is projected to increase, due to the global trend of aging populations and increasing rates of obesity.
Because the development and prognosis of CKD is largely associated with hypertension and diabetes,2 body mass index (BMI), which is closely related to these two metabolic disorders, has been considered an important clinical factor with prognostic significance in the CKD population. Because BMI and metabolic syndrome status change dynamically in the general population, recent population-scale evidence has shown the variability in BMI and metabolic syndrome components has distinct prognostic significance for various adverse outcomes, including kidney function impairment.3,4 Higher BMI variability, or weight cycling, has been reported to increase fat components or reflect metabolic instability and has further been associated with higher risks of cardiovascular diseases and mortality.5 However, there have been few large-scale studies including patients with CKD that have investigated the prognostic significance of BMI or metabolic parameter variability in the context of important health outcomes. Such evidence would encourage health care providers to pay attention to recent trends in BMI and metabolic status, which could help clinicians appropriately interpret important clinical parameters in the growing CKD population. In particular, because higher baseline BMI is associated with a better prognosis in patients with CKD (i.e., “the obesity paradox”),6 which conflicts with the idea that a higher BMI increases the risk of adverse outcomes in the general population,7 additional investigation of the prognostic significance of another BMI parameter, BMI variability, is warranted.
In this study, we aimed to investigate the association between BMI and metabolic syndrome component variability and risks of adverse outcomes in patients with predialysis CKD in a nationwide health screening database linked to claims information in South Korea. We studied the risks of all-cause mortality, myocardial infarction, stroke, and progression to the state requiring maintenance KRT, with consideration of the trends in BMI. We further extended our assessment to the variability of various metabolic syndrome components, which also reflect fluctuations in metabolic status. We hypothesized that a higher BMI or metabolic parameter variability would be associated with higher risks of adverse outcomes in patients with CKD, independent of baseline metabolic parameters.
Methods
Ethics Considerations
The study was performed in accordance with the principles of the Declaration of Helsinki and was approved by the institutional review board of Seoul National University Hospital (E-2001–112–1096). The investigation of the Korea National Health Insurance Service (NHIS) database was approved by the relevant government organization (REQ000034394). The need to obtain informed consent was waived by the above organizations, because the study was observational and investigated anonymous public databases.
Study Setting
This study was a retrospective observational cohort study. The study data comprised nationwide health screening data from the Korea NHIS, which have been previously described.8,9 In South Korea, a free-of-charge nationwide health screening program is provided to the general adult population that includes clinicodemographic assessments, lifestyle evaluations, and laboratory tests.2 Health screening is provided annually or biennially, and the quality of health screening centers is monitored and controlled by the NHIS. The data are also linked to the nationwide claims database, which includes information on all insured medical services nationwide. Because national health insurance is provided to all Korean citizens and essential medical services are generally insured, the database has been utilized for many population-scale studies investigating outcomes, including cardiocerebrovascular diseases, ESKD, and mortality.3,10–12
In this study, BMI variability was identified on the basis of the BMI calculated at every health screening visit. The health screenings also involve laboratory tests and waist circumference measurements to assess major metabolic syndrome components. Predialysis CKD was identifiable on the basis of serum creatinine levels and dipstick albuminuria, which were measured at every health screening visit with ≥1-year intervals, and insurance codes specified ESKD events. The study outcomes and medical histories were collected from the claims database. A graphical description of the time windows for determining the study population, collecting the covariate data, and identifying the outcomes is presented in Figure 1.13
Figure 1.
Graphical depiction of the time windows used to determine the studied variables. S indicates the national health screenings that were mostly performed at annual or biennial intervals.
Study Population
The study identified individuals who had undergone baseline health screenings from 2013 to 2014. As multiple health screenings are needed to calculate BMI or metabolic parameter variability, the study included individuals who had ≥3 health screenings before the baseline visit. After excluding those with missing information for the collected variables, those who had persistent predialysis CKD (eGFR <60 ml/min per 1.73 m2 or dipstick albuminuria ≥1+) in the baseline assessment period were retained in the study population, with the exclusion of patients with ESKD with an eGFR <15 ml/min per 1.73 m2. Because we aimed to investigate the risks of incident events of myocardial infarction, stroke, and maintenance KRT, those who had a history of the outcomes before follow-up were excluded.
Ascertainment of BMI Exposure
BMI values were calculated from the height and weight measured at the health screening visits. BMI variability was primarily determined by variation independent of the mean (VIM).3 The VIM has advantages over other variability indexes, because the parameter is calculated from a regression model to be independent of the mean level. The other well-known variability indexes were also calculated, including standard deviation; coefficient of variation, which divides the standard deviation by the mean value; and average real variability, calculated as the sum of the differences between adjacent values divided by the number of gaps. The collected exposures were divided into quartiles, and high variability was defined as the highest quartile (Q4) variability level. The baseline BMI was also utilized as a supplemental exposure, to determine whether an inverse association between baseline BMI and the survival of patients with CKD also existed in this cohort.
Ascertainment of Metabolic Syndrome Component Exposure
We extended our assessment to various metabolic parameters, because BMI variability was likely to be closely linked to variabilities in metabolic syndrome components. The assessed metabolic parameters included the components of metabolic syndrome defined in the harmonized criteria14: waist circumference, fasting blood glucose, systolic BP, diastolic BP, serum triglycerides, and HDL. We also assessed other cholesterol parameters, including low-density lipoprotein and total cholesterol levels. The VIM was calculated as above and stratified into quartiles.
Construction of Cumulative Metabolic Variability Score
We constructed an ordinal score to reflect the overall burden of metabolic variability. We defined the score by summing the number of high-variability (Q4) metabolic parameters, selecting a component from each metabolic syndrome domain (obesity, impaired glucose tolerance, high BP, and dyslipidemia).3 BMI, fasting blood glucose, systolic BP, and total cholesterol were included in the ordinal scoring as in a previous study.3
Ascertainment of Study Outcomes
The main study outcome was all-cause mortality on the basis of the claims database, which includes nationwide mortality events collected from the death certificates. Additional incident risks of major adverse outcomes in patients with predialysis CKD were collected. Incident KRT was determined by specific insurance coverage codes for “maintenance” KRT in the NHIS data, including hemodialysis, peritoneal dialysis, and transplantation events. As in the previous study,3 myocardial infarction was recorded if an individual had International Classification of Diseases 10th revision (ICD-10) codes I21 or I22 during hospitalization, or if these codes were issued ≥2 times. Stroke was defined as ICD-10 codes I63 or I64 during hospitalization, with claims information for brain magnetic resonance imaging or brain computerized tomography imaging. Follow-up was initiated after the date of the baseline health screening visit at which the variability exposure assessment was completed (day 1) and was censored on the last date of data availability or at the date of death, to consider the competing risk by mortality (Figure 1).
Collection of the Covariate Data
We aimed to include a range of data, including demographic information, lifestyle factors, baseline kidney function, and various metabolic parameters, as covariates.3 We collected baseline age, sex, waist circumference, current smoking history, alcohol intake (>0 g of alcohol intake per day), regular physical activity (moderate-intensity physical activity ≥5 days or vigorous-intensity physical activity ≥3 days per week), low-income status (lower quartile of the nation), diabetes mellitus (ICD-10 codes E11–14 with relevant antidiabetes medication history), hypertension (ICD-10 codes I10–13 or I15 with relevant antihypertensive medication history), dyslipidemia (ICD-10 code E78 with relevant dyslipidemia medication history), chronic lung disease (ICD-10 codes J41–44), cancer (specific insurance code for malignancies in the NHIS data), baseline eGFR, presence of dipstick albuminuria (≥1+), fasting glucose, systolic BP, diastolic BP, high-density lipoprotein, low-density lipoprotein, and triglycerides. We also calculated the variability in eGFR during the exposure assessment period, because including the covariates in the multivariable model could adjust the effects of a direct association between BMI variability and eGFR variability.15 In addition, the number of examinations during the exposure period was also collected as a covariate, because the number of examinations could affect variability parameters or reflect the health-seeking behavior of a subject.
Statistical Analysis
We first assessed the risk of adverse outcomes according to BMI exposure, both ordinal or continuous, by Cox regression analysis. In addition to the univariable model and a model adjusted for age, sex, and the number of exams, a multivariable model including various collected covariates was constructed. To show the cumulative risks of the study outcomes, we plotted adjusted survival curves on the basis of the final multivariable model. We also performed subgroup analyses, with stratification according to sex, baseline BMI ≥25 kg/m2, the presence of diabetes, and baseline eGFR <60 ml/min per 1.73 m2. Furthermore, we assessed interactions by calculating interaction term P values in the multivariable model. We selected BMI ≥25 kg/m2, which is the definition for overweight in the international guidelines, because obesit is determined by the cutoff in the Korean guidelines on the basis of the nationwide BMI distribution.16 Because this value was close to the mean level in the study population, selecting it was further supported by the fact it would divide participants into subgroups with similar numbers.
In addition, because the clinical significance of BMI variability may be different in those with increasing and decreasing trends in BMI, we determined the trends by calculating the linear regression slope of the individuals’ collected BMI values. An increasing trend was defined as a slope ≥0 (positive), and a decreasing trend was defined as a slope <0 (negative). The Cox regression analysis described above was also performed in the subgroups stratified by the BMI changes and the quartiles of BMI variability.
Next, the variability in the collected metabolic parameters and their associations with the adverse outcomes were assessed by Cox regression analysis as above. Finally, the ordinal score for overall metabolic variability was constructed as described above, and its association with the risks of adverse outcomes were assessed by Cox regression analysis, as above.
All statistical analyses were performed with SAS (version 9.4, SAS institute), and two-sided P values <0.05 were considered statistically significant. When the ordinal outcome was included in the regression models as the exposure variable, the P for trend for the higher exposure grades was calculated. Interaction term P values <0.1 were considered indicative of the possible presence of an interaction.
Results
Study Population
We identified 11,651,753 individuals who underwent ≥3 health screenings during the exposure assessment period (Figure 2). There were 5,402,903 individuals with two health screenings; however, although the information may determine CKD, this population was not considered because metabolic variabilities require ≥3 health screenings. Among them, 128,658 individuals had a persistent eGFR <60 ml/min per 1.73 m2 or dipstick albuminuria ≥1. After excluding those with prevalent ESKD, myocardial infarction, and stroke, 84,636 patients with predialysis CKD with identifiable BMI or metabolic parameter variability were included in the study population.
Figure 2.
Study population. MI, myocardial infarction.
Baseline Characteristics
The study population had a median age of 68 (IQR, 60–74) years old, and 51% were male. The median BMI value was 24.6 (IQR, 22.6–26.7) kg/m2. When we stratified the characteristics according to BMI variability (Table 1), the number of health exams did not show a linear trend according to BMI variability. Those with lower BMI variability had a higher proportion of individuals who engaged in regular physical activity. The prevalence of diabetes was the highest in those with high BMI variability, which was also the stratum with the highest proportion of subjects who had a baseline eGFR ≥60 ml/min per 1.73 m2 but had albuminuria. The distributions of the other metabolic parameters were relatively similar among the BMI variability quartiles.
Table 1.
