To assess 1) the association between metabolic dysfunction–associated steatotic liver disease (MASLD) in pediatric obesity and youth-onset type 2 diabetes, 2) the joint effect of MASLD and intermediate hyperglycemia on type 2 diabetes risk, and 3) the effect of obesity treatment on type 2 diabetes risk.
A cohort study using the Swedish Childhood Obesity Treatment Register (Barnobesitas Registret i Sverige [BORIS]) (1999–2020) linked with national registers was conducted. We included 10,346 children with overweight or obesity and 59,336 matched control individuals. MASLD was defined by transaminases and diagnosis code, separately. Type 2 diabetes was ascertained from national registers.
In the obesity cohort, median age at type 2 diabetes diagnosis was 16.9 (quartile 1 [Q1], quartile 3 [Q3]: 14.7, 21.4) years, median follow-up was 8.1 (Q1, Q3: 5.1, 11.7) years. Cumulative incidence of type 2 diabetes at age 30 was 22.7% (obesity and MASLD), 9.9% (obesity alone), and 0.7% (control individuals). MASLD was associated with risk for type 2 diabetes (hazard ratio [HR] 2.71 [95% CI 2.14–3.43]), independently of age, sex, degree of obesity, intermediate hyperglycemia, and parental type 2 diabetes. Joint effect of MASLD and intermediate hyperglycemia increased type 2 diabetes risk (HR 9.04 [6.38–12.79]). Optimal response in obesity treatment reduced the risk (HR 0.23 [0.09–0.57]).
MASLD, defined by transaminases or diagnosis code, in pediatric obesity is associated with increased risk for youth-onset type 2 diabetes. MASLD interacts synergistically with intermediate hyperglycemia to dramatically increase the risk. Optimal response in obesity treatment reduces type 2 diabetes risk, despite MASLD.
Introduction
The rising prevalence of childhood obesity has led to a higher incidence of youth-onset cardiometabolic disorders, including type 2 diabetes and metabolic dysfunction–associated steatotic liver disease (MASLD) (1,2). A bidirectional association between MASLD, whose previous terminology was nonalcoholic fatty liver disease (3), and type 2 diabetes has been shown, particularly in adults (4–6). To which extent the association can be inferred to pediatric population is precarious because the phenotypes of pediatric MASLD and type 2 diabetes differ from the adult types (7,8). Moreover, whether MASLD in pediatric obesity adds excess risk for type 2 diabetes remains unclear.
In pediatric obesity, intermediate hyperglycemia is associated with an increased risk for type 2 diabetes (9). Nevertheless, whether the joint effect of intermediate hyperglycemia and MASLD leads to a greater risk for type 2 diabetes remains unknown. This study assessed 1) the association between pediatric MASLD and the risk for youth-onset type 2 diabetes in patients undergoing pediatric obesity treatment and their general population comparators, 2) the synergistic effect of MASLD and intermediate hyperglycemia on the risk for youth-onset type 2 diabetes, and 3) the effect of pediatric obesity treatment on the risk.
Research Design and Methods
Study Design
A nationwide cohort study of children undergoing obesity treatment enrolled in the Swedish Childhood Obesity Treatment Register (Barnobesitas Registret i Sverige [BORIS], https://www.e-boris.se/in-english/) (1999–2020) was conducted. Children with overweight or obesity (10), aged 9–18 years, and free from type 2 diabetes at baseline were included. They were monitored from the index date (i.e., the date of MASLD diagnosis or corresponding matching date among control individuals) until type 2 diabetes was detected, age 30 years, death, emigration, or end of follow-up (July 2023), whichever occurred first. We excluded secondary causes of steatotic liver disease or elevated transaminases (e.g., Wilson disease, autoimmune hepatitis, and α-1-antitrypsin deficiency) (The full list of exclusion criteria is in Supplementary Table 1). Also excluded were type 1 diabetes, maturity-onset of diabetes in the young, and genetic syndromes. The obesity cohort was matched with ratio 1:5 on sex, birth year, and residential area with the general population.
The effect of obesity treatment on the risk for type 2 diabetes was assessed in a subgroup of the obesity cohort undergoing ≥6 months of obesity treatment. The follow-up was started at the first visit after 6 months of treatment.
All available diagnostic tests for pediatric MASLD have substantial bias. To strengthen the confidence in the finding (11), we applied two separated approaches for identifying MASLD. The first approach was based on ALT levels in outpatient obesity care recorded in the BORIS register. Elevated ALT, in the absence of other hepatic diseases, together with the presence of overweight or obesity, may be used to diagnose MASLD, according to a recent position statement from several global pediatric hepatology societies (12). The second approach used diagnosis code for MASLD recorded in the Patient Register (13). The code was according to ICD-10. The code for MASLD has a high positive predictive value in Swedish patients (14). Table 1 summarizes the detailed description of the approaches. Individuals with ALT-based MASLD and diagnostic code–based MASLD in this study were nonidentical for most cases (diagnostic overlap of ALT-based and diagnostic code–based MASLD in Supplementary Table 2). This study was approved by the Regional Ethical Review Board in Stockholm (2016/922-31/1; 2020-02707).
