Introduction

The burden of chronic kidney disease (CKD) is rising worldwide due to population aging and an increased prevalence of CKD risk factors [1]. An estimated 697.5 million individuals, corresponding to a global prevalence of 9.1%, had a diagnosis of CKD in 2017 [1]. Further, 1.2 million individuals died from CKD in 2017 and the disease is projected to rise in rank as a leading cause of death [1]. CKD is generally more common in women than men; however, the disease is often more severe among men as indicated by the higher rates of end-stage kidney disease (ESKD) and age-standardized mortality [1]. Although CKD is recognized as an important health issue, the awareness of the disease remains low among the general population and healthcare providers [2].

CKD is defined by kidney damage and includes a range of abnormalities in kidney structure or function [3]. The disorder encompasses different disease stages, of which some early stages can be asymptomatic in contrast to the clinically apparent advanced-stages, including ESKD. The disease is associated with a range of adverse effects, including cardiovascular complications, reduced quality of life, and premature mortality [2].

The primary risk factors for CKD vary by country; however, diabetes and hypertension are among the established leading causes, particularly in high-income countries [2,4]. Additionally, increasing evidence supports that excess adult adiposity is a substantial and causal contributor to CKD risk [5–8]. Although overweight and obesity are closely related to both diabetes and hypertension, it is suggested that excess adiposity may have effects on CKD risks independent of these 2 major causes of CKD [9].

In addition to the high prevalence of overweight in adults, overweight and obesity in children now similarly constitute an enormous global health crisis with more than 124 million children living with obesity [10]. Despite the presence of early kidney abnormalities and decreased kidney function among children with severe obesity [7], the potential early origins of CKD are sparsely investigated [11–16]. Among the few studies in this area, some included body size in late adolescence [12,14], and many only provided associations with ESKD [11,12,14]. Further, although some studies investigated associations with patterns of BMI development from child- to adulthood [13,15,16], the majority of these were limited by including subclinical renal damage as a proxy for CKD [15,16]. Therefore, using a large Danish cohort with repeated measurements of childhood height and weight, we prospectively investigated sex-specific associations between childhood BMI trajectories and weight status at individual childhood ages and adult-onset CKD and ESKD, respectively, and the role of adult-onset type 2 diabetes (T2D) on these associations.

Methods

Childhood latent class trajectories and weight status

The children in the CSHRR had up to 12 measurements recorded. All children with a minimum of 2 BMI values (≥95% of children had more than 2 values) were included when estimating sex-specific BMI trajectories using latent class trajectory models, which impute missing values under a missing at random assumption [18–20]. We a priori chose to generate BMI trajectories separately by sex to account for differences in growth among boys and girls. The trajectories were modelled using natural splines with knot points positioned at ages 8, 10, and 12 years (corresponding approximately to the 25th, 50th, and 75th percentiles) to best capture the natural growth patterns of children at these ages as well as to evenly space these points across the age span and were adjusted for birth cohort (5-year intervals). Several parameters and fit indices (i.e., proportion in each trajectory, mean posterior probability, Bayesian information criteria, relative entropy, odds of correct classification, visual inspection of the trajectories) were used to identify the optimal number of trajectories [19]. We tested models with up to 8 different childhood BMI trajectories (S1 and S2 Figs). The best performing model identified 5 noncrossing BMI trajectories for each sex (S3 Fig) and did not include random effects, whereby the variation in BMI level within each trajectory was reduced. These BMI trajectories were termed as below-average, average, above-average, overweight, and obese. Most boys and girls followed the average BMI trajectory, and the fewest belonged to the obese BMI trajectory (Table 1). For each child, a posterior probability, which is a measure of how well a child’s BMI trajectory fits within the identified trajectories, was assigned for each of the 5 trajectories. Furthermore, due to nonlinearity in the associations, at individual childhood ages, BMI values were classified as underweight (<fifth BMI percentile), normal weight (≥fifth and <85th BMI percentile), or overweight, including obesity (≥85th BMI percentile) using the Centers for Disease Control and Prevention (CDC) criteria [21].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of individuals in the CSHRR by childhood BMI trajectory and by sex.