Baseline characteristics of the study population according to BMI variability
| Characteristics | Total (n=84,636) | Q1 (n=21,159) | Q2 (n=21,159) | Q3 (n=21,159) | Q4 (n=21,159) |
|---|---|---|---|---|---|
| Age, yrs | 68 (60, 74) | 68 (60, 74) | 68 (58, 74) | 68 (60, 74) | 70 (62, 76) |
| Male sex | 41,665 (49) | 11,397 (54) | 11,096 (52) | 10,421 (49) | 8751 (41) |
| BMI, kg/m2 | 24.61 (22.59, 26.73) | 25 (23.18, 26.95) | 24.75 (22.86, 26.78) | 24.56 (22.59, 26.67) | 24 (21.72, 26.44) |
| BMI variability | |||||
| Variation independent mean, unit | 0.64 (0.40, 0.98) | 0.27 (0.19, 0.34) | 0.52 (0.46, 0.58) | 0.78 (0.71, 0.87) | 1.3 (1.11, 1.65) |
| Standard deviation, kg/m2 | 0.66 (0.41, 1.02) | 0.27 (0.2, 0.36) | 0.54 (0.47, 0.61) | 0.81 (0.72, 0.92) | 1.35 (1.14, 1.71) |
| Coefficient of variation, % | 2.69 (1.68, 4.11) | 1.13 (0.8, 1.45) | 2.17 (1.93, 2.42) | 3.29 (2.98, 3.67) | 5.47 (4.68, 6.95) |
| Actual real variability, kg/m2 per measure | 0.77 (0.48, 1.19) | 0.33 (0.21, 0.44) | 0.63 (0.5, 0.8) | 0.92 (0.75, 1.17) | 1.54 (1.2, 2.05) |
| Number of health screenings | |||||
| 3 | 74,407 (88) | 18,990 (90) | 17,926 (85) | 18,242 (86) | 19,249 (91) |
| 4 | 5680 (7) | 1248 (6) | 1621 (8) | 1632 (8) | 1179 (6) |
| 5 | 4549 (5) | 921 (4) | 1612 (8) | 1285 (6) | 731 (3) |
| Social factors | |||||
| Urban residence | 38,315 (45) | 9956 (47) | 9861 (47) | 9627 (46) | 8871 (42) |
| Low income (<25th percentile) | 15,790 (19) | 3629 (17) | 3918 (19) | 3985 (19) | 4258 (20) |
| Lifestyle factors | |||||
| Current smoker | 10,374 (12) | 2549 (12) | 2724 (13) | 2699 (13) | 2402 (11) |
| Alcohol intake (> 0 g/day) | 22,862 (27) | 6319 (30) | 6298 (30) | 5840 (28) | 4405 (21) |
| Regular physical activity | 19,242 (23) | 5285 (25) | 5131 (24) | 4848 (23) | 3978 (19) |
| Comorbidities | |||||
| Diabetes mellitus | 25,019 (30) | 5809 (27) | 5919 (28) | 6196 (29) | 7095 (34) |
| Hypertension | 58,703 (69) | 14,587 (69) | 14,483 (68) | 14,736 (70) | 14,897 (70) |
| Cancer | 3608 (4) | 797 (4) | 830 (4) | 894 (4) | 1087 (5) |
| Chronic lung disease | 9689 (11) | 2181 (10) | 2256 (11) | 2443 (12) | 2809 (13) |
| Dyslipidemia | 36,697 (43) | 9136 (43) | 9202 (43) | 9272 (44) | 9087 (43) |
| Chronic heart failure | 4232 (5) | 862 (4) | 913 (4) | 1057 (5) | 1400 (7) |
| Atrial fibrillation | 2327 (3) | 498 (2) | 551 (3) | 600 (3) | 678 (3) |
| Medication history | |||||
| ACE I or ARB | 33,419 (39) | 8429 (40) | 8531 (40) | 8278 (39) | 8181 (39) |
| Insulin | 7425 (9) | 1510 (7) | 1656 (8) | 1838 (9) | 2421 (11) |
| Statin | 34,313 (41) | 8514 (40) | 8559 (40) | 8663 (41) | 8577 (41) |
| Laboratory/anthropometric findings | |||||
| Waist circumference, cm | 86 (80, 91) | 86 (80, 91) | 85 (79, 90) | 84 (79, 90) | 83 (77, 90) |
| Systolic BP, mmHg | 130 (120, 139) | 130 (120, 139) | 130 (120, 139) | 130 (120, 139) | 130 (120, 140) |
| Diastolic BP, mmHg | 79 (70, 83) | 79 (70, 83) | 79 (70, 83) | 79 (70, 83) | 79 (70, 83) |
| Fasting glucose, mg/dl | 101 (92, 117) | 101 (92, 117) | 100 (91, 117) | 100 (91, 117) | 101 (91, 118) |
| Total cholesterol, mg/dl | 188 (163, 216) | 188 (163, 216) | 189 (163, 218) | 188 (162, 217) | 185 (159, 215) |
| HDL, mg/dl | 48 (40, 57) | 48 (40, 57) | 48 (41, 57) | 48 (41, 58) | 49 (41, 59) |
| LDL, mg/dl | 109 (85, 134) | 109 (85, 134) | 109 (85, 135) | 108 (84, 134) | 106 (83, 133) |
| Triglycerides, mg/dl | 134 (96, 188) | 134 (96, 188) | 134 (95, 188) | 131 (94, 184) | 126 (91, 178) |
| Hemoglobin, g/dl | 13.4 (12.2, 14.7) | 13.4 (12.2, 14.7) | 13.3 (12.1, 14.6) | 13.1 (12, 14.4) | 12.7 (11.5, 13.9) |
| Urine albuminuria, ≥ 1+ | 27220 (32) | 6823 (32) | 6976 (33) | 6913 (33) | 6508 (31) |
| Serum Cr,mg/dl | 1.3 (1.1, 1.5) | 1.3 (1.1, 1.5) | 1.3 (1.1, 1.5) | 1.3 (1.1, 1.5) | 1.3 (1.1, 1.5) |
| eGFR, ml/min per 1.73 m2 | 52.9 (45.85, 58.52) | 52.74 (45.46, 58.56) | 52.28 (44.5, 58.35) | 51.41 (42.39, 57.71) | |
| ≥60 | 4905 (6) | 960 (5) | 1085 (5) | 1241 (6) | 1619 (8) |
| ≤30 to <60 | 65844 (78) | 16634 (79) | 16400 (78) | 16345 (77) | 16465 (78) |
| ≤15 to <30 | 13887 (16) | 3565 (17) | 3674 (17) | 3573 (17) | 3075 (15) |
Continuous variables are presented as medians (interquartile ranges), and categorical variables are presented as frequencies (percentages). There were no missing data in the table.
Prognosis According to Baseline BMI and BMI Variability
During the median follow-up of 4.0 (IQR, 3.4–4.5) years, 4782 (6%) mortality, 3276 (4%) maintenance KRT, 1839 (2%) myocardial infarction, and 2666 (3%) stroke events were identified. In the regression analysis, baseline BMI was inversely associated with the risks of all studied adverse outcomes after multivariable adjustment (Supplemental Table 1). However, those with higher BMI variability had significantly higher risks of adverse outcomes (Table 2, Figure 3). Those with high (Q4 VIM) BMI variability had approximately 60% higher hazards for all-cause mortality and 20% higher hazards for myocardial infarction, stroke, and KRT, than those with low variability (Q1 VIM), even after adjusting for various characteristics, including baseline BMI, baseline eGFR, and eGFR variability during the exposure assessment period. The results were similar when other variability indexes were used as the exposure variable (Supplemental Table 2).
Table 2.
Risk of adverse outcomes according to BMI variability (VIM) in patients with predialysis CKD
| Outcome | BMI Variability Independent of Mean Exposure | N | Event | Follow-up Person-years | Incidence Rate (/1000 Person-years) | Univariable Model | Age, Sex, and Number of Exams Adjusted Model | Multivariable Modela | ||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HR (95% CI) | P for trend | Adjusted HR (95% CI) | P for trend | Adjusted HR (95% CI) | P for trend | |||||||
| All-cause mortality | Continuous | NA | 1.37 (1.33 to 1.40) | <0.001 | 1.30 (1.26 to 1.33) | <0.001 | 1.20 (1.17, 1.23) | <0.001 | ||||
| Categorical | Q1 (low) | 21,159 | 847 | 82,939 | 10.21 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | |
| Q2 | 21,159 | 878 | 83,500 | 10.52 | 1.03 (0.93 to 1.13) | 1.07 (0.97 to 1.17) | 1.03 (0.94 to 1.14) | |||||
| Q3 | 21,159 | 1105 | 83,184 | 13.28 | 1.297 (1.186 to 1.419) | 1.298 (1.187 to 1.420) | 1.172 (1.071 to 1.282) | |||||
| Q4 (high) | 21,159 | 1952 | 81,560 | 23.93 | 2.349 (2.167 to 2.547) | 2.068 (1.906 to 2.244) | 1.662 (1.530 to 1.806) | |||||
| Myocardial infarction | Continuous | NA | 1.140 (1.094 to 1.188) | <0.001 | 1.113 (1.068 to 1.160) | <0.001 | 1.072 (1.028 to 1.117) | 0.001 | ||||
| Categorical | Q1 (low) | 21,159 | 410 | 82,361 | 4.98 | Reference | <0.001 | Reference | <0.001 | Reference | 0.003 | |
| Q2 | 21,159 | 388 | 82,892 | 4.68 | 0.937 (0.815 to 1.076) | 0.959 (0.834 to 1.102) | 0.942 (0.820 to 1.083) | |||||
| Q3 | 21,159 | 470 | 82,461 | 5.70 | 1.141 (1.000 to 1.303) | 1.145 (1.002 to 1.307) | 1.092 (0.956, 1.247) | |||||
| Q4 (high) | 21,159 | 571 | 80,783 | 7.07 | 1.421 (1.252 to 1.613) | 1.332 (1.172 to 1.514) | 1.191 (1.046 to 1.356) | |||||
| Stroke | Continuous | NA | 1.128 (1.090 to 1.168) | <0.001 | 1.090 (1.053 to 1.128) | <0.001 | 1.058 (1.020 to 1.093) | 0.002 | ||||
| Categorical | Q1 (low) | 21,159 | 584 | 82,008 | 7.12 | Reference | <0.001 | Reference | <0.001 | Reference | 0.01 | |
| Q2 | 21,159 | 610 | 82,459 | 7.40 | 1.038 (0.926 to 1.162) | 1.07 (0.955 to 1.199) | 1.059 (0.945 to 1.187) | |||||
| Q3 | 21,159 | 652 | 82,106 | 7.94 | 1.114 (0.997 to 1.246) | 1.115 (0.997 to 1.247) | 1.074 (0.96 to 1.202) | |||||
| Q4 (high) | 21,159 | 820 | 80,306 | 10.21 | 1.436 (1.291 to 1.597) | 1.304 (1.172 to 1.451) | 1.189 (1.067 to 1.325) | |||||
| KRT | Continuous | NA | 1.183 (1.147 to 1.220) | <0.001 | 1.241 (1.203 to 1.281) | <0.001 | 1.069 (1.035 to 1.103) | <0.001 | ||||
| Categorical | Q1 (low) | 21,159 | 646 | 82,054 | 7.87 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | |
| Q2 | 21,159 | 754 | 82,395 | 9.15 | 1.16 (1.045 to 1.289) | 1.135 (1.021 to 1.261) | 0.991 (0.892 to 1.101) | |||||
| Q3 | 21,159 | 832 | 81,896 | 10.16 | 1.29 (1.164 to 1.430) | 1.319 (1.190 to 1.462) | 1.066 (0.962 to 1.182) | |||||
| Q4 (high) | 21,159 | 1044 | 79,957 | 13.06 | 1.67 (1.514 to 1.842) | 1.906 (1.728 to 2.104) | 1.201 (1.087 to 1.327) | |||||
Multivariable model was adjusted for age, sex, number of exams, current smoking, drinking alcohol consumption, regular physical activity, low-income status, history of diabetes mellitus, hypertension, dyslipidemia, cancer, chronic lung disease, baseline BMI, waist circumference, fasting glucose, systolic BP, diastolic BP, HDL, baseline eGFR, presence of dipstick albuminuria, and eGFR variability during the exposure assessment period.