Two approaches to identify MASLD
. | Approach A (ALT-based MASLD)–main analysis . | Approach B (diagnosis code–based MASLD) . |
---|---|---|
Study design | Cohort | Matched cohort (up to 5 no-MASLD individuals for each MASLD case) |
Population | Children with overweight or obesity who had transaminases and glucose data at obesity treatment initiation (n = 10,346). | Children with overweight/obesity who had obesity care visit data at ±1 year from the index date (n = 1,557). |
Exposure (MASLD) | Patients with ALT levels more than twice the ULN (n = 1,102) | Patients with ICD-10 code for liver steatosis recorded in the National Patient Register (n = 281). |
Comparison (non-MASLD) | Patients with normal ALT levels (n = 5,361). | Patients without any ICD-10 code for liver steatosis, matched (1: up to 5) by sex, birth year, and date of obesity treatment visit (±1 year) (n = 1,276). |
Outcome (type 2 diabetes) | Type 2 diabetes diagnosis based on ICD-10 codes in the National Patient Register and/or type 2 diabetes medications based on Anatomical Therapeutic Chemical codes in the Prescribed Drug Register. Algorithm to ascertain type 2 diabetes is available in the Supplementary Material. | |
Covariates | Sex, age, degree of obesity, intermediate hyperglycemia at baseline, and parental type 2 diabetes. Additional covariate in subpopulation analysis (n = 2,524): BMI SDS changes in children undergoing at least 6 months of obesity treatment. | Degree of obesity and intermediate hyperglycemia at baseline, parental type 2 diabetes. Additional covariate in subpopulation analysis (n = 778): BMI SDS changes in children undergoing at least 6 months of obesity treatment. |
. | Approach A (ALT-based MASLD)–main analysis . | Approach B (diagnosis code–based MASLD) . |
---|---|---|
Study design | Cohort | Matched cohort (up to 5 no-MASLD individuals for each MASLD case) |
Population | Children with overweight or obesity who had transaminases and glucose data at obesity treatment initiation (n = 10,346). | Children with overweight/obesity who had obesity care visit data at ±1 year from the index date (n = 1,557). |
Exposure (MASLD) | Patients with ALT levels more than twice the ULN (n = 1,102) | Patients with ICD-10 code for liver steatosis recorded in the National Patient Register (n = 281). |
Comparison (non-MASLD) | Patients with normal ALT levels (n = 5,361). | Patients without any ICD-10 code for liver steatosis, matched (1: up to 5) by sex, birth year, and date of obesity treatment visit (±1 year) (n = 1,276). |
Outcome (type 2 diabetes) | Type 2 diabetes diagnosis based on ICD-10 codes in the National Patient Register and/or type 2 diabetes medications based on Anatomical Therapeutic Chemical codes in the Prescribed Drug Register. Algorithm to ascertain type 2 diabetes is available in the Supplementary Material. | |
Covariates | Sex, age, degree of obesity, intermediate hyperglycemia at baseline, and parental type 2 diabetes. Additional covariate in subpopulation analysis (n = 2,524): BMI SDS changes in children undergoing at least 6 months of obesity treatment. | Degree of obesity and intermediate hyperglycemia at baseline, parental type 2 diabetes. Additional covariate in subpopulation analysis (n = 778): BMI SDS changes in children undergoing at least 6 months of obesity treatment. |
Data Source
The BORIS register covers pediatric obesity treatment care in all health care levels, from primary care to university hospitals across Sweden. The register includes patients with and without comorbidities. The treatment mainly comprises behavioral lifestyle modification. During the obesity treatment period of the current study, the new generation antiobesity drugs was not approved for adolescents in Sweden; thus, no patients received pharmacological treatment of obesity. All health care for children to age 18 years is free of charge. BORIS records clinical and laboratory data (e.g., weight, height, fasting glucose, glycated hemoglobin [HbA1c], transaminases). Details about BORIS are published elsewhere (15). BORIS was linked with several Swedish national registers through the Swedish unique personal identity number. Medical diagnosis (e.g., type 2 diabetes, MASLD, and other causes of steatosis) was obtained from the Swedish National Patient Register. This register records medical diagnosis made in specialized care. Prescribed antidiabetes medications in any level of care was obtained from the Swedish National Prescribed Drug Register. Date of death was obtained from the Cause of Death Register. Migration data and general population comparators were obtained from the Total Population Register. Supplementary Table 3 describes details about each register.
Variables
Sex was categorized as male and female, based on legal sex from Total Population Register.
Weight status was measured as BMI standard deviation score (SDS) according to the International Obesity Task Force (10) and categorized as overweight, class I, class II, and class III obesity corresponding to adult BMI of 25 to <30, 30 to <35, 35 to <40, and ≥40 kg/m2.
ALT-based MASLD was defined as ALT two or more times upper limit of normal (ULN), where the ULN was 22 units/L (0.37 μkat/L) for girls and 26 units/L in boys (0.44 μkat/L) (16). No MASLD was defined as ALT at or less than ULN. Patients belonging to neither categories were categorized as the borderline group.
Diagnosis code-based MASLD was identified using ICD-10 code K76.0 fatty (change of) liver, not elsewhere classified (13).
Type 2 diabetes was defined using ICD-10 code E11 (type 2 diabetes) or O24.1 (preexisting type 2 diabetes mellitus, in pregnancy, childbirth and the puerperium) in the Patient Register and/or prescribed antidiabetes medications in the Prescribed Drug Register (Supplementary Table 4). For instance, a patient who was prescribed the combination of metformin and sulfonylureas or was prescribed metformin alone frequently was considered to have type 2 diabetes (the full algorithm to ascertain type 2 diabetes is in Supplementary Fig. 5).
Intermediate hyperglycemia was defined by fasting glucose at baseline between 6.1 and 6.9 mmol/L (110 and 125 mg/dL) (9) and/or HbA1c between 5.7 and 6.4% (39 and 46 mmol/mol) (17).
Parental type 2 diabetes, as a proxy for heredity, was defined as at least one parent with type 2 diabetes (from the Patient Register and/or Prescribed Drug Register) before age 50.