https://doi.org/10.1371/journal.pmed.1004098.t001

Results

During a median of 30.8 (IQR: 18.7 to 40.0) person-years of follow-up, 5,968 men and 3,903 women were diagnosed with CKD and 977 men and 543 women with ESKD (Table 1). The median age at diagnosis for CKD was 67 years (IQR: 57 to 73 years) for men and 66 years (IQR: 55 to 74 years) for women and for ESKD the median age at diagnosis was 62 years (IQR: 51 to 71 years) for men and 62 years (IQR: 50 to 70 years) for women.

A significantly lower rate of CKD was observed for men with the below-average childhood BMI trajectory in comparison with the average trajectory (used as the reference) (Fig 1A and S4 Table). In contrast, men with the above-average BMI, overweight, and obese childhood BMI trajectories had significantly higher rates of CKD, in comparison with the average BMI reference group, and the strength of the association increased in a dose–response manner. The patterns of associations with rates of CKD were similar among women. Women with childhood BMI trajectories that were above-average, overweight, and obese had significantly higher CKD rates than those with the average childhood BMI trajectory (Fig 1B and S4 Table). Although the association with the below-average childhood BMI trajectory did not reach statistical significance among women, it still indicated a lower CKD rate. We did not find that the occurrence of adult-onset T2D significantly modified the associations between childhood BMI trajectories and rates of CKD among men (p-value = 0.19) or women (p-value = 0.90). When including T2D in the analyses, the associations were statistically significant among men with the above-average (IRR = 1.07, 95% CI: 1.00 to 1.14), overweight (IRR = 1.25, 95% CI: 1.14 to 1.38), and obese (IRR = 1.39, 95% CI: 1.13 to 1.72) childhood BMI trajectories and women with the above-average (IRR = 1.18, 95% CI: 1.08 to 1.28), overweight (IRR = 1.24, 95% CI: 1.11 to 1.38), and obese (IRR = 1.54, 95% CI: 1.28 to 1.86) childhood BMI trajectories. When including T2D, associations with the below-average childhood BMI trajectory did not reach statistical significance in men (IRR = 0.93, 95% CI: 0.86 to 1.00) or women (IRR = 0.94, 95% CI: 0.85 to 1.04).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Childhood BMI trajectories and CKD and ESKD.

(A) CKD among men in analyses without (white) and with (black) T2D. (B) CKD among women in analyses without (white) and with (black) T2D. (C) ESKD among men in analyses without (white) and with (black) T2D. (D) ESKD among women in analyses without (white) and with (black) T2D. BMI, body mass index; CI, confidence interval; CKD, chronic kidney disease; ESKD, end-stage kidney disease; T2D, type 2 diabetes.

https://doi.org/10.1371/journal.pmed.1004098.g001

Investigations of ESKD generally showed similar patterns of associations with childhood BMI trajectories as were found for CKD. In analyses that did not include adult-onset T2D, men with the overweight and obese childhood BMI trajectories had significantly higher rates of ESKD than men with the average childhood BMI trajectory, whereas no significant associations were observed for the below- and the above-average childhood BMI trajectories (Fig 1C and S4 Table). Among women in analyses that did not include T2D, those with the above-average, overweight, and obese childhood BMI trajectories had significantly higher ESKD rates than women with the average childhood BMI trajectory, whereas women with the below-average BMI trajectory had a significantly lower ESKD rate (Fig 1D and S2 Table). We did not find that the occurrence of adult-onset T2D significantly modified the associations between childhood BMI trajectories and rates of ESKD among men (p-value = 0.29) or women (p-value = 0.83). When including adult-onset T2D, the associations were not statistically significant among men. Among women, when including T2D, those with the overweight (IRR = 1.50, 95% CI: 1.14 to 1.97) and obese (IRR = 1.97, 95% CI: 1.25 to 3.11) childhood BMI trajectories had significantly higher rates of ESKD compared with the average childhood BMI trajectory, and those with the below-average BMI trajectory (IRR = 0.74, 95% CI: 0.56 to 0.99) had a significantly lower ESKD rate.

