Impact of Parental or First-Degree Family History of Diabetes on Diabetes Incidence and Progression During Long-term Follow-up in the Diabetes Prevention Program Outcomes Study

Abstract

OBJECTIVE

To determine the effects of first-degree family history of diabetes on diabetes incidence in Diabetes Prevention Program (DPP) and Diabetes Prevention Program Outcomes Study (DPPOS) participants.

RESEARCH DESIGN AND METHODS

In the DPP, adults with prediabetes were randomized to an intensive lifestyle intervention, metformin, or placebo and followed for incident diabetes. On study completion 88% of eligible DPP participants reenrolled in DPPOS for long-term follow-up. The present analysis includes all 3,072 participants with family history information through DPPOS, with a median follow-up of 21 years (1,975 had parental history of diabetes [PH] [312 biparental, 947 maternal, 716 paternal], 226 had only sibling history [SH], and 871 denied any family history). The primary outcome is incident diabetes based on American Diabetes Association criteria, with adjustment for demographic and clinical variables, DPP randomization arm, and polygenic risk score (PRS).

RESULTS

Adjusted hazard ratio (HR) was 1.21 (95% CI 1.06, 1.38) for any family history, 1.19 (1.04, 1.35) for PH, and 1.15 (0.91, 1.44) for SH. Biparental history conferred greater hazard (HR 1.44 [95% CI 1.22, 1.69]) than maternal (1.22 [1.08, 1.38]) or paternal (1.22 [1.08, 1.39]) diabetes history alone. PRS explained 32% of the association of any family history with diabetes risk.

CONCLUSIONS

PH increased type 2 diabetes risk after DPP treatment group was controlled for. That effect was only partially explained by PRS, suggesting that rare gene variants, familial, and environmental factors may contribute to type 2 diabetes risk in people with prediabetes.

Graphical Abstract

Introduction

Diabetes, a major public health problem that currently affects ∼37 million adults in the U.S. and 537 million people worldwide, is a leading cause of blindness, amputation, and chronic kidney disease (1–3). Diabetes confers a two- to fourfold increase in the risk for cardiovascular disease (3). The health care costs associated with diabetes amounted to USD 412.9 billion in the U.S. in 2022 (4). Global diabetes prevalence is projected to reach 783.2 million by 2045, with the steepest increase occurring in low- and middle-income regions (2). Type 2 diabetes accounts for 90%–95% of the diabetes burden and is preceded by an intermediate stage of prediabetes (impaired fasting glucose and impaired glucose tolerance [IGT]). Approximately 98 million adults in the U.S. have prediabetes, and most are likely to develop type 2 diabetes over time (1). The risk factors for type 2 diabetes include overweight/obesity, physical inactivity, family history, ethnic and racial background, history of hypertension or dyslipidemia, and conditions associated with insulin resistance, including acanthosis nigricans and metabolic dysfunction–associated steatotic liver disease (5). Women with a history of polycystic ovary disease or gestational diabetes mellitus are at an increased risk of developing type 2 diabetes (5).

Valuable insights into family history as a risk factor for type 2 diabetes emerged from twins and other cohort studies. Those studies have clearly established type 2 diabetes as a heritable disorder (6–8). The concordance rate in monozygotic twins approaches 80%, and the lifetime risk of type 2 diabetes among offspring and siblings of affected patients is ∼40% (6,7). For individuals with insulin resistance and biparental history, the diabetes risk approaches 80% (8). Cross-sectional observations from the Nurses’ Health Study cohort (N = 73,227) showed that having at least one first-degree family member with diabetes was associated with a twofold higher relative risk of type 2 diabetes compared with having no family history of diabetes (9). The increased risk of type 2 diabetes associated with family history probably entails heritable and nongenetic components. The latter include maternal intrauterine environment, dietary and physical activity habits, socioeconomic status, air quality, and shared microbiome (10–17).