Obesity treatment response was measured as BMI SDS reduction (per 0.1-unit decrease). Optimal response was defined as BMI SDS reduction of at least 0.25 units (18,19).
Migrant background was divided into Nordic (born in, and with 1 or 2 parents born in, Sweden, Norway, Denmark, Finland, or Iceland) or non-Nordic.
Statistical Analysis
Statistical analysis was performed using ALT-based MASLD and diagnosis code–based MASLD as the main exposure, separately. MASLD reported in the main text refers to ALT-based MASLD, unless mentioned otherwise. Corresponding analyses with diagnosis code–based MASLD as the main exposure are reported in the Supplementary Material.
The characteristics of the study population are reported as proportions for categorical variables or as median, first quartile (Q1), and third quartile (Q3) for continuous variables.
Effect of MASLD on the Risk for Type 2 Diabetes
The incidence rate of type 2 diabetes per 10,000 person-years (P-Y) and the 95% CI was calculated. Cumulative incidence of type 2 diabetes in individuals with both obesity and MASLD, obesity alone, and the general population comparators was estimated using flexible parametric regression.
Overall hazard ratio (HR) of MASLD on type 2 diabetes in the obesity cohort was calculated using Cox regression with attained age as the timescale. HR was adjusted for the covariates of age, sex, degree of obesity, intermediate hyperglycemia at baseline, and parental type 2 diabetes. Since a time-dependent effect of MASLD on type 2 diabetes was indicated from visual inspection of hazard rates (Supplementary Fig. 6) and Schoenfeld residual test, adjusted HRs over time were also calculated using flexible parametric regression with time-varying effect. The baseline hazard was modeled using 4 df for the spline variables and 3 df for the time-varying effect. Further, by splitting the observation time for each individual to multiple observations, HRs of MASLD on type 2 diabetes during age 9–19 and 20–30 years were estimated using Cox regression.
Joint Effect Between Intermediate Hyperglycemia and MASLD on Type 2 Diabetes
The joint effect was measured on an additive scale (20). Adjusted HRs were estimated from Cox regression. Relative excess risk due to interaction (RERI), attributable proportion due to interaction (AP), and synergy index (S) and their 95% CIs were estimated (20). Synergistic effect is considered evident when RERI and its 95% CI do not include 0, and when S and its 95% CI do not include 1.
Effect of BMI SDS Reduction on the Risk for Type 2 Diabetes
The effect of BMI SDS reduction was estimated using Cox regression, adjusted for all covariates. BMI SDS reduction was treated as a continuous variable and categorical variable, separately. To see whether the effect of BMI SDS reduction on the risk for type 2 diabetes differed by MASLD or by degree of obesity at baseline, interaction terms between those variables were checked.
Sensitivity Analysis
Because guidelines recommend repeated ALT measurements to indicate MASLD (16,21), a sensitivity analysis was performed in a subpopulation that had two ALT measurements at least 6 months apart. MASLD was defined as persistently elevated ALT two or more times ULN, and its association with risk for type 2 diabetes was assessed. Because different ALT cutoffs have been suggested (22), another sensitivity analysis was performed using ALT >35 units/L, also ALT >25 units/L for boys or ALT >22 units/L for girls as the cutoffs. Moreover, because exposure misclassification bias is possible both when using ALT and diagnosis code to identify MASLD, we performed sensitivity analysis to quantify the impact of the misclassification (23). We developed different scenarios of varying sensitivity and specificity of MASLD classification and then accounted for misclassification and random error with 1,000 times simulation. Bias-adjusted HRs and 95% CIs were reported.
Data and Resource Availability
Patient-level data are not publicly available because of third-party data. Data are available from the Swedish Childhood Obesity Treatment Register, Statistics Sweden, and the Swedish National Board of Health and Welfare for researchers with permission of the Swedish Ethical Review Authority.
Results
Study Population
At baseline, there were 1,122 individuals with MASLD and 9,294 individuals without MASLD in the obesity cohort. Of these, 1.8% (20 of 1,122) of individuals with MASLD and 0.5% (50 of 9,294) of individuals without MASLD were excluded because of type 2 diabetes prior to obesity treatment. Therefore, 10,346 individuals (MASLD, n = 1,102) in the obesity cohort were included in the analysis.
The median age at type 2 diabetes diagnosis was 16.9 (Q1, Q3: 14.7, 21.4) years in the obesity cohort and 22.8 (18.7, 26.6) years in the general population comparators. Table 2 summarizes characteristics between groups.
