Considerations in the Study of Body Mass Index Variability : Journal of the American Society of Nephrology

Journal Logo

Up Front Matters: Editorials

Considerations in the Study of Body Mass Index Variability

Gregg, L. Parker1,2,3; Navaneethan, Sankar D.1,2,3,4

Author Information
JASN 32(10):p 2395-2397, October 2021. | DOI: 10.1681/ASN.2021060844
  • Free

Clinical risk prediction models generally incorporate data points from an isolated point in time to prognosticate risk at the bedside. However, in clinical practice, it is impossible to ignore variability in biologic parameters (such as BP and blood sugar) that are dynamic, and that some individuals exceed typical degrees of variability in these measures. Prior epidemiologic investigations reported that high variability of various biologic parameters is associated with adverse clinical outcomes. Body mass index (BMI) variability has been previously associated with mortality and cardiovascular events in the Framingham cohort and in patients with diabetes mellitus.1,2 In the Korean general population, the highest quartiles of variability in fasting glucose, BP, cholesterol, and body weight were associated with emergency hospitalization and death, with a significantly stronger relationship among individuals with CKD stages 3–5 than those without CKD.3 Given that weight gain and weight loss are known to affect kidney hemodynamics and other kidney pathology,4 and is common in CKD,5 it is vital to better understand the relationship between BMI variability and adverse health outcomes among individuals with CKD.

To address this knowledge gap, Park et al.6 conducted a nationally representative, retrospective cohort study of Korean patients with nondialysis CKD stages 1–4. Using claims data, they evaluated the association of variability in BMI and other metabolic parameters with death, myocardial infarction, stroke, and progression to dialysis dependence or kidney transplantation. CKD was defined as the presence of persistent eGFR <60 ml/min per 1.73 m2 or dipstick albuminuria >1+. BMI variability was primarily measured as the variability independent of the mean, with sensitivity analyses presented using the SD, coefficient of variation, and average real variability, which are more commonly used measures of variability in other such studies.7 Individuals with high BMI variability were more likely to have diabetes mellitus and heart failure and less likely to engage in regular physical activity. The highest quartile of BMI variability was associated with increased risk of all studied outcomes, independent of baseline BMI, diabetes status, waist circumference, eGFR, and other relevant confounders. Variability of systolic and diastolic BP; waist circumference; total, LDL, and HDL cholesterol; and fasting blood glucose were independently associated with adverse outcomes. Outcome rates were higher in those losing weight over time, with diabetes, and with CKD stages 3–4 versus 1–2.

To address whether an increased cumulative burden of variability in multiple metabolic parameters was associated with adverse outcomes, the authors constructed a score on the basis of the sum of parameters in the highest quartile of variability, including BMI, fasting blood glucose, systolic BP, and total cholesterol. They demonstrated that, when compared with a score of zero, high variability in two, three, or four parameters was associated with all outcomes. Only 38.1% of participants had one highly variable metabolic parameter, indicating that high variability in these parameters may not happen in concert in many patients, which is surprising because change in weight is often accompanied by change in other metabolic parameters in clinical practice.

This analysis has several strengths, including the availability of comprehensive comorbidity details, careful consideration of confounding factors, and availability of clinical outcome data. Despite these, several factors should be considered in the interpretation of the results and variability in observational studies. First, most measures of variability reflect the range and distribution of measurements without taking into account the order in which those measurements were obtained. There may be important physiologic differences between a steadily rising BMI, a steadily falling BMI, or one that is fluctuating within a range. Still, each of these three phenotypes may have the same calculated BMI variability (Figure 1A). In this study, the authors addressed this question by evaluating the slope of BMI change, and found that BMI variability was associated with outcomes independent of the slope. Second, real-world retrospective study participants may have a different number of measurements, with a wide range of time elapsed between measures (Figure 1B). Those with more frequent contact with the healthcare system may have more chronic illnesses or have greater healthcare needs that could confound the association of high variability with adverse outcomes. Third, calculating the variance of just a few data points may not accurately represent the true range that the individual experiences (Figure 1C). A low number of measurements per participant it is not likely to capture the extremes of BMI and may underestimate variation in those with fewer measurements.

Figure 1.:
Several considerations affect the interpretation of BMI variability in real-world observational research. These hypothetical graphs illustrate some limitations in the calculation of BMI variability in real-world retrospective observational studies. (A) Different patterns in the trajectory of BMI may indicate important differences in physiology but generate the same calculated variability over a discrete period of time. (B) The interval of time between measurements may be inconsistent and leave long gaps without measured data points. (C) For individuals with a small number of available data points, calculated BMI variability may underestimate actual variability.

In contrast to cholesterol or BP variability, which may imply noncompliance with treatment or worsening clinical status, BMI variability may be attributed to several reasons. Although BMI is frequently used as a surrogate marker of adiposity, muscle mass and total body water are also dynamic potential contributors to changes in body weight, and, therefore, BMI variability. Fluctuation in extracellular water could be more pronounced in individuals with CKD and heart failure (either due to volume overload or diuretic therapy) and confounds the relationship between BMI variability and cardiovascular outcomes. Although an increase in BMI may relate to an unhealthy lifestyle, those with a decline in BMI could reflect worsening health status. In large retrospective datasets, it is challenging to differentiate individuals with a decrease in BMI due to an amputation or severe cachexia, indicating more severe underlying disease. In contrast, a decrease in BMI due to intentional weight loss could have positive effects.