Previous reports on the effect of family history on diabetes risk were derived mostly from cross-sectional studies of individuals who were not receiving any interventions to prevent diabetes. Would the impact of family history on diabetes risk be discernible in people with prediabetes and multiple risk factors for diabetes who are receiving interventions for diabetes prevention? The prospective Diabetes Prevention Program (DPP) provided a unique opportunity for addressing that question (18). In the DPP, adults at high risk for type 2 diabetes were randomized to intensive lifestyle intervention (ILI), metformin, or placebo and followed for progression to diabetes (8). On completion of the study, participants in the original DPP randomization groups were offered lifestyle training. In the Diabetes Prevention Program Outcomes Study (DPPOS), surviving DPP participants were reenrolled for long-term follow-up, with analysis of outcomes by intention to treat based on their original DPP assignment. In a previous report from the DPP, a polygenic risk score (PRS) based on 34 type 2 diabetes–associated gene variants predicted progression to diabetes (19).

Parental history probably embodies the comprehensive genetic landscape for a given individual, including known and unknown gene variants (20). Pragmatically, family history is more accessible to clinicians than genotyping information and has the potential to capture exposure to environmental risk factors. Based on the foregoing considerations, we investigated the role of parental and nonparental first-degree family history of diabetes in the risks of progression from prediabetes to diabetes and progression of diagnosed diabetes among the multiethnic DPP/DPPOS participants. We examined the extent to which the effects of family history on glycemic outcomes were mediated by clinical and genetic factors. Further, we assessed the interactions of race and ethnicity and DPP treatment arm (ILI, metformin, or placebo) with the effects of family history on diabetes risk. We found that a family history of diabetes (vs. no family history) conferred a greater risk of progression to type 2 diabetes; the risk was highest for biparental history of diabetes, followed by single-parent history and sibling history, respectively. Thus, the magnitude of diabetes risk is related to the degree of closeness of the affected family members. Our findings could improve risk stratification and inform more targeted approaches to diabetes prevention.

Research Design and Methods

Participants and Study Design

The DPP (clinical trial reg. no. NCT00004992, ClinicalTrials.gov) was a clinical trial in which adults with prediabetes were randomized to ILI or blinded treatment with metformin or placebo from 1996 until 2001 and followed for incident diabetes. The trial was performed at 27 centers involving 3,234 participants who met eligibility criteria. Eligible participants were age ≥25 years, with BMI ≥24 kg/m² (≥22 kg/m² in Asian Americans), fasting plasma glucose (FPG) 5.3–6.9 mmol/L (≤6.9 mmol/L at American Indian sites), and 2-h plasma glucose (2-hPG) 7.8–11.0 mmol/L during a 75-g oral glucose tolerance test (OGTT). The differential eligibility criteria were based on the epidemiology of diabetes in Asian American and American Indian adults (18). All participants had IGT and high-normal to impaired fasting glucose at baseline, according to American Diabetes Association (ADA) criteria. Individuals with a history of diabetes and those taking medications that alter glucose tolerance were excluded. Details of the DPP study methods have previously been published (21).

On termination of the DPP, all participants were offered ILI in a group format and invited to join the DPPOS, regardless of diabetes status. Approximately 88% of eligible participants in each DPP treatment group joined DPPOS, which was initiated in September 2002 (clinical trial reg. no. NCT0038727, ClinicalTrials.gov). Long-term data from the DPPOS have been analyzed by intention to treat based on the original DPP treatment group (22). The present analysis includes 3,072 participants who provided family history information and follow-up data through DPPOS, with median follow-up time 21 years.

The institutional review boards at each participating center approved the DPP and DPPOS study protocols, and written informed consent was obtained from all participants prior to initiation of the study. An independent data safety monitoring board, appointed by the primary funding agency (National Institute of Diabetes and Digestive and Kidney Diseases [NIDDK]), provided study oversight.

Ascertainment of Family History of Diabetes

Family history of diabetes was ascertained with a standard questionnaire completed during baseline visits. Participants reported a history of diabetes in their mother and father separately and indicated each parent’s year of birth, age at diabetes diagnosis, and vital status (alive or dead). Information on year of death was collected from participants with deceased parents. The participants were also asked to provide information on the number of their natural brothers and sisters and indicate how many had been diagnosed with diabetes. Using the information obtained, we classified participants as having parental history of diabetes (PH) (subclassified into maternal, paternal, or biparental), sibling history (SH) but not parental history of diabetes, any first-degree history of diabetes (parental and sibling), and no family history of diabetes (No-FH).