Figure 3.
Adjusted survival curves showing the cumulative risks of the study outcome. The y axes indicate cumulative adjusted incidence probability, and the x axes indicate the time (years). The survival curves are stratified by BMI variability quartiles (black: Q1, low variability; red: Q2; green: Q3; and blue: Q4, high variability). The survival tables are presented below the adjusted survival curves. A multivariable model was adjusted for age, sex, number of exams, current smoking, alcohol consumption, regular physical activity, low-income status, history of diabetes mellitus, hypertension, dyslipidemia, cancer, chronic lung disease, baseline BMI, waist circumference, fasting glucose, systolic BP, diastolic BP, HDL, baseline eGFR, presence of dipstick albuminuria, and eGFR variability during the exposure assessment period.
Subgroup Analysis Results for BMI Variability
After stratification by sex, there was a suspected interaction with all-cause mortality; otherwise, the interactions were nonsignificant (Supplemental Table 3). Nevertheless, the risk of all-cause mortality was significantly higher in those with higher BMI variability in both sex subgroups, and this was similar to the result for KRT risk in the fully adjusted model (Supplemental Table 4). Although the statistical significance remained in the model adjusted by age, sex, and number of health screenings, the significance was attenuated for stroke or myocardial infarction outcome.
Baseline BMI ≥25 kg/m2 had possible interactions with all-cause mortality and KRT outcomes (Supplemental Table 3). However, the results were significant for both all-cause mortality and the need for KRT regardless of whether baseline BMI was ≥25 kg/m2 (Supplemental Table 5). Statistical significance was attenuated for myocardial infarction and stroke in those with baseline BMI ≥25 kg/m2.
Diabetes had a significant interaction with the need for KRT (Supplemental Table 3). However, the association between high BMI variability and KRT risk remained significant regardless of the presence of diabetes (Supplemental Table 6). Some attenuation was identified for myocardial infarction or stroke risk, yet there were nonsignificant interactions with diabetes for outcomes according to the presence of diabetes.
Baseline eGFR <60 ml/min per 1.73 m2 showed significant interactions with the risk of KRT but not other adverse outcomes (Supplemental Table 3). However, the risk for necessity of KRT was higher in those with higher BMI variability in both subgroups, including those with reduced eGFR and with preserved eGFR (Supplemental Table 7). There was some attenuation of the significance for myocardial infarction and stroke in those with baseline eGFR ≥60 ml/min per 1.73 m2; however, the assessed number of subjects was small, with a low incidence rate of adverse outcomes in the subgroup.
Analysis of Trends in BMI
When the slope of the change in BMI was calculated, there were 38,157 (45%) individuals with an increasing BMI trend (≥0 regression beta), whereas 46,479 (55%) individuals had a decreasing trend in BMI. The analysis of the trends in BMI yielded significant results, indicating a possible interaction for all-cause mortality (Supplemental Table 3). When we assessed the prognostic significance in the subgroups stratified by trend in BMI, higher BMI variability was significantly associated with higher risks of all-cause mortality and KRT in the multivariable model (Table 3). Although statistical significance was not achieved for stroke in those with an increasing BMI trend or for myocardial infarction and stroke in those with a decreasing BMI trend, the interaction term analysis indicated there was an absence of a significant interaction for the outcomes according to the trend in BMI (Supplemental Table 3).
Table 3.
Risk of adverse outcomes according to BMI variability in patients with predialysis CKD stratified by trends in BMI
| Subgroup | Outcome | BMI Variability Independent of Mean Exposure | N | Event | Follow-up Person- years | Incidence rate (/1000 Person-years) | Univariable Model | Age, Sex, and Number of Exams Adjusted Model | Multivariable Modela | ||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HR (95% CI) | P for trend | Adjusted HR (95% CI) | P for trend | Adjusted HR (95% CI) | P for trend | ||||||||
| Increasing trend (regression β≥0) | All-cause mortality | Continuous | NA | 1.241 (1.189 to 1.295) | <0.001 | 1.217 (1.167 to 1.269) | <0.001 | 1.165 (1.118 to 1.215) | <0.001 | ||||
| Categorical | Q1 (low) | 9861 | 387 | 38,706 | 10.00 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | ||
| Q2 | 9974 | 364 | 39,515 | 9.21 | 0.917 (0.794 to 1.058) | 0.917 (0.794 to 1.058) | 0.992 (0.86 to 1.146) | ||||||
| Q3 | 9719 | 467 | 38,386 | 12.17 | 1.211 (1.058 to 1.385) | 1.211 (1.058 to 1.385) | 1.213 (1.06 to 1.389) | ||||||
| Q4 (high) | 8603 | 597 | 33,405 | 17.87 | 1.79 (1.575 to 2.034) | 1.79 (1.575 to 2.034) | 1.531 (1.345 to 1.742) | ||||||
| Myocardial infarction | Continuous | NA | 1.142 (1.071 to 1.217) | <0.001 | 1.132 (1.063 to 1.206) | <0.001 | 1.093 (1.025 to 1.164) | 0.006 | |||||
| Categorical | Q1 (low) | 9861 | 179 | 38,449 | 4.66 | Reference | <0.001 | Reference | 0.001 | Reference | 0.04 | ||
| Q2 | 9974 | 178 | 39,232 | 4.54 | 0.971 (0.789 to 1.195) | 1.01 (0.82 to 1.243) | 0.997 (0.81 to 1.228) | ||||||
| Q3 | 9719 | 213 | 38,045 | 5.60 | 1.198 (0.982 to 1.461) | 1.229 (1.008 to 1.5) | 1.186 (0.972 to 1.448) | ||||||
| Q4 (high) | 8603 | 221 | 33,096 | 6.68 | 1.436 (1.179 to 1.748) | 1.409 (1.156 to 1.718) | 1.265 (1.036 to 1.544) | ||||||
| Stroke | Continuous | NA | 1.114 (1.057 to 1.174) | <0.001 | 1.101 (1.045 to 1.161) | <0.001 | 1.066 (1.012 to 1.023) | 0.02 | |||||
| Categorical | Q1 (low) | 9861 | 273 | 38,290 | 7.13 | Reference | <0.001 | Reference | 0.002 | Reference | 0.10 | ||
| Q2 | 9974 | 265 | 39,045 | 6.79 | 0.951 (0.803 to 1.126) | 1.008 (0.851 to 1.194) | 1.005 (0.848 to 1.19) | ||||||
| Q3 | 9719 | 297 | 37,908 | 7.83 | 1.098 (0.932 to 1.294) | 1.139 (0.966 to 1.343) | 1.101 (0.934 to 1.298) | ||||||
| Q4 (high) | 8603 | 317 | 32,914 | 9.63 | 1.355 (1.152 to 1.593) | 1.32 (1.122 to 1.553) | 1.198 (1.017 to 1.411) | ||||||
| KRT | Continuous | NA | 1.142 (1.087 to 1.199) | <0.001 | 1.181 (1.124 to 1.240) | <0.001 | 1.049 (0.998 to 1.104) | 0.06 | |||||
| Categorical | Q1 (low) | 9861 | 290 | 38,324 | 7.57 | Reference | <0.001 | Reference | <0.001 | Reference | 0.01 | ||
| Q2 | 9974 | 343 | 39,008 | 8.79 | 1.159 (0.992 to 1.356) | 1.132 (0.968 to 1.325) | 0.943 (0.806 to 1.105) | ||||||
| Q3 | 9719 | 327 | 37,857 | 8.64 | 1.14 (0.973 to 1.335) | 1.161 (0.991 to 1.36) | 0.931 (0.794 to 1.092) | ||||||
| Q4 (high) | 8603 | 383 | 32,803 | 11.68 | 1.551 (1.332 to 1.807) | 1.702 (1.461 to 1.984) | 1.160 (0.993 to 1.355) | ||||||
| Decreasing trend (regression β <0) | All-cause mortality | Continuous | NA | 1.438 (1.390 to 1.488) | <0.001 | 1.341 (1.296 to 1.387) | <0.001 | 1.213 (1.171 to 1.256) | <0.001 | ||||
| Categorical | Q1 (low) | 11,298 | 460 | 44,233 | 10.40 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | ||
| Q2 | 11,185 | 514 | 43,985 | 11.69 | 1.12 (0.988 to 1.271) | 1.12 (0.988 to 1.271) | 1.069 (0.942 to 1.212) | ||||||
| Q3 | 11,440 | 638 | 44,798 | 14.24 | 1.367 (1.213 to 1.541) | 1.367 (1.213 to 1.541) | 1.147 (1.016 to 1.294) | ||||||
| Q4 (high) | 12,556 | 1355 | 48,155 | 28.14 | 2.713 (2.441 to 3.016) | 2.713 (2.441 to 3.016) | 1.719 (1.538 to 1.92) | ||||||
| Myocardial infarction | Continuous | NA | 1.135 (1.075 to 1.198) | <0.001 | 1.099 (1.040 to 1.160) | <0.001 | 1.062 (1.003 to 1.123) | 0.04 | |||||
| Categorical | Q1 (low) | 11,298 | 231 | 43,912 | 5.26 | Reference | <0.001 | Reference | 0.001 | Reference | 0.05 | ||
| Q2 | 11,185 | 210 | 43,660 | 4.81 | 0.911 (0.756 to 1.098) | 0.918 (0.761 to 1.107) | 0.902 (0.748 to 1.088) | ||||||
| Q3 | 11,440 | 257 | 44,416 | 5.79 | 1.097 (0.918 to 1.31) | 1.078 (0.902 to 1.288) | 1.032 (0.863 to 1.234) | ||||||
| Q4 (high) | 12,556 | 350 | 47,687 | 7.34 | 1.395 (1.182 to 1.647) | 1.275 (1.078 to 1.508) | 1.153 (0.969 to 1.372) | ||||||
| Stroke | Continuous | NA | 1.136 (1.086 to 1.189) | <0.001 | 1.081 (1.033 to 1.131) | <0.001 | 1.047 (0.999 to 1.097) | 0.06 | |||||
| Categorical | Q1 (low) | 11,298 | 311 | 43,718 | 7.11 | Reference | <0.001 | Reference | 0.003 | Reference | 0.15 | ||
| Q2 | 11,185 | 345 | 43,414 | 7.95 | 1.116 (0.958 to 1.301) | 1.123 (0.963 to 1.309) | 1.102 (0.945 to 1.285) | ||||||
| Q3 | 11,440 | 355 | 44,199 | 8.03 | 1.129 (0.969 to 1.314) | 1.096 (0.941 to 1.276) | 1.052 (0.902 to 1.227) | ||||||
| Q4 (high) | 12,556 | 503 | 47,392 | 10.61 | 1.493 (1.296 to 1.719) | 1.295 (1.122 to 1.494) | 1.178 (1.015 to 1.365) | ||||||
| KRT | Continuous | NA | 1.206 (1.158 to 1.255) | <0.001 | 1.280 (1.228 to 1.334) | <0.001 | 1.071 (1.027 to 1.117) | 0.001 | |||||
| Categorical | Q1 (low) | 11,298 | 356 | 43,730 | 8.14 | Reference | <0.001 | Reference | <0.001 | Reference | 0.009 | ||
| Q2 | 11,185 | 411 | 43,387 | 9.47 | 1.162 (1.008 to 1.339) | 1.14 (0.989 to 1.314) | 1.006 (0.872 to 1.16) | ||||||
| Q3 | 11,440 | 505 | 44,039 | 11.47 | 1.41 (1.231 to 1.615) | 1.448 (1.265 to 1.659) | 1.157 (1.009 to 1.326) | ||||||
| Q4 (high) | 12,556 | 661 | 47,154 | 14.02 | 1.735 (1.526 to 1.974) | 2.042 (1.794 to 2.325) | 1.196 (1.046 to 1.367) | ||||||
Multivariable model was adjusted for age, sex, number of exams, current smoking, alcohol consumption, regular physical activity, low-income status, history of diabetes mellitus, hypertension, dyslipidemia, cancer, chronic lung disease, baseline BMI, waist circumference, fasting glucose, systolic BP, diastolic BP, HDL, baseline eGFR, presence of dipstick albuminuria, and eGFR variability during the exposure assessment period.