Characteristics of the obesity cohort and the general population comparators
. | Obesity cohort(n = 10,346) . | No MASLD(n = 5,361) . | Borderline(n = 3,883) . | MASLD(n = 1,102) . | General populationcomparators (n = 59,336) . |
---|---|---|---|---|---|
Sex | |||||
Boys | 5,889 (56.9) | 2,935 (54.8) | 2,196 (56.6) | 758 (68.8) | 27,898 (56.9) |
Girls | 4,457 (43.1) | 2,426 (45.2) | 1,687 (43.4) | 344 (31.2) | 21,092 (43.1) |
Age, years | 12.5 (10.8, 14.5) | 12.3 (10.8, 14.2) | 12.6 (10.7, 14.7) | 13.4 (11.3, 15.5) | 12.5 (10.8, 14.5) |
Nordic origin | 6,918 (68.7) | 3,545 (68.1) | 2,650 (70.1) | 723 (67.0) | 34,965 (74.1) |
BMI SDS | 2.69 (2.43, 2.99) | 2.60 (2.37, 2.89) | 2.77 (2.50, 3.07) | 2.89 (2.61, 3.21) | |
Overweight | 1,082 (10.5) | 751 (14.0) | 282 (7.3) | 49 (4.4) | |
Class I obesity | 5,814 (56.2) | 3,238 (60.4) | 2,066 (53.2) | 510 (46.3) | |
Class II obesity | 2,559 (24.7) | 1,093 (20.4) | 1,107 (28.5) | 359 (32.6) | |
Class III obesity | 891 (8.6) | 279 (5.2) | 428 (11.0) | 184 (16.7) | |
Intermediate hyperglycemia | 1,110 (10.7) | 517 (9.6) | 447 (11.5) | 146 (13.3) | |
Parental diabetes | 2,311 (22.3) | 1,100 (20.5) | 920 (23.7) | 291 (26.4) | 6,115 (12.5) |
Age at end of follow-up (years) | 20.8 (17.7, 24.7) | 20.7 (17.6, 24.4) | 20.9 (17.7, 24.7) | 21.0 (17.8, 25.3) | 21.0 (17.9, 24.9) |
Length of follow-up (years) | 8.1 (5.1, 11.7) | 8.1 (5.2, 11.7) | 8.1 (5.2, 11.7) | 7.7 (4.5, 11.6) | 8.3 (5.4, 11.8) |
Reason follow-up ended | |||||
Type 2 diabetes | 515 (5.0) | 190 (3.5) | 208 (5.4) | 117 (10.6) | 117 (0.2) |
Death | 26 (0.3) | 11 (0.2) | 13 (0.3) | 2 (0.2) | 78 (0.2) |
Emigration | 97 (0.9) | 52 (1.0) | 36 (0.9) | 9 (0.8) | 910 (1.9) |
End of study period | 9,708 (93.8) | 5,108 (95.3) | 3,626 (93.4) | 974 (88.4) | 47,885 (97.7) |
. | Obesity cohort(n = 10,346) . | No MASLD(n = 5,361) . | Borderline(n = 3,883) . | MASLD(n = 1,102) . | General populationcomparators (n = 59,336) . |
---|---|---|---|---|---|
Sex | |||||
Boys | 5,889 (56.9) | 2,935 (54.8) | 2,196 (56.6) | 758 (68.8) | 27,898 (56.9) |
Girls | 4,457 (43.1) | 2,426 (45.2) | 1,687 (43.4) | 344 (31.2) | 21,092 (43.1) |
Age, years | 12.5 (10.8, 14.5) | 12.3 (10.8, 14.2) | 12.6 (10.7, 14.7) | 13.4 (11.3, 15.5) | 12.5 (10.8, 14.5) |
Nordic origin | 6,918 (68.7) | 3,545 (68.1) | 2,650 (70.1) | 723 (67.0) | 34,965 (74.1) |
BMI SDS | 2.69 (2.43, 2.99) | 2.60 (2.37, 2.89) | 2.77 (2.50, 3.07) | 2.89 (2.61, 3.21) | |
Overweight | 1,082 (10.5) | 751 (14.0) | 282 (7.3) | 49 (4.4) | |
Class I obesity | 5,814 (56.2) | 3,238 (60.4) | 2,066 (53.2) | 510 (46.3) | |
Class II obesity | 2,559 (24.7) | 1,093 (20.4) | 1,107 (28.5) | 359 (32.6) | |
Class III obesity | 891 (8.6) | 279 (5.2) | 428 (11.0) | 184 (16.7) | |
Intermediate hyperglycemia | 1,110 (10.7) | 517 (9.6) | 447 (11.5) | 146 (13.3) | |
Parental diabetes | 2,311 (22.3) | 1,100 (20.5) | 920 (23.7) | 291 (26.4) | 6,115 (12.5) |
Age at end of follow-up (years) | 20.8 (17.7, 24.7) | 20.7 (17.6, 24.4) | 20.9 (17.7, 24.7) | 21.0 (17.8, 25.3) | 21.0 (17.9, 24.9) |
Length of follow-up (years) | 8.1 (5.1, 11.7) | 8.1 (5.2, 11.7) | 8.1 (5.2, 11.7) | 7.7 (4.5, 11.6) | 8.3 (5.4, 11.8) |
Reason follow-up ended | |||||
Type 2 diabetes | 515 (5.0) | 190 (3.5) | 208 (5.4) | 117 (10.6) | 117 (0.2) |
Death | 26 (0.3) | 11 (0.2) | 13 (0.3) | 2 (0.2) | 78 (0.2) |
Emigration | 97 (0.9) | 52 (1.0) | 36 (0.9) | 9 (0.8) | 910 (1.9) |
End of study period | 9,708 (93.8) | 5,108 (95.3) | 3,626 (93.4) | 974 (88.4) | 47,885 (97.7) |
Data are presented as n (%) or as median (Q1, Q3).
The highest incidence rate (IR) of type 2 diabetes was found in patients with overweight or obesity and MASLD (IR 131.1 [95% CI 109.4–157.0] per 10,000 P-Y) and the lowest was observed in the general population comparators (IR 2.8 [2.3–3.3] per 10,000 P-Y). Supplementary Fig. 7 shows the cumulative incidence over time by groups. The IR of type 2 diabetes in diagnosis code–based MASLD was even higher (IR 316.8 [242.1–414.7] per 10,000 P-Y) than that of ALT-based MASLD. Supplementary Table 8 and Supplementary Fig. 9 show characteristics and incidence of type 2 diabetes in patients with diagnosis code–based MASLD and the control group.