Covariates

The covariates assessed at baseline that were used as adjustment variables in our models included age, sex, race and ethnicity, education, treatment arm (placebo, metformin, or ILI), BMI, systolic blood pressure, triglycerides, PRS, FPG, and 2-hPG. BMI was calculated as weight in kilograms divided by the square of height in meters. For the biochemical analyses, venous blood was collected and processed at each DPP/DPPOS clinic using a standardized manual of operations. Serum and plasma samples were stored briefly at –20°C at local sites and then shipped in batches on dry ice to the central biochemical laboratory for measurements.

Genotyping and Construction of PRS

The PRS was calculated using the EUR 1000 Genomes HapMap3 linkage disequilibrium reference files (23). Posterior weights were used to calculate the PRS in the Mass General Brigham (MGB) Biobank with the PLINK --score function. To account for PRS variability in our multiancestry cohort, a modified PRS adjustment strategy was implemented based on previously published methods (24,25). Briefly, a linear model of each disease-specific PRS was fitted against genetic ancestry probabilities. Adjusted PRS were then calculated as the residual between the predicted and actual PRS in the entire data set (26). This adjusted PRS was then standardized to a normal distribution, with a mean of 0 and variance of 1. The effects of this standardized score can be interpreted as a hazard ratio (HR) corresponding to 1 SD difference in the PRS.

Outcomes

The primary outcome was progression to diabetes. The secondary outcome was progression of diagnosed diabetes. FPG was measured every 6 months and an OGTT was performed annually until diabetes diagnosis, based on ADA criteria (FPG ≥7 mmol/L or OGTT 2-hPG ≥11.1 mmol/L, confirmed with a second test within 6 weeks) (16). Diabetes progression was defined as reaching HbA_1c level ≥7% in participants with incident diabetes. In the DPP, participants who progressed to diabetes remained on their randomized interventions (ILI, metformin, placebo) until hyperglycemia reached HbA_1c level ≥7%. As the ADA Standards of Care in Diabetes recommend an optimal treatment target of HbA_1c <7%, all participants with diagnosed diabetes whose HbA_1c reached ≥7% were eligible for diabetes management outside the study. Thus, the HbA_1c threshold level of ≥7% represents the point of deployment of additional medications for diabetes management by community physicians.

Statistical Analysis

A total of 3,072 participants (95% of the DPP cohort) with family history information were included in the primary outcome analysis (incident diabetes), and 2,795 (86% of the DPP cohort) were included in the secondary outcome analysis (progression to HbA_1c ≥7%). Baseline characteristics of participants with self-reported No-FH, PH, or SH were compared. Demographic variables selected for this analysis included age, sex, self-reported race and ethnicity (American Indian, Asian American, non-Hispanic Black, Hispanic, non-Hispanic White), education (0–12 years, >12 years), and income (USD <35K, USD 35K to <75K, USD ≥75K, refused to answer). Clinical measures included BMI, waist circumference, HbA_1c, systolic and diastolic blood pressure, HDL cholesterol, triglycerides, FPG, 2-hPG, insulin-to-glucose ratio (IGR), and PRS. Kaplan-Meier curves and log-rank tests were used to analyze the cumulative incidence of the primary and secondary outcomes by family history groups.

Cox proportional hazards models with robust estimates for the SEs and the Efron method for tie handling were used to assess the univariate and multivariate association of family history with incident diabetes. In multivariate analyses we also assessed whether baseline clinical measures were significant predictors of the outcomes. Multicollinearity was assessed using the variance inflation factor, tolerance, and comparison of eigenvalues and condition index values. Through visual examination of Kaplan-Meier curves we assessed the Cox proportional hazards assumption. For significant variables in the Cox proportional hazards model for dichotomized family history, a mediation analysis was conducted. The mediation “adjusted” model included family history defined as a dichotomous variable (any first-degree family history/No-FH), the indicated mediator variable, and demographic variables. The mediation “crude” models included family history defined as a dichotomous variable (any first-degree family history/No-FH) and demographic variables. Each potential mediator was assessed independent of the other in separate models, followed by assessment together in a single model. The proportion of the family history association mediated was calculated as 1 − β_adjusted/β_crude (27), where β_adjusted is the coefficient for family history in the Cox proportional hazards model including the potential mediator and β_crude is the coefficient for family history in the Cox proportional hazards model not including the potential mediator. Finally, tests were conducted of family history effect heterogeneity for the significant log-rank tests outcomes among treatment arms (ILI, metformin, placebo) and subgroups age, sex, and race and ethnicity. Wald tests were used to assess interaction of family history with demographic variables and treatment arm. All statistical analyses were done in 2024–2025, with R version 4.2.1 and SAS version 9.4.