Prognosis According to Metabolic Parameter Variability
When variabilities of other metabolic parameters were considered (Table 4), the metabolic parameters with higher variability that were significantly associated with a higher risk of all-cause mortality in the fully adjusted model were waist circumference, fasting blood glucose, systolic BP, diastolic BP, total cholesterol, LDL cholesterol, and HDL cholesterol. For the risk of myocardial infarction, fasting blood glucose, and high-density lipoprotein cholesterol variabilities, with both categorical and continuous exposures, were significant. For the risk of stroke, variabilities in fasting blood glucose, diastolic BP, and high-density lipoprotein were prominently associated with the outcomes in the multivariable model. Fasting blood glucose and total cholesterol variabilities were significantly associated with the risk of the need for KRT in the fully adjusted model.
Table 4.
Prognostic significance of variabilities in various metabolic parameters in patients with predialysis CKD
| Metabolic parameter | Outcome | Variability Independent of Mean Exposure | N | Event | Follow-up Person- years | Incidence rate (/1000 Person- years) | Univariable Model | Age, Sex, and Number of Exams Adjusted Model | Multivariable Modela | ||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HR (95% CI) | P | Adjusted HR (95% CI) | P | Adjusted HR (95% CI) | P | ||||||||
| Waist circumference | All-cause mortality | Continuous | NA | 1.141 (1.112 to 1.171) | <0.001 | 1.125 (1.096 to 1.154) | <0.001 | 1.097 (1.069 to 1.126) | <0.0001 | ||||
| Categorical | Q1 (low) | 21,176 | 1026 | 82,940 | 12.37 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | ||
| Q2 | 21,121 | 1087 | 82,976 | 13.10 | 1.057 (0.971 to 1.151) | 1.066 (0.979 to 1.161) | 1.044 (0.958 to 1.137) | ||||||
| Q3 | 21,190 | 1158 | 83,178 | 13.92 | 1.123 (1.033 to 1.222) | 1.142 (1.05 to 1.243) | 1.103 (1.014 to 1.2) | ||||||
| Q4 (high) | 21,149 | 1511 | 82,088 | 18.41 | 1.49 (1.377 to 1.613) | 1.427 (1.317 to 1.546) | 1.321 (1.219 to 1.431) | ||||||
| Myocardial infarction | Continuous | NA | 1.035 (0.994 to 1.078) | 0.10 | 1.024 (0.983 to 1.067) | 0.26 | 1.016 (0.974 to 1.058) | 0.46 | |||||
| Categorical | Q1 (low) | 21,176 | 455 | 82,316 | 5.53 | Reference | 0.003 | Reference | 0.02 | Reference | 0.04 | ||
| Q2 | 21,121 | 455 | 82,295 | 5.53 | 0.998 (0.877 to 1.137) | 1.003 (0.881 to 1.142) | 1.002 (0.88 to 1.142) | ||||||
| Q3 | 21,190 | 412 | 82,555 | 4.99 | 0.901 (0.788 to 1.029) | 0.905 (0.792 to 1.034) | 0.895 (0.783 to 1.023) | ||||||
| Q4 (high) | 21,149 | 517 | 81,330 | 6.36 | 1.15 (1.014 to 1.305) | 1.112 (0.979 to 1.263) | 1.086 (0.956 to 1.234) | ||||||
| Stroke | Continuous | NA | 1.054 (1.019 to 1.09) | 0.003 | 1.034 (0.999 to 1.07) | 0.05 | 1.027 (0.992 to 1.062) | 0.13 | |||||
| Categorical | Q1 (low) | 21,176 | 650 | 81,948 | 7.93 | Reference | 0.002 | Reference | 0.08 | Reference | 0.19 | ||
| Q2 | 21,121 | 609 | 81,958 | 7.43 | 0.936 (0.838 to 1.046) | 0.942 (0.843 to 1.052) | 0.942 (0.843 to 1.052) | ||||||
| Q3 | 21,190 | 670 | 82,084 | 8.16 | 1.028 (0.923 to 1.146) | 1.029 (0.924 to 1.147) | 1.024 (0.919 to 1.141) | ||||||
| Q4 (high) | 21,149 | 737 | 80,890 | 9.11 | 1.149 (1.034 to 1.277) | 1.084 (0.974 to 1.206) | 1.061 (0.953 to 1.18) | ||||||
| KRT | Continuous | NA | 1.002 (0.972 to 1.033) | 0.89 | 1.062 (1.029 to 1.096) | <0.001 | 1.007 (0.976 to 1.04) | 0.64 | |||||
| Categorical | Q1 (low) | 21,176 | 818 | 81,730 | 10.01 | Reference | 0.999 | Reference | <0.001 | Reference | 0.26 | ||
| Q2 | 21,121 | 820 | 81,739 | 10.03 | 1.001 (0.909 to 1.103) | 0.996 (0.904 to 1.098) | 0.919 (0.834 to 1.013) | ||||||
| Q3 | 21,190 | 826 | 81,966 | 10.08 | 1.006 (0.913 to 1.108) | 1.058 (0.961 to 1.166) | 0.981 (0.89 to 1.081) | ||||||
| Q4 (high) | 21,149 | 812 | 80,866 | 10.04 | 1.006 (0.913 to 1.108) | 1.198 (1.086 to 1.321) | 1.003 (0.909 to 1.106) | ||||||
| Fasting blood glucose | All-cause mortality | Continuous | NA | 1.201 (1.17 to 1.232) | <0.001 | 1.2 (1.17 to 1.232) | <0.001 | 1.113 (1.084 to 1.143) | <0.001 | ||||
| Categorical | Q1 (low) | 21,160 | 974 | 82,496 | 11.81 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | ||
| Q2 | 21,153 | 1006 | 82,889 | 12.14 | 1.025 (0.939 to 1.12) | 1.059 (0.97 to 1.157) | 1.016 (0.93 to 1.11) | ||||||
| Q3 | 21,164 | 1144 | 83,084 | 13.77 | 1.161 (1.066 to 1.265) | 1.197 (1.099 to 1.304) | 1.079 (0.99 to 1.176) | ||||||
| Q4 (high) | 21,159 | 1658 | 82,713 | 20.05 | 1.691 (1.562 to 1.83) | 1.701 (1.572 to 1.842) | 1.368 (1.26 to 1.485) | ||||||
| Myocardial infarction | Continuous | NA | 1.208 (1.159 to 1.259) | <0.001 | 1.206 (1.157 to 1.257) | <0.001 | 1.109 (1.062 to 1.157) | <0.001 | |||||
| Categorical | Q1 (low) | 21,160 | 377 | 81,923 | 4.60 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | ||
| Q2 | 21,153 | 372 | 82,344 | 4.52 | 0.979 (0.849 to 1.13) | 1.001 (0.867 to 1.155) | 0.965 (0.836 to 1.113) | ||||||
| Q3 | 21,164 | 452 | 82,423 | 5.48 | 1.187 (1.036 to 1.361) | 1.206 (1.052 to 1.383) | 1.081 (0.942 to 1.24) | ||||||
| Q4 (high) | 21,159 | 638 | 81,807 | 7.80 | 1.688 (1.487 to 1.918) | 1.686 (1.484 to 1.916) | 1.328 (1.163 to 1.516) | ||||||
| Stroke | Continuous | NA | 1.193 (1.152 to 1.235) | <0.001 | 1.193 (1.153 to 1.235) | <0.001 | 1.106 (1.067 to 1.146) | <0.001 | |||||
| Categorical | Q1 (low) | 21,160 | 528 | 81,647 | 6.47 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | ||
| Q2 | 21,153 | 586 | 81,931 | 7.15 | 1.105 (0.983 to 1.243) | 1.138 (1.012 to 1.28) | 1.097 (0.975 to 1.234) | ||||||
| Q3 | 21,164 | 666 | 82,043 | 8.12 | 1.254 (1.119 to 1.406) | 1.286 (1.147 to 1.442) | 1.162 (1.036 to 1.303) | ||||||
| Q4 (high) | 21,159 | 886 | 81,258 | 10.90 | 1.686 (1.514 to 1.878) | 1.699 (1.525 to 1.893) | 1.362 (1.218 to 1.523) | ||||||
| KRT | Continuous | NA | 1.351 (1.309 to 1.395) | <0.001 | 1.333 (1.291 to 1.376) | <0.001 | 1.059 (1.024 to 1.095) | 0.001 | |||||
| Categorical | Q1 (low) | 21,160 | 543 | 81,719 | 6.64 | Reference | <0.001 | Reference | <0.001 | Reference | 0.008 | ||
| Q2 | 21,153 | 639 | 81,950 | 7.80 | 1.172 (1.045 to 1.314) | 1.122 (1 to 1.258) | 1.034 (0.921 to 1.159) | ||||||
| Q3 | 21,164 | 825 | 81,801 | 10.09 | 1.516 (1.36 to 1.689) | 1.442 (1.293 to 1.607) | 1.098 (0.984 to 1.226) | ||||||
| Q4 (high) | 21,159 | 1269 | 80,832 | 15.70 | 2.366 (2.14 to 2.616) | 2.247 (2.032 to 2.485) | 1.179 (1.06 to 1.311) | ||||||
| Systolic BP | All-cause mortality | Continuous | NA | 1.155 (1.126 to 1.185) | <0.001 | 1.12 (1.092 to 1.149) | <0.001 | 1.075 (1.048 to 1.102) | <0.001 | ||||
| Categorical | Q1 (low) | 21,247 | 1029 | 82,886 | 12.41 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | ||
| Q2 | 21,070 | 1021 | 82,880 | 12.32 | 0.988 (0.906 to 1.078) | 1.019 (0.934 to 1.111) | 1.004 (0.921 to 1.095) | ||||||
| Q3 | 21,160 | 1193 | 82,917 | 14.39 | 1.156 (1.063 to 1.256) | 1.14 (1.048 to 1.239) | 1.082 (0.995 to 1.176) | ||||||
| Q4 (high) | 21,159 | 1539 | 82,500 | 18.65 | 1.499 (1.385 to 1.622) | 1.384 (1.279 to 1.498) | 1.228 (1.134 to 1.329) | ||||||
| Myocardial infarction | Continuous | NA | 1.079 (1.036 to 1.125) | <0.001 | 1.062 (1.019 to 1.106) | 0.004 | 1.035 (0.993 to 1.078) | 0.10 | |||||
| Categorical | Q1 (low) | 21,247 | 426 | 82,262 | 5.18 | Reference | 0.001 | Reference | 0.03 | Reference | 0.38 | ||