In well-conducted experimental studies, repeated episodes of caloric restriction and ad libitum refeeding resulted in regaining of body fat and cell size after initial weight loss.8,9 Despite this, weight cycling significantly increased the life span, relative to those remaining with obesity, and had a similar benefit to those with sustained modest weight loss. These data contrast with observational study data in humans, such as in Park et al.,6 that note harmful associations with BMI variability which likely captures weight cycling. However, given the known harmful effects of adiposity, despite lack of high-quality clinical trial evidence, clinicians should encourage weight loss in those with higher BMI and the avoidance weight fluctuations, as is recommended by clinical practice guidelines.10

In summary, although the relationship between BMI variability and adverse outcomes satisfies some of the Hill criteria of causality, such as the strength of associations and temporality, several other criteria are yet to be satisfied. Despite our lack of ability to infer causality and unclear reasons for the variability in BMI noted in the study, these data highlight the need to identify patients with CKD who are losing or gaining weight, explore the reasons, and follow them more closely for occurrence of adverse outcomes.


L.P. Gregg and S.D. Navaneethan report being employees of the US Department of Veterans Affairs. S.D. Navaneethan reports serving as a scientific advisor for, or member of, American Journal of Kidney Disease, American Journal of Nephrology, Cardiorenal Medicine, CJASN, Current Opinion in Nephrology and Hypertension, and Kidney Disease Improving Global Outcomes (guideline writing committee member); receiving personal fees from Bayer, Boehringer-Ingelheim, Reata, Tricida, and Vifor; and receiving grants from Keryx.


S.D. Navaneethan is supported by the U.S. Department of Veterans AffairsOffice of Research and Development grant 1I01HX002917-01A1, and by National Institute of Diabetes and Digestive and Kidney Diseases grant R01DK101500. This work was also supported, in part, by the U.S. Department of Veterans Affairs Center for Innovations in Quality, Effectiveness and Safety grant CIN 13-413 (via the Michael E. DeBakey Veterans Affairs Medical Center, Houston, TX).

Published online ahead of print. Publication date available at

See related article, “The prognostic significance of body mass index and metabolic parameter variabilities in predialysis chronic kidney disease: A nationwide observational cohort study,” on pages .


The interpretation and reporting of these data are the responsibility of the authors and in no way should be viewed as official policy or interpretation of the Department of Veterans Affairs or the US Government.


1. Lissner L, Odell PM, D’Agostino RB, Stokes J 3rd, Kreger BE, Belanger AJ, et al.: Variability of body weight and health outcomes in the Framingham population. N Engl J Med 324: 1839–1844, 1991
2. Bangalore S, Fayyad R, DeMicco DA, Colhoun HM, Waters DD: Body weight variability and cardiovascular outcomes in patients with type 2 diabetes mellitus. Circ Cardiovasc Qual Outcomes 11: e004724, 2018
3. Lee SH, Han K, Kwon HS, Yoon KH, Kim MK: Effect of variability in blood pressure, glucose and cholesterol concentrations, and body weight on emergency hospitalization and 30-day mortality in the general population. J Am Heart Assoc 9: e017475, 2020
4. Liu DJX, Stock E, Broeckx BJG, Daminet S, Meyer E, Delanghe JR, et al.: Weight-gain induced changes in renal perfusion assessed by contrast-enhanced ultrasound precede increases in urinary protein excretion suggestive of glomerular and tubular injury and normalize after weight-loss in dogs. PLoS One 15: e0231662, 2020
5. Ku E, Kopple JD, Johansen KL, McCulloch CE, Go AS, Xie D, et al.; CRIC Study Investigators: Longitudinal weight change during CKD progression and its association with subsequent mortality. Am J Kidney Dis 71: 657–665, 2018
6. Park S, Cho S, Lee S, Kim Y, Park S, Kim YC, et al.: The prognostic significance of body mass index and metabolic parameter variabilities in predialysis chronic kidney disease: A nationwide observational cohort study. J Am Soc Nephrol 32: 2595–2612, 2021
7. Gregg LP, Hedayati SS, Yang H, Van Buren PN, Banerjee S, Navaneethan SD, et al.: Association of blood pressure variability and diuretics with cardiovascular events in patients with chronic kidney disease stages 1-5. Hypertension 77: 948–959, 2021
8. Smith DL Jr, Yang Y, Nagy TR, Patki A, Vasselli JR, Zhang Y, et al.: Weight cycling increases longevity compared with sustained obesity in mice. Obesity (Silver Spring) 26: 1733–1739, 2018
9. List EO, Berryman DE, Wright-Piekarski J, Jara A, Funk K, Kopchick JJ: The effects of weight cycling on lifespan in male C57BL/6J mice. Int J Obes 37: 1088–1094, 2013
10. Friedman AN, Kaplan LM, le Roux CW, Schauer PR: Management of obesity in adults with CKD. J Am Soc Nephrol 32: 777–790, 2021

BMI; CKD; mortality

Copyright © 2021 by the American Society of Nephrology