Results

Baseline Characteristics of Participants

All 3,072 DPP participants with family history and follow-up data through DPPOS were included in the present analysis (median follow-up in these individuals 21 years [52,570 person-years]). Of these, 1,975 participants had PH, 226 had SH (but not PH), and 871 had No-FH. Mean ± SD age at baseline was 49.3 ± 10.1, 55.2 ± 11.3, and 51.7 ± 11.0 years for the PH, SH, and No-FH groups, respectively (P < 0.001). Mean baseline FPG and 2-hPG levels were similar for the three groups, as were BMI, waist circumference, and diastolic blood pressure and HDL cholesterol levels. The groups differed significantly in baseline HbA_1c, systolic blood pressure, triglycerides, and the distribution of race and ethnicity and education status (Supplementary Table 1).

Overall Glycemic Outcomes

During total follow-up (mean ± SD 17 ± 7 years), incident diabetes occurred in 59.6%, 57.5%, and 48.1% of participants in the PH, SH, and No-FH groups (P < 0.001), respectively, and progression of diabetes to HbA_1c levels ≥7% occurred in 29.7%, 22.3%, and 21.4% (P < 0.001) (Supplementary Table 2).

Impact of Family History on Incident Diabetes

Having a PH or first-degree family history of diabetes significantly increased risk of progression from prediabetes to diabetes and progression of diabetes to HbA_1c ≥7% (Figs. 1 and 2). In Cox regression models, HR for progression to diabetes was 1.29 (95% CI 1.15, 1.44) for any first-degree family history, 1.30 (1.16, 1.46) for PH, and 1.19 (0.97, 1.45) for SH (Supplementary Tables 1 and 3). Adjustment for demographic and clinical variables, DPP treatment arm, and PRS attenuated the HRs: 1.21 (1.06, 1.38) for any first-degree family history, 1.19 (1.04, 1.35) for PH, and 1.15 (0.91, 1.44) for SH (Table 1 and Supplementary Table 3). Having an SH alone was associated with 18% increased risk of incident diabetes, which was not statistically significant due to wide CIs (Table 1).

Figure 1.

Kaplan-Meier curves for the primary (progression of prediabetes to diabetes) and secondary (progression of diabetes to HbA_1c ≥7%) outcomes with nominal log-rank test P values comparing family history groups (No-FH, PH, and SH only).

Figure 2.

Degrees of family history risks of outcomes progression from prediabetes to diabetes and progression of diabetes to HbA_1c ≥7%. HRs and 95% CIs for family history effects on both outcomes. Results are compiled from separate models, described in Tables 1 and 2 and Supplementary Tables 2 and 3. Models include adjustment for demographic variables (age, sex, race and ethnicity, and education), PRS, clinical baseline variables (BMI, triglycerides, systolic blood pressure [systolic blood pressure], FPG), and treatment arm.

Table 1.

HRs and asymptotic 95% CIs for main effect family history^a on progression from prediabetes to diabetes and progression of diabetes