| Q2 | 21,070 | 421 | 82,283 | 5.12 | 0.985 (0.86 to 1.126) | 1.005 (0.878 to 1.15) | 0.995 (0.87 to 1.139) | ||||||
| Q3 | 21,160 | 466 | 82,233 | 5.67 | 1.091 (0.957 to 1.244) | 1.085 (0.951 to 1.238) | 1.05 (0.921 to 1.198) | ||||||
| Q4 (high) | 21,159 | 526 | 81,718 | 6.44 | 1.239 (1.091 to 1.408) | 1.185 (1.043 to 1.347) | 1.099 (0.966 to 1.25) | ||||||
| Stroke | Continuous | NA | 1.084 (1.048 to 1.122) | <0.001 | 1.061 (1.026 to 1.098) | <0.001 | 1.038 (1.003 to 1.074) | 0.03 | |||||
| Categorical | Q1 (low) | 21,247 | 616 | 81,928 | 7.52 | Reference | <0.001 | Reference | <0.001 | Reference | 0.13 | ||
| Q2 | 21,070 | 618 | 81,876 | 7.55 | 1.003 (0.897 to 1.121) | 1.033 (0.923 to 1.155) | 1.027 (0.918 to 1.148) | ||||||
| Q3 | 21,160 | 655 | 81,853 | 8.00 | 1.063 (0.953 to 1.187) | 1.06 (0.95 to 1.184) | 1.034 (0.926 to 1.154) | ||||||
| Q4 (high) | 21,159 | 777 | 81,223 | 9.57 | 1.272 (1.144 to 1.414) | 1.203 (1.082 to 1.337) | 1.126 (1.012 to 1.252) | ||||||
| KRT | Continuous | NA | 1.128 (1.094 to 1.164) | <0.001 | 1.156 (1.12 to 1.192) | <0.001 | 1.028 (0.997 to 1.061) | 0.08 | |||||
| Categorical | Q1 (low) | 21,247 | 676 | 81,868 | 8.26 | Reference | <0.001 | Reference | <0.001 | Reference | 0.34 | ||
| Q2 | 21,070 | 778 | 81,729 | 9.52 | 1.15 (1.037 to 1.275) | 1.11 (1.001 to 1.23) | 0.999 (0.9 to 1.108) | ||||||
| Q3 | 21,160 | 853 | 81,641 | 10.45 | 1.264 (1.143 to 1.398) | 1.256 (1.135 to 1.389) | 1.048 (0.947 to 1.161) | ||||||
| Q4 (high) | 21,159 | 969 | 81,064 | 11.95 | 1.449 (1.313 to 1.598) | 1.542 (1.397 to 1.701) | 1.077 (0.975 to 1.188) | ||||||
| Diastolic BP | All-cause mortality | Continuous | NA | 1.139 (1.111 to 1.169) | <0.001 | 1.096 (1.068 to 1.124) | <0.001 | 1.059 (1.032 to 1.086) | <0.001 | ||||
| Categorical | Q1 (low) | 21,161 | 1008 | 82,778 | 12.18 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | ||
| Q2 | 21,157 | 1145 | 82,989 | 13.80 | 1.13 (1.038 to 1.23) | 1.068 (0.982 to 1.163) | 1.035 (0.951 to 1.126) | ||||||
| Q3 | 21,025 | 1085 | 82,602 | 13.14 | 1.076 (0.987 to 1.172) | 1.068 (0.98 to 1.163) | 1.043 (0.957 to 1.136) | ||||||
| Q4 (high) | 21,293 | 1544 | 82,813 | 18.64 | 1.531 (1.414 to 1.658) | 1.332 (1.23 to 1.442) | 1.193 (1.101 to 1.292) | ||||||
| Myocardial infarction | Continuous | NA | 1.056 (1.013 to 1.1) | 0.01 | 1.036 (0.995 to 1.079) | 0.09 | 1.012 (0.971 to 1.054) | 0.58 | |||||
| Categorical | Q1 (low) | 21,161 | 439 | 82,144 | 5.34 | Reference | <0.001 | Reference | 0.03 | Reference | 0.23 | ||
| Q2 | 21,157 | 452 | 82,347 | 5.49 | 1.024 (0.898 to 1.168) | 0.992 (0.87 to 1.131) | 0.969 (0.849 to 1.105) | ||||||
| Q3 | 21,025 | 415 | 82,040 | 5.06 | 0.944 (0.825 to 1.079) | 0.941 (0.823 to 1.077) | 0.918 (0.803 to 1.05) | ||||||
| Q4 (high) | 21,293 | 533 | 81,965 | 6.50 | 1.216 (1.071 to 1.379) | 1.133 (0.998 to 1.286) | 1.049 (0.923 to 1.191) | ||||||
| Stroke | Continuous | NA | 1.115 (1.077 to 1.154) | <0.001 | 1.088 (1.052 to 1.126) | <0.001 | 1.066 (1.031 to 1.103) | <0.001 | |||||
| Categorical | Q1 (low) | 21,161 | 554 | 81,875 | 6.77 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | ||
| Q2 | 21,157 | 685 | 81,891 | 8.36 | 1.236 (1.105 to 1.382) | 1.191 (1.065 to 1.332) | 1.161 (1.038 to 1.299) | ||||||
| Q3 | 21,025 | 612 | 81,595 | 7.50 | 1.108 (0.987 to 1.243) | 1.11 (0.99 to 1.246) | 1.089 (0.971 to 1.222) | ||||||
| Q4 (high) | 21,293 | 815 | 81,519 | 10.00 | 1.479 (1.327 to 1.647) | 1.353 (1.215 to 1.508) | 1.265 (1.135 to 1.41) | ||||||
| KRT | Continuous | NA | 1.126 (1.092 to 1.161) | <0.001 | 1.138 (1.103 to 1.173) | <0.001 | 1.012 (0.981 to 1.043) | 0.47 | |||||
| Categorical | Q1 (low) | 21,161 | 710 | 81,749 | 8.69 | Reference | <0.001 | Reference | <0.001 | Reference | 0.28 | ||
| Q2 | 21,157 | 723 | 81,939 | 8.82 | 1.015 (0.915 to 1.125) | 1.027 (0.926 to 1.139) | 0.934 (0.842 to 1.037) | ||||||
| Q3 | 21,025 | 867 | 81,243 | 10.67 | 1.228 (1.112 to 1.356) | 1.187 (1.075 to 1.311) | 1.028 (0.93 to 1.136) | ||||||
| Q4 (high) | 21,293 | 976 | 81,371 | 11.99 | 1.385 (1.257 to 1.525) | 1.448 (1.314 to 1.595) | 1.005 (0.912 to 1.108) | ||||||
| Total cholesterol | All-cause mortality | Continuous | NA | 1.08 (1.053 to 1.108) | <0.001 | 1.092 (1.064 to 1.12) | <0.001 | 1.037 (1.01 to 1.065) | 0.007 | ||||
| Categorical | Q1 (low) | 21,159 | 1053 | 82,816 | 12.71 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | ||
| Q2 | 21,158 | 1188 | 83,021 | 14.31 | 1.123 (1.034 to 1.221) | 1.154 (1.062 to 1.253) | 1.105 (1.017 to 1.201) | ||||||
| Q3 | 21,160 | 1193 | 83,096 | 14.36 | 1.126 (1.037 to 1.224) | 1.15 (1.059 to 1.25) | 1.046 (0.962 to 1.137) | ||||||
| Q4 (high) | 21,159 | 1348 | 82,249 | 16.39 | 1.291 (1.191 to 1.4) | 1.34 (1.236 to 1.453) | 1.151 (1.058 to 1.252) | ||||||
| Myocardial infarction | Continuous | NA | 1.102 (1.057 to 1.148) | <0.001 | 1.101 (1.057 to 1.147) | <0.001 | 1.034 (0.99 to 1.079) | 0.13 | |||||
| Categorical | Q1 (low) | 21,159 | 387 | 82,255 | 4.70 | Reference | <0.001 | Reference | <0.001 | Reference | 0.18 | ||
| Q2 | 21,158 | 445 | 82,379 | 5.40 | 1.146 (1 to 1.313) | 1.162 (1.014 to 1.332) | 1.119 (0.977 to 1.283) | ||||||
| Q3 | 21,160 | 493 | 82,373 | 5.98 | 1.269 (1.111 to 1.45) | 1.28 (1.12 to 1.462) | 1.159 (1.014 to 1.325) | ||||||
| Q4 (high) | 21,159 | 514 | 81,489 | 6.31 | 1.343 (1.177 to 1.532) | 1.342 (1.176 to 1.531) | 1.111 (0.968 to 1.274) | ||||||
| Stroke | Continuous | NA | 1.049 (1.014 to 1.086) | 0.006 | 1.049 (1.014 to 1.086) | 0.005 | 0.999 (0.965 to 1.035) | 0.97 | |||||
| Categorical | Q1 (low) | 21,159 | 615 | 81,797 | 7.52 | Reference | 0.05 | Reference | 0.04 | Reference | 0.67 | ||
| Q2 | 21,158 | 660 | 81,906 | 8.06 | 1.071 (0.96 to 1.196) | 1.096 (0.982 to 1.223) | 1.065 (0.954 to 1.189) | ||||||
| Q3 | 21,160 | 682 | 82,037 | 8.31 | 1.105 (0.991 to 1.232) | 1.12 (1.004 to 1.249) | 1.034 (0.926 to 1.153) | ||||||
| Q4 (high) | 21,159 | 709 | 81,140 | 8.74 | 1.163 (1.044 to 1.296) | 1.167 (1.048 to 1.301) | 1.009 (0.901 to 1.129) | ||||||
| KRT | Continuous | NA | 1.371 (1.328 to 1.416) | <0.001 | 1.403 (1.359 to 1.449) | <0.001 | 1.068 (1.033 to 1.104) | <0.001 | |||||
| Categorical | Q1 (low) | 21,159 | 491 | 82,120 | 5.98 | Reference | <0.001 | Reference | <0.001 | Reference | 0.001 | ||
| Q2 | 21,158 | 646 | 82,085 | 7.87 | 1.315 (1.17 to 1.479) | 1.266 (1.125 to 1.423) | 1.083 (0.963 to 1.219) | ||||||
| Q3 | 21,160 | 929 | 81,684 | 11.37 | 1.904 (1.707 to 2.124) | 1.889 (1.694 to 2.108) | 1.194 (1.069 to 1.334) | ||||||
| Q4 (high) | 21,159 | 1210 | 80,413 | 15.05 | 2.532 (2.28 to 2.812) | 2.654 (2.39 to 2.948) | 1.218 (1.092 to 1.359) | ||||||
| LDL cholesterol | All-cause mortality | Continuous | NA | 1.106 (1.078 to 1.134) | <0.001 | 1.091 (1.063 to 1.119) | <0.001 | 1.029 (1.002 to 1.058) | 0.04 | ||||
| Categorical | Q1 (low) | 21,159 | 1055 | 83,059 | 12.70 | Reference | <0.001 | Reference | <0.001 | Reference | 0.03 | ||
| Q2 | 21,159 | 1111 | 83,101 | 13.37 | 1.051 (0.966 to 1.144) | 1.037 (0.954 to 1.129) | 0.964 (0.886 to 1.049) | ||||||
| Q3 | 21,159 | 1214 | 82,858 | 14.65 | 1.154 (1.062 to 1.253) | 1.123 (1.034 to 1.22) | 0.994 (0.913 to 1.081) | ||||||
| Q4 (high) | 21,159 | 1402 | 82,164 | 17.06 | 1.348 (1.245 to 1.46) | 1.292 (1.193 to 1.4) | 1.083 (0.995 to 1.178) | ||||||