The cumulative incidences for CKD were highest for the childhood BMI trajectories with the highest mean BMI, at older adult ages, among men and among individuals with adult-onset T2D (Table 2). A similar pattern was observed for ESKD, although the cumulative incidences were lower than for CKD (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Cumulative incidences of CKD and ESKD by childhood BMI trajectory.

Sex-specific 10-year cumulative incidences (in percentages) of CKD and ESKD in individuals aged 55 and 65 years with and without T2D, respectively, by childhood BMI trajectories when taking competing risk of death into account.

https://doi.org/10.1371/journal.pmed.1004098.t002

When categorizing childhood weight status at ages 7, 10, and 13 years, most boys and girls were classified as having a normal weight, which was used as the reference group, and the fewest as having underweight (Table 3). Men and women with overweight at childhood ages had higher rates of CKD in comparison with individuals with normal weight in childhood, whereas childhood underweight was generally not associated with adult-onset CKD (Table 3). We did not find strong indications of an effect modification by T2D in the analyses of childhood age-specific weight status and rates of CKD among men (p-values ≥ 0.39) or women (p-values ≥ 0.23), except for 1 significant interaction with weight status at age 13 years among boys (p-value = 0.03). When including adult-onset T2D in the analyses, the rates of CKD were significantly higher among men (IRR_{age 13 years} = 1.37, 95% CI: 1.24 to 1.52) and women (IRR_{age 13 years} = 1.36, 95% CI: 1.22 to 1.53) with childhood overweight in comparison to individuals with childhood normal weight and the estimates were higher with older childhood ages (Table 3). Both men and women with overweight at childhood ages had significantly higher rates of ESKD than those with normal weight, while underweight in childhood was not associated with ESKD (Table 3). We did not find indications of an effect modification by T2D in the analyses of childhood age-specific weight status and rates of ESKD among men (p-values ≥ 0.50) or women (p-values ≥ 0.43). When including adult-onset T2D in the analyses for ESKD, the associations with overweight in childhood were not statistically significant for men (IRR_{age 13 years} = 1.22, 95% CI: 0.94 to 1.58) but were statistically significant for women at ages 10 and 13 years (IRR_{age 13 years} = 1.83, 95% CI: 1.40 to 2.39).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Childhood weight status and CKD and ESKD.

Sex-specific IRRs and corresponding 95% CIs of the associations between childhood weight status at ages 7, 10, and 13 years and CKD and ESKD, respectively, in analyses with and without adult-onset T2D and duration with T2D.

https://doi.org/10.1371/journal.pmed.1004098.t003

Discussion

Men and women with childhood BMI trajectories above average had higher rates of CKD compared to individuals with the average childhood BMI trajectory, even when including the development and duration of adult T2D in the analyses. Although similar associations were observed for ESKD, when including T2D, these were only statistically significant among women but not among men. Additionally, overweight at individual childhood ages was associated with higher rates of CKD and ESKD in comparison to childhood normal weight. When T2D was included in these analyses, associations with ESKD were only statistically significant among women.