Model	Group/covariate	Prediabetes to diabetes	P*	Progression of diabetes to HbA1c ≥7%	P
Unadjusted†	PH	1.30 (1.16, 1.46)	<0.0001	1.34 (1.13, 1.59)	0.001
	SH only	1.19 (0.97, 1.45)	0.0952	0.95 (0.68, 1.32)	0.749
Demographics‡	PH	1.27 (1.13, 1.42)	<0.0001	1.24 (1.04, 1.47)	0.018
	SH only	1.18 (0.97, 1.44)	0.1039	1.03 (0.74, 1.43)	0.875
Demographics/PRS/clinical measures/treatment arm§	PH	1.19 (1.04, 1.35)	0.0089	1.25 (1.03, 1.53)	0.026
	SH only	1.15 (0.91, 1.44)	0.2350	1.04 (0.72, 1.52)	0.821
	PRS	1.41 (1.25, 1.60)		1.30 (1.08, 1.57)
	BMI	1.02 (1.01, 1.03)		1.03 (1.02, 1.04)
	Triglycerides	1.16 (1.11, 1.21)		1.21 (1.13, 1.30)
	Fasting insulin	1.00 (1.00, 1.00)		1.00 (1.00, 1.00)
	IGR	1.00 (1.00, 1.00)		1.00 (1.00, 1.00)
	SBP	1.00 (1.00, 1.01)		1.00 (0.99, 1.01)
	Lifestyle	0.79 (0.69, 0.90)		0.82 (0.67, 1.00)
	Metformin	0.86 (0.76, 0.99)		0.71 (0.59, 0.86)
Demographics/PRS/clinical measures (including FPG)/treatment arm¶	PH	1.19 (1.04, 1.36)	0.0093	1.24 (1.02, 1.52)	0.033
	SH only	1.18 (0.94, 1.47)	0.1473	1.11 (0.77, 1.60)	0.559

^aNo-FH (reference category), PH (maternal and/or paternal), and SH only. *Nominal P values, comparing the hazards of the given outcome, are shown relative to reference category No-FH. †Univariate model of main effect of family history. ‡Multivariate model with adjustment for demographic variables (age, sex, race and ethnicity, and education). §Multivariate model with adjustment for demographic variables (age, sex, race and ethnicity, and education), PRS, clinical baseline variables (BMI, triglycerides, fasting insulin, IGR, and systolic blood pressure), and treatment arm. ¶Multivariate model with adjustment for demographic variables (age, sex, race and ethnicity, and education), PRS, clinical baseline variables (BMI, triglycerides, systolic blood pressure, and FPG), and treatment arm. No adjustment was done for multiple comparisons, since the analyses are considered exploratory.

Impact of Family History on Progression of Diabetes to HbA_1c ≥7%

Having a PH significantly increased the risk of progression of diagnosed diabetes to HbA_1c ≥7%. In Cox regression models, HR for progression to HbA_1c ≥7% was 1.30 (95% CI 1.09, 1.54) for any first-degree family history and 1.34 (1.13, 1.59) for PH (Table 1 and Supplementary Table 3). Full adjustment for demographic and clinical variables, DPP treatment arm, and PRS attenuated the effects on progression of diabetes (HR 1.24 [95% CI 1.01, 1.51] for any first-degree family history and 1.25 [1.03, 1.53] for PH). Having SH alone was not significantly associated with progression of hyperglycemia to HbA_1c ≥7% (Table 1 and Supplementary Table 3).

Impact of Number of Affected Parents

Of the 1,975 DPP participants with a PH, 947 gave a maternal history and 716 a paternal history and 312 reported that both parents were affected. Biparental history conferred greater hazard for incident diabetes compared with maternal or paternal diabetes history alone in the fully adjusted model (Table 2 and Fig. 2). Biparental diabetes history also increased risk of early progression of diabetes to hyperglycemia of HbA_1c ≥7%, after adjustment for demographic, anthropometric, and clinical (BMI, DPP treatment arm, systolic blood pressure, triglycerides) variables and PRS (Table 2 and Fig. 2). The HRs for incident diabetes and progression of diabetes were comparable for participants with maternal versus paternal diabetes history (Table 2 and Fig. 2).

Table 2.

HRs (95% CI) for main effect family history^a and outcomes progression from prediabetes to diabetes and progression of diabetes