| Myocardial infarction | Continuous | NA | 1.05 (1.008 to 1.094) | 0.02 | 1.037 (0.995 to 1.08) | 0.08 | 0.947 (0.907 to 0.99) | 0.02 | |||||
| Categorical | Q1 (low) | 21,159 | 446 | 82,380 | 5.41 | Reference | 0.047 | Reference | 0.16 | Reference | 0.07 | ||
| Q2 | 21,159 | 431 | 82,491 | 5.22 | 0.964 (0.844 to 1.1) | 0.956 (0.838 to 1.092) | 0.89 (0.779 to 1.017) | ||||||
| Q3 | 21,159 | 458 | 82,164 | 5.57 | 1.029 (0.903 to 1.172) | 1.009 (0.885 to 1.149) | 0.866 (0.758 to 0.99) | ||||||
| Q4 (high) | 21,159 | 504 | 81,462 | 6.19 | 1.146 (1.009 to 1.301) | 1.104 (0.972 to 1.255) | 0.842 (0.735 to 0.964) | ||||||
| Stroke | Continuous | NA | 1.058 (1.023 to 1.095) | 0.001 | 1.047 (1.011 to 1.083) | 0.009 | 0.976 (0.941 to 1.012) | 0.19 | |||||
| Categorical | Q1 (low) | 21,159 | 612 | 82,063 | 7.46 | Reference | 0.01 | Reference | 0.07 | Reference | 0.50 | ||
| Q2 | 21,159 | 651 | 82,006 | 7.94 | 1.064 (0.953 to 1.188) | 1.06 (0.949 to 1.184) | 0.999 (0.895 to 1.116) | ||||||
| Q3 | 21,159 | 687 | 81,738 | 8.40 | 1.127 (1.011 to 1.257) | 1.108 (0.993 to 1.235) | 0.985 (0.881 to 1.101) | ||||||
| Q4 (high) | 21,159 | 716 | 81,072 | 8.83 | 1.186 (1.065 to 1.321) | 1.148 (1.03 to 1.279) | 0.928 (0.828 to 1.04) | ||||||
| KRT | Continuous | NA | 1.325 (1.284 to 1.368) | <0.001 | 1.319 (1.278 to 1.362) | <0.001 | 1.001 (0.967 to 1.035) | 0.98 | |||||
| Categorical | Q1 (low) | 21,159 | 469 | 82,361 | 5.69 | Reference | <0.001 | Reference | <0.001 | Reference | 0.06 | ||
| Q2 | 21,159 | 739 | 82,028 | 9.01 | 1.583 (1.41 to 1.777) | 1.556 (1.386 to 1.747) | 1.152 (1.026 to 1.294) | ||||||
| Q3 | 21,159 | 941 | 81,434 | 11.56 | 2.035 (1.822 to 2.274) | 2.039 (1.825 to 2.278) | 1.112 (0.993 to 1.245) | ||||||
| Q4 (high) | 21,159 | 1127 | 80,479 | 14.00 | 2.476 (2.223 to 2.758) | 2.426 (2.178 to 2.702) | 1.053 (0.94 to 1.178) | ||||||
| HDL cholesterol | All-cause mortality | Continuous | NA | 1.242 (1.21 to 1.274) | <0.001 | 1.142 (1.113 to 1.172) | <0.001 | 1.085 (1.054 to 1.116) | <0.001 | ||||
| Categorical | Q1 (low) | 21,159 | 856 | 82,673 | 10.35 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | ||
| Q2 | 21,157 | 1062 | 83,126 | 12.78 | 1.229 (1.123 to 1.345) | 1.169 (1.069 to 1.28) | 1.137 (1.039 to 1.245) | ||||||
| Q3 | 21,164 | 1208 | 82,957 | 14.56 | 1.401 (1.284 to 1.53) | 1.239 (1.135 to 1.353) | 1.16 (1.06 to 1.269) | ||||||
| Q4 (high) | 21,156 | 1656 | 82,426 | 20.09 | 1.934 (1.781 to 2.101) | 1.516 (1.396 to 1.647) | 1.303 (1.191 to 1.426) | ||||||
| Myocardial infarction | Continuous | NA | 1.242 (1.191 to 1.295) | <0.001 | 1.185 (1.137 to 1.236) | <0.001 | 1.093 (1.044 to 1.145) | <0.001 | |||||
| Categorical | Q1 (low) | 21,159 | 337 | 82,168 | 4.10 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 | ||
| Q2 | 21,157 | 410 | 82,526 | 4.97 | 1.207 (1.045 to 1.395) | 1.171 (1.013 to 1.352) | 1.102 (0.953 to 1.274) | ||||||
| Q3 | 21,164 | 441 | 82,293 | 5.36 | 1.302 (1.13 to 1.501) | 1.212 (1.051 to 1.397) | 1.067 (0.922 to 1.234) | ||||||
| Q4 (high) | 21,156 | 651 | 81,510 | 7.99 | 1.942 (1.702 to 2.215) | 1.693 (1.483 to 1.932) | 1.336 (1.156 to 1.544) | ||||||
| Stroke | Continuous | NA | 1.183 (1.143 to 1.225) | <0.001 | 1.123 (1.085 to 1.163) | 0.04 | 1.044 (1.005 to 1.085) | 0.03 | |||||
| Categorical | Q1 (low) | 21,159 | 517 | 81,874 | 6.31 | Reference | <0.001 | Reference | 0.03 | Reference | 0.02 | ||
| Q2 | 21,157 | 627 | 82,114 | 7.64 | 1.208 (1.075 to 1.357) | 1.172 (1.043 to 1.317) | 1.107 (0.985 to 1.245) | ||||||
| Q3 | 21,164 | 650 | 81,871 | 7.94 | 1.256 (1.119 to 1.41) | 1.164 (1.037 to 1.307) | 1.034 (0.918 to 1.164) | ||||||
| Q4 (high) | 21,156 | 872 | 81,021 | 10.76 | 1.704 (1.529 to 1.9) | 1.461 (1.31 to 1.63) | 1.178 (1.045 to 1.328) | ||||||
| KRT | Continuous | NA | 1.272 (1.233 to 1.312) | <0.001 | 1.268 (1.228 to 1.309) | <0.001 | 1.03 (0.995 to 1.067) | 0.10 | |||||
| Categorical | Q1 (low) | 21,159 | 580 | 81,821 | 7.09 | Reference | <0.001 | Reference | <0.001 | Reference | 0.41 | ||
| Q2 | 21,157 | 685 | 82,134 | 8.34 | 1.174 (1.051 to 1.311) | 1.133 (1.014 to 1.265) | 1.039 (0.929 to 1.162) | ||||||
| Q3 | 21,164 | 861 | 81,663 | 10.54 | 1.487 (1.339 to 1.652) | 1.429 (1.286 to 1.589) | 1.055 (0.946 to 1.177) | ||||||
| Q4 (high) | 21,156 | 1150 | 80,683 | 14.25 | 2.017 (1.825 to 2.228) | 1.979 (1.789 to 2.189) | 1.098 (0.983 to 1.227) | ||||||
| Triglycerides | All-cause mortality | Continuous | NA | 0.985 (0.961 to 1.011) | 0.25 | 1.038 (1.012 to 1.065) | 0.004 | 1.018 (0.992 to 1.044) | 0.17 | ||||
| Categorical | Q1 (low) | 21,159 | 1203 | 82,338 | 14.61 | Reference | 0.66 | Reference | 0.03 | Reference | 0.47 | ||
| Q2 | 21,159 | 1212 | 82,744 | 14.65 | 1 (0.923 to 1.083) | 1.068 (0.986 to 1.157) | 1.043 (0.963 to 1.13) | ||||||
| Q3 | 21,159 | 1201 | 83,012 | 14.47 | 0.986 (0.91 to 1.068) | 1.089 (1.006 to 1.18) | 1.061 (0.979 to 1.149) | ||||||
| Q4 (high) | 21,159 | 1166 | 83,088 | 14.03 | 0.956 (0.882 to 1.036) | 1.124 (1.037 to 1.219) | 1.055 (0.973 to 1.144) | ||||||
| Myocardial infarction | Continuous | NA | 1.004 (0.964 to 1.046) | 0.84 | 1.03 (0.989 to 1.073) | 0.16 | 1.021 (0.98 to 1.064) | 0.33 | |||||
| Categorical | Q1 (low) | 21,159 | 469 | 81,639 | 5.74 | Reference | 0.42 | Reference | 0.24 | Reference | 0.37 | ||
| Q2 | 21,159 | 429 | 82,124 | 5.22 | 0.907 (0.795 to 1.033) | 0.94 (0.824 to 1.071) | 0.933 (0.819 to 1.064) | ||||||
| Q3 | 21,159 | 473 | 82,345 | 5.74 | 0.995 (0.876 to 1.131) | 1.048 (0.922 to 1.191) | 1.034 (0.91 to 1.176) | ||||||
| Q4 (high) | 21,159 | 468 | 82,388 | 5.68 | 0.984 (0.866 to 1.118) | 1.065 (0.937 to 1.211) | 1.036 (0.911 to 1.178) | ||||||
| Stroke | Continuous | NA | 0.984 (0.951 to 1.017) | 0.34 | 1.016 (0.982 to 1.051) | 0.37 | 1.006 (0.973 to 1.041) | 0.73 | |||||
| Categorical | Q1 (low) | 21,159 | 691 | 81,225 | 8.51 | Reference | 0.32 | Reference | 0.32 | Reference | 0.49 | ||
| Q2 | 21,159 | 669 | 81,676 | 8.19 | 0.962 (0.865 to 1.07) | 1.009 (0.907 to 1.122) | 0.999 (0.898 to 1.111) | ||||||
| Q3 | 21,159 | 631 | 81,994 | 7.70 | 0.903 (0.811 to 1.006) | 0.966 (0.867 to 1.076) | 0.952 (0.854 to 1.061) | ||||||
| Q4 (high) | 21,159 | 675 | 81,984 | 8.23 | 0.966 (0.869 to 1.074) | 1.069 (0.961 to 1.189) | 1.038 (0.933 to 1.155) | ||||||
| KRT | Continuous | NA | 1.059 (1.027 to 1.092) | <0.001 | 1.041 (1.009 to 1.074) | 0.01 | 0.996 (0.965 to 1.027) | 0.79 | |||||
| Categorical | Q1 (low) | 21,159 | 731 | 81,280 | 8.99 | Reference | 0.001 | Reference | 0.04 | Reference | 0.22 | ||
| Q2 | 21,159 | 798 | 81,562 | 9.78 | 1.086 (0.982 to 1.201) | 1.035 (0.936 to 1.144) | 1.051 (0.95 to 1.163) | ||||||
| Q3 | 21,159 | 880 | 81,710 | 10.77 | 1.194 (1.083 to 1.318) | 1.13 (1.024 to 1.247) | 1.076 (0.975 to 1.188) | ||||||
| Q4 (high) | 21,159 | 867 | 81,750 | 10.61 | 1.176 (1.066 to 1.298) | 1.112 (1.008 to 1.227) | 0.983 (0.89 to 1.086) | ||||||
Multivariable model was adjusted for age, sex, number of exams, current smoking, drinking alcohol, regular physical activity, low-income state, history of diabetes mellitus, hypertension, dyslipidemia, cancer, chronic lung disease, baseline BMI, waist circumference, fasting glucose, systolic BP, diastolic BP, HDL, baseline eGFR, presence of dipstick albuminuria, and eGFR variability during the exposure assessment period.