Association Between MASLD and the Risk for Type 2 Diabetes
In the obesity cohort, MASLD was associated with increased risk for type 2 diabetes (HR 2.71 [95% CI 2.14–3.43]), independently of sex, age, degree of obesity, intermediate hyperglycemia, and parental type 2 diabetes (Table 3). The association was prominent during adolescence (HR 3.99 [2.99–5.32]) and diminished in adulthood (HR 1.28 [0.79–2.07]). The MASLD group had ∼10 additional cases of type 2 diabetes per 1,000 patients during 10 years of follow-up. Supplementary Fig. 10 presents HRs varying over time. Female sex, older age, higher degree of obesity, intermediate hyperglycemia, and parental type 2 diabetes were also associated with an increased risk for type 2 diabetes (Table 3).
HRs of ALT-based MASLD and other covariates for developing type 2 diabetes in the obesity cohort
Variables at baseline . | 9–30 years of age . | 9–19 years of age . | 20–30 years of age . | ||||||
---|---|---|---|---|---|---|---|---|---|
IR per 10,000 P-Y . | UnadjustedHR (95% CI) . | AdjustedHR (95% CI) . | IR per 10,000 P-Y . | UnadjustedHR (95% CI) . | AdjustedHR (95% CI) . | IR per 10,000 P-Y . | UnadjustedHR (95% CI) . | AdjustedHR (95% CI) . | |
No MASLD | 41.2 | Ref | Ref | 33.9 | Ref | Ref | 59.3 | Ref | Ref |
Borderline | 63.1 | 1.54 (1.27–1.88)** | 1.33 (1.09–1.62)* | 69.3 | 2.06 (1.60–2.66)** | 1.66 (1.29–2.15)** | 56.4 | 0.95 (0.67–1.33) | 0.95 (0.67–1.34) |
MASLD | 131.1 | 3.22 (2.54–4.03)** | 2.71 (2.14–3.43)** | 180.1 | 5.35 (4.03–7.09)** | 3.99(2.99–5.32)** | 69.7 | 1.15 (0.71–1.84) | 1.28 (0.79–2.07) |
Males | 43.4 | Ref | Ref | 49.7 | Ref | Ref | 32.9 | Ref | Ref |
Females | 78.1 | 1.80 (1.51–2.14)** | 1.92 (1.61–2.29)** | 75.3 | 1.53 (1.23–1.89)** | 1.71 (1.38–2.12)** | 92.4 | 2.78 (1.98–3.89)** | 2.80 (1.99–3.93)** |
Age (years) at baseline | 1.15 (1.10–1.20)** | 1.09 (1.04–1.13)** | 1.30 (1.22–1.39)** | 1.21 (1.13–1.29)** | 1.02 (0.95–1.10) | 0.99 (0.92–1.06) | |||
Overweight/class I obesity | 35.2 | Ref | Ref | 29.9 | Ref | Ref | 45.0 | Ref | Ref |
Class II obesity | 85.4 | 2.50 (2.05–3.04)** | 2.09 (1.71–2.56)** | 103.7 | 3.53 (2.75–4.52)** | 2.73 (2.12–3.51)** | 70.9 | 1.51 (1.06–2.15)* | 1.42 (0.99–2.05) |
Class III obesity | 148.0 | 4.49 (3.56–5.65)** | 2.99 (2.35–3.81)** | 232.3 | 8.23 (6.20–10.91)** | 4.91 (3.66–6.59)** | 72.0 | 1.88 (1.25–2.83)* | 1.59 (1.04–2.45)* |
No intermediate hyperglycemia | 29.8 | Ref | Ref | 49.0 | Ref | Ref | 55.4 | Ref | Ref |
Intermediate hyperglycemia | 136 | 2.71 (2.21–3.32)** | 2.36 (1.93–2.91)** | 165.6 | 2.25 (2.64–4.26)** | 2.73 (2.15–3.47)** | 95.7 | 1.72 (1.13–2.62)* | 1.64 (1.08–2.51)* |
No parental type 2 diabetes | 45.9 | Ref | Ref | 47.2 | Ref | Ref | 49.6 | Ref | Ref |
Parental type 2 diabetes | 100.8 | 2.20 (1.85–2.62)** | 1.94 (1.62–2.31)** | 107.9 | 2.29 (1.84–2.85)** | 1.88 (1.51–2.34)** | 90.1 | 1.81 (1.31–2.50)** | 1.73 (1.25–2.39)** |
Variables at baseline . | 9–30 years of age . | 9–19 years of age . | 20–30 years of age . | ||||||
---|---|---|---|---|---|---|---|---|---|
IR per 10,000 P-Y . | UnadjustedHR (95% CI) . | AdjustedHR (95% CI) . | IR per 10,000 P-Y . | UnadjustedHR (95% CI) . | AdjustedHR (95% CI) . | IR per 10,000 P-Y . | UnadjustedHR (95% CI) . | AdjustedHR (95% CI) . | |
No MASLD | 41.2 | Ref | Ref | 33.9 | Ref | Ref | 59.3 | Ref | Ref |
Borderline | 63.1 | 1.54 (1.27–1.88)** | 1.33 (1.09–1.62)* | 69.3 | 2.06 (1.60–2.66)** | 1.66 (1.29–2.15)** | 56.4 | 0.95 (0.67–1.33) | 0.95 (0.67–1.34) |
MASLD | 131.1 | 3.22 (2.54–4.03)** | 2.71 (2.14–3.43)** | 180.1 | 5.35 (4.03–7.09)** | 3.99(2.99–5.32)** | 69.7 | 1.15 (0.71–1.84) | 1.28 (0.79–2.07) |
Males | 43.4 | Ref | Ref | 49.7 | Ref | Ref | 32.9 | Ref | Ref |
Females | 78.1 | 1.80 (1.51–2.14)** | 1.92 (1.61–2.29)** | 75.3 | 1.53 (1.23–1.89)** | 1.71 (1.38–2.12)** | 92.4 | 2.78 (1.98–3.89)** | 2.80 (1.99–3.93)** |
Age (years) at baseline | 1.15 (1.10–1.20)** | 1.09 (1.04–1.13)** | 1.30 (1.22–1.39)** | 1.21 (1.13–1.29)** | 1.02 (0.95–1.10) | 0.99 (0.92–1.06) | |||
Overweight/class I obesity | 35.2 | Ref | Ref | 29.9 | Ref | Ref | 45.0 | Ref | Ref |