Model	Group	Prediabetes to diabetes	P*	Progression of diabetes to HbA1c ≥7%	P
Unadjusted†	Biparental	1.44 (1.22, 1.69)	<0.0001	1.53 (1.20, 1.94)	<0.001
	Maternal	1.22 (1.08, 1.38)	0.0010	1.29 (1.07, 1.55)	0.007
	Paternal	1.22 (1.08, 1.39)	0.0017	1.36 (1.12, 1.65)	0.002
Demographics‡	Biparental	1.38 (1.18, 1.63)	<0.0001	1.34 (1.06, 1.71)	0.015
	Maternal	1.19 (1.05, 1.34)	0.0047	1.19 (0.99, 1.43)	0.070
	Paternal	1.20 (1.06, 1.37)	0.0041	1.23 (1.01, 1.50)	0.036
Demographics/PRS/clinical measures/treatment arm§	Biparental	1.31 (1.09, 1.58)	0.0042	1.37 (1.03, 1.84)	0.032
	Maternal	1.15 (1.01, 1.32)	0.0407	1.26 (1.01, 1.56)	0.037
	Paternal	1.12 (0.97, 1.30)	0.1143	1.18 (0.94, 1.47)	0.150
Demographics/PRS/clinical measures (including FPG)/treatment arm¶	Biparental	1.42 (1.19, 1.70)	0.0001	1.46 (1.11, 1.92)	0.007
	Maternal	1.11 (0.97, 1.28)	0.1257	1.20 (0.97, 1.48)	0.099
	Paternal	1.10 (0.95, 1.27)	0.2025	1.16 (0.93, 1.44)	0.199

^aNo parental history (reference category), maternal history, paternal history, or biparental diabetes history. *Nominal P values comparing the hazards of the given outcome are shown relative to reference category No-FH. †Univariate model of main effect family history. ‡Multivariate model with adjustment for demographic variables (age, sex, race and ethnicity, and education). §Multivariate model with adjustment for demographic variables (age, sex, race and ethnicity, and education), PRS, clinical baseline variables (BMI, triglycerides, fasting insulin, IGR, and SBP), and treatment arm. ¶Multivariate model with adjustment for demographic variables (age, sex, race and ethnicity, and education), PRS, clinical baseline variables (BMI, triglycerides, systolic blood pressure, FPG), and treatment arm. No adjustment was done for multiple comparisons, since the analyses are considered exploratory.

Mediation Analysis

We examined significant mediators of the effects of family history of diabetes on outcomes of the study. The mediation analysis included PRS and baseline FPG, as these variables were significant in both outcome models for any first-degree family history. The proportion of the family history association mediated was calculated as described in Statistical Analysis. Our calculations showed that PRS explained 32.4% and 4.7% of the association of family history with risk of incident diabetes and progression of diabetes, respectively. Other mediators included FPG, fasting insulin, and IGR, each accounting for <10% of the effect of family history on risk of incident diabetes (Supplementary Table 4). We also examined the C-statistics in models with and without the PRS to assess its improvement of prediction of diabetes. The C-statistic was 0.668 without the PRS and 0.674 with PRS included, indicating only modest improvement in risk prediction with PRS versus with commonly known risk factors.

We found no significant interactions between the effects of family history on diabetes outcomes and subgroups defined by age, sex, race and ethnicity, or DPP treatment arm. Thus, pairwise comparisons of family history groups within the strata of treatment arms and age, sex, and race and ethnicity subgroups were not conducted.

Conclusions

In the multiethnic DPP/DPPOS population, PH was associated with increased risk of progression from prediabetes to diabetes after adjustment for demographic, socioeconomic, clinical, and genetic variables. Biparental history conferred a greater diabetes hazard than maternal or paternal diabetes history alone. Furthermore, risk of diabetes progression to HbA_1c ≥7% was significantly greater for participants with PH and increased further among participants with biparental history. In contrast, having only an affected sibling had a weaker effect on risk of progression from prediabetes to diabetes and no effect on risk of progression of hyperglycemia to HbA_1c ≥7%. DPP participants who developed diabetes remained on their randomized interventions until HbA_1c was ≥7%, which triggered referral for diabetes management by community physicians. That outcome (reaching HbA_1c level ≥7%), indicating transition to treatment intensification, was associated with PH but not family history.