The BMI trajectories we used have the advantage of being able to account for dynamic changes in body size across childhood. We identified 5 sex-specific BMI trajectories, which align with the contemporary United States CDC 2000 growth reference [21], and we found that CKD rates are elevated among children with BMI growth patterns below levels classified as child overweight and obesity by this reference. A limited number of studies have investigated associations between patterns of BMI growth in early life and later CKD [13,15,16]. Using the 1946 British Birth cohort, BMI trajectories created from categories of overweight (yes/no) at ages 2 to 20 years were evaluated in relation to CKD [13]. The results showed that trajectories of consistent overweight or trajectories of developing overweight were positively related to CKD risks in midlife (60 to 64 years) [13]. Although the trajectories investigated in our study are not directly comparable with those in the British study due to differences in ages examined and how the trajectories were created, both studies consistently found that patterns of early life BMI development that include overweight are important in relation to CKD risk. Nevertheless, the findings in the previous study were limited by the low number of cases (<100). Two recent studies using data from China and Australia, respectively, found that trajectories with high BMI values from child (6 to 15 or 7 to 15 years) to adult life (36 to 45 or 36 to 50 years) were positively associated with risks of subclinical renal damage in midlife used as a proxy for CKD [15,16]. These associations remained after diabetes was accounted for (either through adjustment or exclusion). Nonetheless, these studies were limited by low case numbers (257 and 82, respectively) and are not directly comparable with our study due to the age ranges investigated and as they included a subclinical outcome.

Two studies investigated if categories of body size measured at conscription examinations in late adolescence or early adulthood was associated with ESKD [12,14]. An Israeli study of male and female conscripts found that individuals with overweight and obesity at age 17 years had higher ESKD risks compared to those with normal weight, whereas no associations were found for underweight [12]. In a study of Swedish male conscripts, young adult overweight and obesity at ages 18 to 19 years was similarly associated with an increased ESKD risk compared with normal weight, whereas underweight was not [14]. Although the ages at BMI assessment differed from our study and the potential role of T2D was not evaluated, the findings from these 2 previous studies are comparable with ours and emphasize that overweight and obesity at single ages early in life may be markers of ESKD disease processes, whereas underweight most likely is not.

When including T2D as a time-varying covariate, which accounts for the time exposed to T2D before a CKD diagnosis, the associations between childhood exposures and CKD were strong and statistically significant. For ESKD, when T2D was included as a time-varying covariate, the associations with childhood BMI trajectories in women were strong and significant. In men, however, the associations were attenuated. As T2D is likely a mediator in the associations between childhood body size and adult-onset CKD, thus when including T2D in the analyses, we are not estimating the total effect of childhood body size. The impact of T2D on the development of CKD is further evident from the 10-year cumulative incidences, whereby men and women with T2D across all childhood BMI trajectories had higher cumulative incidences of CKD than individuals without T2D.

The biological mechanisms underlying the associations between excess adiposity and CKD are likely multifactorial and complex but are not fully understood. The main factors suggested to explain these links include the alterations of renal or systemic hemodynamics (including hypertension), adverse metabolic effects such as insulin resistance and chronic inflammation, and renal lipid accumulation, all contributing to kidney damage [7,9]. Another potential explanation is the tracking of body size, as it has been shown that excess adiposity at childhood ages is likely to track into adult ages [28]. Nonetheless, tracking of body size is not likely to entirely explain the association as children with severe obesity already display kidney abnormalities [7]. Additionally, a large mendelian randomization study found evidence for a causal relationship between adult BMI levels of overweight or obesity with CKD [8], supporting that the associations in our study reflect an effect of BMI.

Among the major strengths of the present study are the repeated measurements of weight and height in a large cohort of school-aged children, whereby we were able to evaluate associations with both body size at single time points as well as BMI development during childhood. Thus, we were able to take full advantage of the richness of this unique data resource by including all available childhood BMI values as opposed to restricting to information at specific ages. Selection into the cohort is limited as health examinations were performed at both public and private schools. The risk of recall bias related to memory of child body size is eliminated as the measurements were recorded as they were taken. Due to a long follow-up period with virtually no loss to follow-up as well as obtaining information on diagnoses in national and validated health registers in a nation with universal healthcare access, an adequate number of unselected cases were included. Furthermore, the risk of reverse causality affecting our associations through the inclusion of undiagnosed kidney diseases is minimized as we used a longitudinal study design with a period of minimum 15 years between the exposures and outcomes. Finally, although our cohort is historical, whereby the comparability with more contemporary populations can be questioned, historical data resources are a prerequisite to obtain a long follow-up period when examining long-term health outcomes and our study also included more recent birth cohorts.