Prognosis According to Cumulative Metabolic Variability Score
We constructed the cumulative metabolic variability score by summing the number of parameters from among BMI, fasting blood glucose, systolic BP, and total cholesterol that had high variability (Q4). There were 29,405, 32,282, 17,190, 5062, and 697 individuals with 0–4 highly variable metabolic syndrome components in the study population, respectively (Table 5). In the regression analysis, a higher cumulative metabolic variability score was significantly associated with higher risks of all assessed adverse outcomes.
Table 5.
Risk of adverse outcomes according to the number of metabolic parameters with high variability (Q4) from each domain
| Outcome | Number of High Variability (Q4) Componentsa | N | Event | Follow-up Person-years | Incidence Rate (/1000 Person-years) | Univariable Model | Age, Sex, and Number of Exams Adjusted Model | Multivariable Modelb | |||
|---|---|---|---|---|---|---|---|---|---|---|---|
| HR (95% CI) | P | Adjusted HR (95% CI) | P | Adjusted HR (95% CI) | P | ||||||
| All-cause mortality | 0 | 29,405 | 1023 | 115,685 | 8.84 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 |
| 1 | 32,282 | 1765 | 126,694 | 13.93 | 1.574 (1.457 to 1.700) | 1.475 (1.366 to 1.594) | 1.320 (1.221 to 1.427) | ||||
| 2 | 17,190 | 1332 | 66,765 | 19.95 | 2.26 (2.083 to 2.452) | 2.002 (1.845 to 2.173) | 1.590 (1.462 to 1.729) | ||||
| 3 | 5062 | 580 | 19,357 | 29.96 | 3.406 (3.076 to 3.771) | 2.959 (2.671 to 3.277) | 2.141 (1.926 to 2.379) | ||||
| 4 | 697 | 82 | 2682 | 30.58 | 3.465 (2.767 to 4.339) | 3.243 (2.589 to 4.062) | 2.108 (1.678 to 2.649) | ||||
| Myocardial infarction | 0 | 29,405 | 481 | 114,961 | 4.18 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 |
| 1 | 32,282 | 679 | 125,690 | 5.40 | 1.29 (1.148 to 1.45) | 1.237 (1.101 to 1.391) | 1.114 (0.990 to 1.253) | ||||
| 2 | 17,190 | 492 | 66,073 | 7.45 | 1.782 (1.571 to 2.02) | 1.654 (1.458 to 1.876) | 1.338 (1.176 to 1.524) | ||||
| 3 | 5062 | 162 | 19,129 | 8.47 | 2.029 (1.698 to 2.424) | 1.861 (1.556 to 2.224) | 1.373 (1.143 to 1.650) | ||||
| 4 | 697 | 25 | 2644 | 9.46 | 2.263 (1.514 to 3.383) | 2.138 (1.430 to 3.196) | 1.423 (0.948 to 2.137) | ||||
| Stroke | 0 | 29,405 | 731 | 114,485 | 6.39 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 |
| 1 | 32,282 | 1001 | 125,047 | 8.01 | 1.254 (1.140 to 1.379) | 1.187 (1.079 to 1.306) | 1.086 (0.986 to 1.196) | ||||
| 2 | 17,190 | 649 | 65,773 | 9.87 | 1.547 (1.392 to 1.72) | 1.403 (1.262 to 1.56) | 1.171 (1.050 to 1.306) | ||||
| 3 | 5062 | 247 | 18,965 | 13.02 | 2.046 (1.771 to 2.363) | 1.82 (1.574 to 2.103) | 1.403 (1.209 to 1.628) | ||||
| 4 | 697 | 38 | 2610 | 14.56 | 2.287 (1.65 to 3.168) | 2.12 (1.530 to 2.938) | 1.491 (1.072 to 2.074) | ||||
| KRT | 0 | 29,405 | 704 | 114,674 | 6.14 | Reference | <0.001 | Reference | <0.001 | Reference | <0.001 |
| 1 | 32,282 | 1201 | 124,905 | 9.62 | 1.569 (1.430 to 1.722) | 1.699 (1.548 to 1.865) | 1.129 (1.028 to 1.241) | ||||
| 2 | 17,190 | 900 | 65,375 | 13.77 | 2.258 (2.046 to 2.492) | 2.550 (2.309 to 2.815) | 1.218 (1.100 to 1.349) | ||||
| 3 | 5062 | 393 | 18,778 | 20.93 | 3.454 (3.053 to 3.908) | 4.065 (3.591 to 4.601) | 1.410 (1.240 to 1.603) | ||||
| 4 | 697 | 78 | 2569 | 30.36 | 5.027 (3.979 to 6.351) | 5.957 (4.713 to 7.529) | 1.248 (0.980 to 1.589) | ||||
Number of high variability components were calculated by summing the presence of high variability (Q4) in a metabolic parameter from each metabolic syndrome domain (BMI, fasting blood glucose, systolic BP, and total cholesterol).
Multivariable model was adjusted for age, sex, number of exams, current smoking, alcohol consumption, regular physical activity, low-income status, history of diabetes mellitus, hypertension, dyslipidemia, cancer, chronic lung disease, baseline BMI, waist circumference, fasting glucose, systolic BP, diastolic BP, HDL, baseline eGFR, presence of dipstick albuminuria, and eGFR variability during the exposure assessment period.
Discussion
In this observational study, we demonstrated that higher BMI variability was significantly associated with higher risks of all-cause mortality, myocardial infarction, stroke, and progression to the state requiring KRT in patients with predialysis CKD. Higher BMI variability was a significant risk factor for adverse outcomes regardless of obesity, sex, diabetes, reduced baseline eGFR, or even positive or negative trends in BMI. In addition, certain metabolic parameter variabilities showed significant associations with prognosis in patients with predialysis CKD. Furthermore, the cumulative metabolic variability score analysis showed that those with a higher burden of metabolic variability suffered from a worse prognosis. Thus, our study suggests the prognostic significance of BMI and metabolic parameter variabilities in patients with predialysis CKD.
Unlike the traditional concept of a higher BMI causing metabolic disorders, thus leading to higher risks of cardiovascular diseases or all-cause mortality in the general population, an inverse association between BMI and prognosis has been reported in patients with CKD.6 Even among those with obesity or morbid obesity, a higher BMI was associated with better survival in patients with ESKD.17,18 In individuals with predialysis CKD, a previous study investigating >0.4 million US veterans reported a U-shaped association between baseline BMI and the risk of mortality.19 In individuals with a BMI <40 kg/m2, a higher BMI was associated with a lower risk of mortality, and those with a BMI <20 kg/m2 had the highest risk of mortality, further supporting the so-called “obesity paradox” in patients with CKD. There were similar findings for other adverse outcomes, including the progression of kidney function impairment.4 Previous studies investigated whether these findings were associated with residual confounding effects by implementing marginal structural model analysis to account for time-varying confounders and reported that the inverse association between BMI and the risk of mortality was observed with all utilized statistical methods.20 Thus, the consensus is that higher baseline BMI is associated with a better prognosis in patients with CKD, although the biologic plausibility needs further investigation. The unexplained “obesity paradox,” which is different from the findings in the general population, confused the interpretation of the importance of BMI in patients with CKD, although BMI is one of the most well-recognized modifiable risk factors for various metabolic disorders and cardiovascular diseases.
In this study, we primarily aimed to investigate the clinical significance of BMI variability, rather than baseline BMI levels, in patients with predialysis CKD. In this large-scale cohort, we again identified inverse associations between baseline BMI and mortality and other major adverse outcomes in this predialysis CKD population. Furthermore, we identified that higher BMI variability was significantly associated with higher risks of all-cause mortality, KRT, myocardial infarction, and stroke. The significance remained even after the regression models were adjusted for various demographics, lifestyle factors, metabolic parameters, and baseline BMI, eGFR, or eGFR variability. The inclusion of a large number of patients with predialysis CKD with sequential measurements of BMI enabled subgroup analyses, and we identified the risks of adverse outcomes were significantly higher in those with higher BMI variability regardless of sex, obesity, diabetes, reduced eGFR, and trends in BMI. Although some statistical significance was attenuated in the multivariable models for some outcomes other than mortality, the interaction term analysis indicated the possibility of significant interactions with the variables used to determine the subgroups was minimal for the outcomes with attenuated associations, suggesting the findings may be generally applied to the studied population of patients with predialysis CKD. These findings encourage health care providers in the field of nephrology to carefully assess not only baseline BMI values, but also variability in BMI, because high BMI variability was an independent risk factor for major adverse outcomes in patients with CKD.
In addition, there were certain metabolic parameters, the variabilities of which showed significant associations with the risks of adverse outcome in patients with predialysis CKD. The most prominent association was identified for fasting blood glucose variability, the prognostic significance of which has been previously suggested in the general population and in those with diabetes.3,21 Variabilities in waist circumference, systolic BP, and some cholesterol levels also showed certain notable associations, further supporting the clinical importance of metabolic variability in patients with predialysis CKD. In the results, the null findings in the fully adjusted model should be interpreted with caution, because we stringently adjusted for many clinicodemographic characteristics to ensure the robustness of the analysis; however, some adjusted variables might have certain mediating effects on the influence of previous variabilities in metabolic parameters on adverse outcomes. For instance, the fact that components with significant associations with the risk of requirement for KRT were relatively rare might be because the adjustment for baseline eGFR or eGFR variability might have already reflected the effect of metabolic variability on kidney function. Thus, although our results may prioritize the metabolic parameters with particularly important variabilities in patients with predialysis CKD, additional investigation is warranted to ascertain the individual effects of variabilities in metabolic parameters. In addition, that metabolic parameters are certainly correlated with each other (e.g., high waist circumference, indicating central obesity, being a risk factor for impaired glucose tolerance or high blood pressure) should be minded when interpreting the findings.
Previous studies reported the importance of metabolic variability in the general population and in those with diabetes, and a higher degree of variability in metabolic parameters was significantly associated with a worse prognosis.3,22,23 A previous study showed large fluctuations in weight lead to a decrease in muscle mass, but an increase in abdominal fat mass, which may lead to higher risks of metabolic and cardiovascular diseases.5,9 In addition, those with high BMI variability have unstable metabolic parameters, such as blood pressure, which has also been reported to be associated with adverse outcomes.24 In an animal experiment, weight cycling induced increased inflammation in adipose tissue, resulting in insulin resistance.25 In addition, high BMI variability may indicate repetitive overshoot in metabolic stress, which may lead to the accumulation of pathologic changes in the cardiovascular system during the period of worsening metabolic health.26 The above suggested mechanisms may also support the presence of the effects of high BMI variability on adverse outcomes in patients with CKD, because metabolic health is an important prognostic determinant in people with impaired kidney function. The results of the analysis of the metabolic variability score further emphasize the importance of the cumulative burden of variability in metabolic status in patients with predialysis CKD, supporting this mechanistic explanation. A future clinical trial should investigate the effect of minimizing BMI fluctuations or compare the effect of such interventions with that of rapid weight reduction, because rapid weight loss may cause weight regain or cycling, to further confirm the causality of the identified association between BMI variability and prognosis in patients with CKD.27
The study has several limitations. First, whether high BMI variability or metabolic parameter variability was associated with intentional or unintentional weight changes or lifestyle modifications could not be investigated due to the retrospective nature of the study. It should be noted that the effects of intentional weight changes may be different from those of BMI changes related to external factors. Because of this limitation, it is impossible to disregard the possibility that intentionally reducing weight at a modest rate may still be helpful for metabolic health in patients with CKD, despite the fact that doing so would increase BMI variability.7 Second, there was the possibility of selection bias, because those with severe illness would be less likely to undergo general health screenings (healthy volunteer bias), which was reflected in the relatively low prevalence of CKD in the utilized database when compared with the prevalence of CKD of 8.2% in the general population in Korea.28 In addition, the distinct age ranges and Asian ethnicities in this study suggest the need for more studies in other populations to improve the generalizability of the results. Third, survivorship bias should also be considered, because the study population had to undergo successive health screenings within a certain period to be included in this study. Fourth, measurement bias should be considered as follows: although the health screening centers were quality controlled, the weight scales were not directly standardized, and some patients might have visited different centers for their multiple health screenings. In addition, the dipstick albuminuria test has limitations and might have missed modest degrees of albuminuria. Last, the follow-up duration was relatively short; thus, the long-term association between metabolic variability and adverse outcomes in patients with predialysis CKD could not be investigated in this study.