Class II obesity | 85.4 | 2.50 (2.05–3.04)** | 2.09 (1.71–2.56)** | 103.7 | 3.53 (2.75–4.52)** | 2.73 (2.12–3.51)** | 70.9 | 1.51 (1.06–2.15)* | 1.42 (0.99–2.05) |
Class III obesity | 148.0 | 4.49 (3.56–5.65)** | 2.99 (2.35–3.81)** | 232.3 | 8.23 (6.20–10.91)** | 4.91 (3.66–6.59)** | 72.0 | 1.88 (1.25–2.83)* | 1.59 (1.04–2.45)* |
No intermediate hyperglycemia | 29.8 | Ref | Ref | 49.0 | Ref | Ref | 55.4 | Ref | Ref |
Intermediate hyperglycemia | 136 | 2.71 (2.21–3.32)** | 2.36 (1.93–2.91)** | 165.6 | 2.25 (2.64–4.26)** | 2.73 (2.15–3.47)** | 95.7 | 1.72 (1.13–2.62)* | 1.64 (1.08–2.51)* |
No parental type 2 diabetes | 45.9 | Ref | Ref | 47.2 | Ref | Ref | 49.6 | Ref | Ref |
Parental type 2 diabetes | 100.8 | 2.20 (1.85–2.62)** | 1.94 (1.62–2.31)** | 107.9 | 2.29 (1.84–2.85)** | 1.88 (1.51–2.34)** | 90.1 | 1.81 (1.31–2.50)** | 1.73 (1.25–2.39)** |
HRs and the 95% CIs were estimated using Cox regression. Adjusted HRs were calculated from mutually adjusted models.
*P < 0.05;
**P < 0.001.
Of patients starting obesity treatment at the age of 12, baseline probability of having type 2 diabetes by age 19 was estimated:
1. In a girl with MASLD but no intermediate hyperglycemia at baseline, the estimated probability for type 2 diabetes by age 19 was 9.2%.
2. In a girl with intermediate hyperglycemia but no MASLD, the estimated probability was 6.4%.
3. In a girl with both MASLD and intermediate hyperglycemia, the estimated probability was 23.6%.
A positive association between MASLD and type 2 diabetes was also observed when the main exposure was diagnosis-code based MASLD (adjusted HR 4.16 [95% CI 2.80–6.17]) (Supplementary Fig. 11 and Supplementary Table 12). The association remained when migrant background was included as a covariate (adjusted HR 4.07 [2.73–6.07]).
Synergistic Effect of MASLD and Intermediate Hyperglycemia
In the obesity cohort, the incidence for type 2 diabetes was dramatically higher in the group of patients with both MASLD and intermediate hyperglycemia at baseline than in those with MASLD alone or intermediate hyperglycemia alone (Fig. 1A). Compared with patients with obesity alone, patients with both MASLD and intermediate hyperglycemia had a HR for developing type 2 diabetes of 9.04 (95% CI 6.38–12.79), MASLD only of 2.16 (1.63–2.87), and intermediate hyperglycemia only of 1.79 (1.22–2.60) (see Fig. 1B). A synergistic effect between MASLD and intermediate hyperglycemia was found (RERI 6.09 [3.10–9.07] and S 4.12 [2.39–7.10], all P < 0.001). Within the group with both intermediate hyperglycemia and MASLD, 67% of the type 2 diabetes cases were attributed by the synergistic effect (AP 0.67, P < 0.001).
A: Cumulative incidence of type 2 diabetes by MASLD and intermediate hyperglycemia (IH). The shaded areas indicate the 95% CI of the cumulative incidence. B: Joint effect MASLD and IH for developing type 2 diabetes in the obesity cohort.
A: Cumulative incidence of type 2 diabetes by MASLD and intermediate hyperglycemia (IH). The shaded areas indicate the 95% CI of the cumulative incidence. B: Joint effect MASLD and IH for developing type 2 diabetes in the obesity cohort.
Effect of BMI SDS Reduction on the Risk for Type 2 Diabetes
In a subpopulation undergoing obesity treatment of ≥6 months (median treatment duration 17.3 [Q1, Q3: 12.9, 27.8] months), every 0.1-unit decrease in BMI SDS was associated with a relative risk reduction for type 2 diabetes of 9% (adjusted HR 0.91 [95% CI 0.88–0.93]). There was no interaction between BMI SDS reduction and MASLD (P = 0.665) or between BMI SDS reduction and degree of obesity at baseline (P = 0.173). Moreover, an optimal response in obesity treatment was associated with a relative risk reduction for type 2 diabetes of at least 43% (adjusted HR 0.23 [0.09–0.57]). The subpopulation had similar distribution of sex and intermediate hyperglycemia as the overall study population (P > 0.05), but the subpopulation was on average 6 months younger (P < 0.001) and had higher proportion of class 2 obesity (39% in subpopulation vs. 32% in overall study population, P < 0.001).