Our study also has some limitations. Our follow-up period spans several decades; thus, it is likely that diagnostic changes have occurred as well as changes in the awareness of CKD by healthcare providers. We did not have a high and systematic coverage of laboratory tests available during the follow-up period [29]; thus, we were unable to include estimated glomerular filtration rate or albuminuria levels in the case ascertainment as recommended by the Kidney Disease Improving Global Outcomes guidelines [3]. In Denmark, patients with either mild or moderate CKD are treated in primary care, whereas patients seen in a hospital setting represent the more severe CKD cases [30]. As such, a Danish study found, unsurprisingly, that the prevalence of CKD was markedly lower using hospital-based (775 per 100,000 person-years) versus different laboratory-based (4,637 to 8,327 per 100,000 person-years) diagnoses of CKD, whereas the 1-year mortality risk was profoundly higher among individuals with hospital-based (22%) than laboratory-based CKD diagnoses (7% to 9%) [30]. Reflecting this, the median age at ESKD diagnosis was lower than the median age at a CKD diagnosis in our study. Thus, our findings are primarily generalizable to more severe cases of CKD. A small Danish validation study found that diagnoses of moderate to severe kidney disease in the Danish National Patient Register had a positive predictive value of 100% [31], suggesting that the diagnoses included in the present study are of high validity. The definition and assessment of T2D has changed across time. Whereas older cohorts only captured more advanced hospital-diagnosed cases, milder T2D cases are also registered in more recent decades, and due to the age structure of our cohort, this is when most patients with T2D in our cohort are registered. Further, we were limited by not being able to take T2D treatment regimens into account and nor did we perform formal mediation analyses of T2D on the associations between childhood body size and adult-onset CKD. To do this would have required information on potential confounders of the T2D and CKD association, and, unfortunately, this information was not systematically available for our study. Due to the nature of the available national health registers, the quality of information on hypertension (either from diagnostic codes or prescriptions) was estimated to be too poor due to its lack of specificity to evaluate its effects on the investigated associations. The identified childhood BMI trajectories rely in part on a subjective evaluation of the optimal model. However, as multiple fit indices were carefully examined, it is likely that the child BMI development and variation is accurately captured by the 5 identified trajectories for each sex, and this is also supported by the similarities with the CDC growth reference. In addition, the uncertainty of trajectory membership was accounted for by including the posterior probabilities of belonging to the trajectories in the analyses as opposed to an all-or-nothing assignment.

The children in our study were born across 57 years of birth, spanning periods before, during, and after World War II, during which numerous significant societal and medical changes occurred, including significant improvements in the healthcare system and important socioeconomic changes. Nonetheless, it is reassuring for our findings that we did not observe any birth cohort effects on these. The children included in our study were mainly of Northern European Caucasian descent, as such, our findings are likely generalizable to many Western populations. However, the potential impact of race and ethnicity on the findings requires further investigation.

In conclusion, in this study, we observed that men and women with childhood BMI trajectories above average had higher rates of adult-onset CKD and ESKD in comparison with individuals with average childhood BMI trajectories. Further, overweight at individual childhood ages was positively related to CKD and ESKD rates in comparison with childhood normal weight. Our findings when including T2D suggest that childhood body size provides information about adult-onset CKD beyond the information provided by T2D. With a worldwide growing population of children with overweight and obesity, our findings highlight the importance of continued efforts to attain or maintain a healthy weight and further add to the understanding of CKD disease etiology and potential modifiable risk factors.

Supporting information

Acknowledgments

The CSHRR was initiated and planned by Dr. Thorkild I.A. Sørensen and built by the Institute of Preventive Medicine, The Capital Region of Denmark.