In conclusion, high BMI variability is associated with a worse prognosis in patients with predialysis CKD. In addition to the assessment of metabolic status at a single time point, health care providers in the field of nephrology should pay attention to recent trends in metabolic parameters in CKD patients, because a higher degree of fluctuation may be associated with future risks of mortality or major adverse outcomes.
Disclosures
K. Joo reports receiving research funding from Baxter. All remaining authors have no disclosures.
Funding
This work was supported by the Ministry of Trade, Industry & Energy (Korea) Development of Early Diagnosis Technology for Acute/Chronic Renal Failure the Industrial Strategic Technology Development Program—Development of Biocore Technology grant 10077474.
Supplementary Material
Acknowledgments
The corresponding author attests that all of the listed authors meet the authorship criteria and that no others meeting the criteria have been omitted. K. Joo, D. Kim, H. Lee, and S. Park conceptualized and designed the study; S. Cho, K. Han, K. Joo, D. Kim, Y. C. Kim, Y. S. Kim, J. Lee, S. Lee, C. Lim, and S. Park advised on statistical aspects and interpreted the data; K. Han and S. Park performed the main statistical analysis, assisted by S. Lee; K. Joo, Y. S. Kim; H. Lee, J. Lee, and C. Lim; D. Kim offered advice regarding the data interpretation, obtained funding, and provided supervision for the overall project; all of the authors participated in drafting the manuscript, reviewed the manuscript, and approved the final version to be published. This work was performed under a project associated with the memorandum of understanding between the Korean Society of Nephrology and the NHIS of Korea. The study used the database from the NHIS of the Republic of Korea (REQ000034394).
Footnotes
Published online ahead of print. Publication date available at www.jasn.org.
See related editorial, “Considerations in the study of body mass index variability,” on pages 2395–2397.
Supplemental Material
This article contains the following supplemental material online at http://jasn.asnjournals.org/lookup/suppl/doi:10.1681/ASN.2020121694/-/DCSupplemental.
Supplemental Table 1. Risk of adverse outcomes according to baseline BMI in patients with predialysis CKD.
Supplemental Table 2. Risk of adverse outcomes according to BMI variability by other indexes in patients with predialysis CKD.
Supplemental Table 3. Interaction term analysis results for the subgroup analysis.
Supplemental Table 4. Risk of adverse outcomes according to BMI variability in patients with predialysis CKD stratified by sex.
Supplemental Table 5. Risk of adverse outcomes according to BMI variability in patients with predialysis CKD stratified by baseline BMI ≥25 kg/m2.
Supplemental Table 6. Risk of adverse outcomes according to BMI variability in patients with predialysis CKD stratified by the presence of diabetes mellitus.
Supplemental Table 7. Risk of adverse outcomes according to BMI variability in patients with predialysis CKD stratified by baseline eGFR <60 ml/min per 1.73 m2.
References
- 1.GBD Chronic Kidney Disease Collaboration: Global, regional, and national burden of chronic kidney disease, 1990–2017: A systematic analysis for the Global Burden of Disease Study 2017. Lancet 395: 709–733, 2020 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Park S, Lee S, Kim Y, Lee Y, Kang MW, Han K, et al. : Reduced risk for chronic kidney disease after recovery from metabolic syndrome: A nationwide population-based study. Kidney Res Clin Pract 39: 180–191, 2020 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kim MK, Han K, Park YM, Kwon HS, Kang G, Yoon KH, et al. : Associations of variability in blood pressure, glucose and cholesterol concentrations, and body mass index with mortality and cardiovascular outcomes in the general population. Circulation 138: 2627–2637, 2018 [DOI] [PubMed] [Google Scholar]
- 4.Ryu S, Chang Y, Woo HY, Kim SG, Kim DI, Kim WS, et al. : Changes in body weight predict CKD in healthy men. J Am Soc Nephrol 19: 1798–1805, 2008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.van der Kooy K, Leenen R, Seidell JC, Deurenberg P, Hautvast JG: Effect of a weight cycle on visceral fat accumulation. Am J Clin Nutr 58: 853–857, 1993 [DOI] [PubMed] [Google Scholar]
- 6.Kalantar-Zadeh K, Rhee CM, Chou J, Ahmadi SF, Park J, Chen JL, et al. : The obesity paradox in kidney disease: How to reconcile it with obesity management. Kidney Int Rep 2: 271–281, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Wade KH, Chiesa ST, Hughes AD, Chaturvedi N, Charakida M, Rapala A, et al. : Assessing the causal role of body mass index on cardiovascular health in young adults: Mendelian randomization and recall-by-genotype analyses. Circulation 138: 2187–2201, 2018 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Seong SC, Kim YY, Park SK, Khang YH, Kim HC, Park JH, et al. : Cohort profile: the National Health Insurance Service-National Health Screening Cohort (NHIS-HEALS) in Korea. BMJ Open 7: e016640, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Cheol Seong S, Kim YY, Khang YH, Heon Park J, Kang HJ, Lee H, et al. : Data resource profile: The national health information database of the national health insurance service in South Korea. Int J Epidemiol 46: 799–800, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Park S, Lee S, Kim Y, Lee Y, Kang MW, Han K, et al. : Altered risk for cardiovascular events with changes in the metabolic syndrome status: A nationwide population-based study of approximately 10 million persons. Ann Intern Med 171: 875–884, 2019 [DOI] [PubMed] [Google Scholar]
- 11.Park S, Lee S, Kang MW, Han K, Kim Y, An JN, et al. : Postdischarge major adverse cardiovascular events of ICU survivors who received acute renal replacement therapy. Crit Care Med 46: e1047–e1054, 2018 [DOI] [PubMed] [Google Scholar]
- 12.Park S, Lee S, Kim Y, Lee Y, Kang MW, Cho S, et al. : Association of CKD with incident tuberculosis. Clin J Am Soc Nephrol 14: 1002–1010, 2019 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Schneeweiss S, Rassen JA, Brown JS, Rothman KJ, Happe L, Arlett P, et al. : Graphical depiction of longitudinal study designs in health care databases. Ann Intern Med 170: 398–406, 2019 [DOI] [PubMed] [Google Scholar]
- 14.Alberti KG, Eckel RH, Grundy SM, Zimmet PZ, Cleeman JI, Donato KA, et al. ; International Diabetes Federation Task Force on Epidemiology and Prevention; Hational Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; International Association for the Study of Obesity: Harmonizing the metabolic syndrome: a joint interim statement of the International Diabetes Federation Task Force on Epidemiology and Prevention; National Heart, Lung, and Blood Institute; American Heart Association; World Heart Federation; International Atherosclerosis Society; and International Association for the Study of Obesity. Circulation 120: 1640–1645, 2009 [DOI] [PubMed] [Google Scholar]
- 15.Kawamoto R, Kohara K, Tabara Y, Miki T, Ohtsuka N, Kusunoki T, et al. : An association between body mass index and estimated glomerular filtration rate. Hypertens Res 31: 1559–1564, 2008 [DOI] [PubMed] [Google Scholar]
- 16.Seo MH, Lee WY, Kim SS, Kang JH, Kang JH, Kim KK, et al. ; Committee of Clinical Practice Guidelines, Korean Society for the Study of Obesity (KSSO): 2018 Korean Society for the Study of Obesity guideline for the management of obesity in Korea. J Obes Metab Syndr 28: 40–45, 2019 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Kalantar-Zadeh K, Block G, Humphreys MH, Kopple JD: Reverse epidemiology of cardiovascular risk factors in maintenance dialysis patients. Kidney Int 63: 793–808, 2003 [DOI] [PubMed] [Google Scholar]
- 18.Ahmadi SF, Zahmatkesh G, Streja E, Molnar MZ, Rhee CM, Kovesdy CP, et al. : Body mass index and mortality in kidney transplant recipients: A systematic review and meta-analysis. Am J Nephrol 40: 315–324, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lu JL, Kalantar-Zadeh K, Ma JZ, Quarles LD, Kovesdy CP: Association of body mass index with outcomes in patients with CKD. J Am Soc Nephrol 25: 2088–2096, 2014 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Doshi M, Streja E, Rhee CM, Park J, Ravel VA, Soohoo M, et al. : Examining the robustness of the obesity paradox in maintenance hemodialysis patients: A marginal structural model analysis. Nephrol Dial Transplant 31: 1310–1319, 2016 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Muggeo M, Zoppini G, Bonora E, Brun E, Bonadonna RC, Moghetti P, et al. : Fasting plasma glucose variability predicts 10-year survival of type 2 diabetic patients: The Verona Diabetes Study. Diabetes Care 23: 45–50, 2000 [DOI] [PubMed] [Google Scholar]
- 22.Kim MK, Han K, Kim HS, Park YM, Kwon HS, Yoon KH, et al. : Cholesterol variability and the risk of mortality, myocardial infarction, and stroke: a nationwide population-based study. Eur Heart J 38: 3560–3566, 2017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Nam GE, Kim W, Han K, Lee CW, Kwon Y, Han B, et al. : Body weight variability and the risk of cardiovascular outcomes and mortality in patients with type 2 diabetes: A nationwide cohort study. Diabetes Care 43: 2234–2241, 2020 [DOI] [PubMed] [Google Scholar]
- 24.Bae EH, Lim SY, Han KD, Oh TR, Choi HS, Kim CS, et al. : Association between systolic and diastolic blood pressure variability and the risk of end-stage renal disease. Hypertension 74: 880–887, 2019 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Anderson EK, Gutierrez DA, Kennedy A, Hasty AH: Weight cycling increases T-cell accumulation in adipose tissue and impairs systemic glucose tolerance. Diabetes 62: 3180–3188, 2013 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Montani JP, Viecelli AK, Prévot A, Dulloo AG: Weight cycling during growth and beyond as a risk factor for later cardiovascular diseases: The ‘repeated overshoot’ theory. Int J Obes 30: S58–S66, 2006 [DOI] [PubMed] [Google Scholar]
- 27.Korkeila M, Rissanen A, Kaprio J, Sorensen TI, Koskenvuo M: Weight-loss attempts and risk of major weight gain: A prospective study in Finnish adults. Am J Clin Nutr 70: 965–975, 1999 [DOI] [PubMed] [Google Scholar]
- 28.Park JI, Baek H, Jung HH: Prevalence of chronic kidney disease in Korea: the Korean National Health and Nutritional Examination Survey 2011–2013. J Korean Med Sci 31: 915–923, 2016 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.