In the model with diagnosis code–based MASLD as an exposure, BMI SDS reduction was associated with lower risk for type 2 diabetes (HR 0.87 [95% CI 0.79–0.95], P = 0.003 for every 0.1-unit decrease) (Supplementary Table 13).
Sensitivity Analysis
When MASLD was identified based on repeated measures of ALT, which was available in a subpopulation (43%, n = 4,445), the association between MASLD and type 2 diabetes remained, although the strength of the association was attenuated (adjusted HR 1.70 [95% CI 1.06–2.72], with persistently normal ALT as the reference) (Supplementary Table 14). The association also remained when the exposure was ALT >35 units/L (adjusted HR 1.92 [1.60–2.32]) or ALT >25 units/L for boys or ALT >22 units/L for girls (adjusted HR 1.70 [1.42–2.04]) (see Supplementary Fig. 15). The association between MASLD and type 2 diabetes also persisted after correction for misclassification bias of exposure (HR 5.62 [2.65–38.51]) (Supplementary Table 16).
Conclusions
Using data from one of the globally largest pediatric obesity treatment registers, this study demonstrates that pediatric patients with both obesity and MASLD have an increased risk for youth-onset type 2 diabetes compared with pediatric patients with obesity alone and the general population comparators. The increased risk is independent of intermediate hyperglycemia and heredity of type 2 diabetes. Furthermore, there is a synergistic effect between MASLD and intermediate hyperglycemia on the risk for youth-onset type 2 diabetes. Nevertheless, a decrease in the degree of obesity in pediatric years reduces the risk of youth-onset type 2 diabetes, despite MASLD.
Although obesity per se is a risk factor for type 2 diabetes, our study indicates that MASLD increases the risk for youth-onset type 2 diabetes in the population with pediatric obesity, with most cases manifesting before age 20. Even after adjustment for heredity and intermediate hyperglycemia, the strong association between MASLD and type 2 diabetes remained. Moreover, the study indicated a diminished effect of MASLD on type 2 diabetes after age 20. The reasons for this are unclear, but some possible explanations include detection bias due to health-seeking behavior or screening practices in the adult health care system; if a patient is prone to develop type 2 diabetes early in life, pediatric obesity with MASLD may advance the onset; or weight trajectories after pediatric obesity treatment might modify the association.
Although intermediate hyperglycemia has been known as an important predictor for type 2 diabetes (24), the current study interestingly indicates that MASLD, compared with intermediate hyperglycemia, contributes to a higher incidence rate for adolescent type 2 diabetes. We also demonstrate a synergistic effect of MASLD and intermediate hyperglycemia, where 67% of all type 2 diabetes cases were attributed to the synergy. Our findings emphasize the importance of comprehensive care, including glucose and liver screening, in pediatric obesity treatment. Routine glucose screening is typically a part of pediatric obesity management, whereas MASLD screening is still often overlooked (25). Moreover, the recommendation for MASLD screening varies across pediatric obesity guidelines (26–28). For instance, American guidelines (28) recommend liver screening for children with obesity aged ≥10 years, whereas U.K. guidelines (27) recommend liver screening for all children with obesity referred to specialized multidisciplinary care.
Insulin resistance and systemic low-grade inflammation are both drivers of type 2 diabetes and MASLD (29), and hence, a bidirectional association between them is plausible. However, our study indicates that among most children and adolescents with obesity, MASLD precedes type 2 diabetes. Of all patients with MASLD in our obesity cohort, <2% had type 2 diabetes at baseline. Moreover, in line with our findings, some studies have shown that pediatric dysglycemia may be found earlier or simultaneously as MASLD (19,30,31), but type 2 diabetes is likely to manifest later (32).
In our study, the incidence rate for type 2 diabetes in the MASLD group was markedly higher when MASLD was identified from diagnosis code than ALT-based MASLD (316 vs. 131 per 10,000 P-Y). Different methods to identify MASLD and distribution of sex, age, degree of obesity, and health-seeking behavior might contribute to this variation of incidence rate. It is possible that diagnosis code–based MASLD (i.e., MASLD diagnosed in hospital specialized care) represents more severe forms and therefore associated with higher risk for type 2 diabetes. Elevated ALT might identify more individuals with inflammation or a severe form of MASLD, rather than solely steatosis (33). Among children with biopsy specimen–proven MASLD, steatohepatitis or fibrosis appear to have an increased risk for type 2 diabetes than steatosis alone (32). Additionally, type 2 diabetes is more prevalent among those with elevated ALT than normal ALT (33). Moreover, those with liver steatosis and dysglycemia have higher odds for steatohepatitis (30). MASLD has also been associated with poorer glycemic control and increased risk for vascular complications among adults with type 2 diabetes (34). Accordingly, the interplay between adiposity, MASLD, and type 2 diabetes affect may not only the occurrence of the diseases but also their progression.
In line with our previous study showing the benefit of optimal obesity treatment response to almost eliminate the risk for pediatric MASLD (19), the current study adds that a reduction of 0.25 BMI SDS units in real-life pediatric obesity treatment is clinically meaningful to reduce the risk for developing type 2 diabetes before age 30 (relative risk reduction of at least 43%), regardless of MASLD. Given that younger onset of type 2 diabetes is associated with significantly higher cardiometabolic morbidities (8) and faster deterioration of glycemic control (35), our finding highlights the importance of obtaining an optimal treatment response in pediatric obesity treatment to prevent such consequences.
Despite the favorable effect of optimal response in pediatric obesity treatment, generally only few children with obesity reached the BMI SDS reduction target of 0.25 units. In Sweden, approximately one in five children with obesity aged ≥10 years obtained this target within 1 year of obesity treatment (36). The proportion of children obtaining the optimal target is henceforth likely to be higher given the availability of incretin-based medications (37,38) for the pediatric population. Nevertheless, the current study further indicates that even a modest BMI SDS reduction of 0.1 units, compared with zero reduction, is associated with relative risk reduction for type 2 diabetes by 7–12%, independently of MASLD or intermediate hyperglycemia at baseline. Given that this study lacked data on liver fat content, it remains unclear whether the effect of weight loss on reduced risk of type 2 diabetes was a direct effect or mediated by decreased liver fat.
We used different approaches to identify MASLD. However, MASLD may be present without elevated ALT (22), and ∼3.5% children with biopsy specimen–proven MASLD have been shown to have normal ALT (39). In the current study, this proportion is unknown, and it also remains unclear to which extent MASLD in the absence of elevated ALT increases diabetes risk. Independently of divergent perspectives on the use of ALT to indicate pediatric MASLD, this study showed that elevated transaminases, either taken at a single point in time or repeatedly measured, are associated with an increased risk for type 2 diabetes. Moreover, the association was consistent even when different ALT cutoffs were applied. Given the universal availability of ALT, the selection bias is minimized. Hence, finding elevated ALT in obesity clinical practice should raise clinicians’ awareness of the individual future risk for type 2 diabetes. On the other hand, given the high accuracy of the MASLD diagnosis code in Swedish specialized care (14), the internal validity of the study is strengthened. Nevertheless, MASLD is still often underdiagnosed in practice, and not all patients with probable MASLD undergo further evaluation.
Therefore, there are limitations to be acknowledged. First, despite using two approaches to identify MASLD, exposure misclassification is possible. Yet, we attempted to quantify for the misclassification and found a persistent association between MASLD and type 2 diabetes after correction for misclassification.
Second, because histopathology and imaging data were lacking, we could not assess the spectrum of MASLD and how it is associated with type 2 diabetes.
Third, to ascertain type 2 diabetes, we used diagnosis codes and antidiabetes medications from national registers. This method could not identify patients with type 2 diabetes managed in primary care without any antidiabetes medications. However, the number of these patients, if any, is likely to be very few because lifestyle modifications only are recommended for prevention, and not treatment, of type 2 diabetes in Sweden. Moreover, outcome misclassification is possible when antidiabetes prescription was used as a proxy for type 2 diabetes.
Fourth, using data of real-life obesity treatment results in each patient having a different frequency of follow-up, depending on the individual needs and the treatment centers. Detection bias, therefore, is possible.
Lastly, genes related to MASLD and youth-onset type 2 diabetes were unavailable in our data, whereas MASLD and type 2 diabetes have been indicated to share genetic architecture in European adult populations (40). As an attempt to adjust for heredity, parental type 2 diabetes was included as a covariate. Nonetheless, an advantage of using data on a nationwide cohort was the generalizability of our findings across geographical areas and migrant background. Moreover, because the national registers cover all people residing in Sweden, loss to follow-up is minimal.
In conclusion, children with obesity and MASLD, based on ALT or diagnosis code, have a higher risk for developing youth-onset type 2 diabetes compared with their peers with obesity alone and the general population comparators. MASLD interacts synergistically with intermediate hyperglycemia to dramatically increase the risk for type 2 diabetes. An optimal response in pediatric obesity treatment reduces the risk, regardless of MASLD.
This article contains supplementary material online at https://doi.org/10.2337/figshare.26999779.
Article Information
Acknowledgment. The authors extend their sincere thanks to all pediatric obesity centers contributing to the BORIS register.
Funding. This work was supported by the Freemason Foundation for Children’s Welfare in Stockholm, the Center for Innovative Medicine (CIMED) (grant no. FoUI-988950), the foundation of Sällskapet Barnavård, and Anna-Lisa & Arne Gustafsson foundation.
Duality of Interest. P.D. reports a leadership or fiduciary role on scientific/medical committee: Member of the steering committee for the Swedish Childhood Obesity Treatment Register, chairman of a working group developing the Swedish national guidelines for pediatric obesity treatment, and secretary of the Swedish Childhood Obesity Association. C.M. reports consulting fees from Novo Nordisk, Rhythm, Oriflame Wellness, DeFaire Medical, and Evira AB; honoraria for lectures from Novo Nordisk, Oriflame Wellness, and AstraZeneca; payment for expert testimony from Novo Nordic Foundation and Rhythm; leadership or fiduciary role on scientific/medical committee: board member of European Society for Paediatric Endocrinology (ESPE) Obesity working group, board member of the Swedish Pediatric Obesity Society, and Register holder for the Swedish Childhood Obesity Treatment Register. E.H. reports commissioned research for Novo Nordisk (2023), but not for the current study; honoraria for lectures from Novo Nordisk; a leadership or fiduciary role on scientific/medical committee: Member of the steering committee for the Swedish Childhood Obesity Treatment Register and secretary of Obesity COBWEB (COllaBoration in health economic modelling of OverWEight and oBesity). No other potential conflicts of interest relevant to this article were reported.
Author Contributions. R.R.P. wrote the first draft of the manuscript. R.R.P. performed statistical analysis. All authors edited, reviewed, and approved the final version of the manuscript and were involved in the interpretation of the results. R.R.P., C.M., and E.H. conceptualized the study. E.H. performed data acquisition. R.R.P and E.H. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Prior Presentation. Parts of this study were presented in abstract form at the 54th Annual Meeting of the European Society for Paediatric Gastroenterology, Hepatology and Nutrition (ESPGHAN), Copenhagen, Denmark, 22–25 June 2022.
Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Steven E. Kahn and Jonathan E. Shaw.