Our findings are consistent with previous reports showing increased risks of diabetes associated with SH, PH, and biparental history of diabetes (7–9,28–33). However, there are some unique observations from our study. The magnitude of diabetes risk associated with PH in the present study is lower than the twofold or greater increased risk often reported for first-degree family history (28–30). Differences in study populations might explain part of the disagreement in the risk estimates. DPP participants had a narrow age range of 50.4 ± 10.6 years (mean ± SD), prevalent IGT, overweight/obesity, among other diabetes risk factors (18,21). Consequently, diabetes incidence was 11/100-person-years in the placebo arm (24). In contrast, the previous reports were obtained from a general population with a broader age range and lower risk burden for diabetes. In one such study, diabetes incidence rate was ∼5/1,000 person-years (30). Thus, our findings might reflect the additional impact of family history among people already at high risk of diabetes based on older age, higher BMI, and preexisting dysglycemia. A much greater impact of family history than was observed in the present study has been reported in study populations at lower baseline risk for diabetes (28–30). Furthermore, two-thirds of participants had been exposed to either ILI or metformin treatment during the DPP phase of the study (18). Previous reports on the effect of family history on incident diabetes typically involved populations that were not undergoing interventions to alter diabetes outcomes. Despite these differences, we found a significant risk attributable to family history in the present study.

Notably, we found that the risk for developing diabetes was increased to a similar magnitude regardless of parental sex, unlike previous reports showing higher risk for maternal versus paternal history (29–32). In one study maternal history increased diabetes risk in lean participants, whereas paternal history did so in individuals with overweight (32). In the report by Scott et al. (29), diabetes hazard was particularly high for individuals with a history of maternal diabetes diagnosed at a younger age (<50 years) (29), as was also reported from the Framingham Offspring Study (33) Thus, there appear to be complex interactions among the diabetogenic effects associated with family history, sex, parental age at diagnosis, and adiposity status of the index subject. Additionally, we observed that SH had a weaker effect on diabetes risk than PH. In the Danish registry-based study, the diabetes risks associated with parental or full sibling history were comparable and both were higher than that of half-sibling history (30). As we did not distinguish between full- and half-sibling status in our study, our analysis could have been affected by inadvertent inclusion of the latter. However, sensitivity analyses excluding those with SH yielded similar associations between PH and diabetes development.

In the present study, PRS explained 32% of the association of family history with risk of incident diabetes, higher than previous estimates (34,35). Across common chronic disorders (including type 2 diabetes), PRS explained on average 10% of the effect of first-degree family history on disease incidence (29). In the InterAct study, PRS (based on 35 type 2 diabetes–associated gene variants) explained only 2% of the association of family history with increased diabetes risk (30). In a recent report, first-degree family history conferred higher odds for diabetes compared with PRS (odds ratio 2.32 [95% CI 2.20, 2.32] vs. 1.75 [1.71, 1.79]), although no mediation analysis was presented (36).

Differences in study populations, number of gene variants included, demographic characteristics, and duration of follow-up might explain the disparate estimates of the mediation effect of PRS (37–39). Nonetheless, the data are consistent in showing that common variants identified so far in genome-wide association studies do not explain most of the variation in the risk of type 2 diabetes. It is plausible that the inclusion of rare and less frequent variants might improve risk estimates. However, based on C-statistics, the PRS added minimal value to commonly known clinical risk factors in predicting diabetes in the present study, as was also noted in a previous report from the DPP (19). Thus, clinical information and measures that are readily available in practice do provide adequate predictive information regarding type 2 diabetes risk without resorting to genetic testing.

Our present findings along with previous reports indicate that environmental factors, currently rare gene variants, and possibly epigenetic mediators might account for a substantial proportion of the association between family history and diabetes risk (19,29,34). PH integrates the effects of common, rare, and unknown gene variants and exposure to diabetogenic environmental risk factors, including behavioral, socioeconomic, aerosols, sleep habits, and endocrine-disrupting chemicals (11–17,35,40,41). Additionally, there is evidence that individuals in the same household share 12%–32% of their gut and oral microbiomes (17,41). Gut flora interact with the host energy metabolism, modulate bile acid availability, and influence signaling via fibroblast growth factor 19 and other pathways related to the development of obesity, insulin resistance, diabetes, and cardiometabolic disorders (41,42). Some of the environmental factors may also act via epigenetic mechanisms to increase diabetes risk (43). Insulin resistance and β-cell dysfunction predicted progression from prediabetes to diabetes in the DPP (44). The genetic and environmental determinants of the increased risk associated with family history of diabetes might exert their effects via induction of insulin resistance or β-cell dysfunction (8,44,45). Interestingly, the known diabetes gene variants are associated predominantly with β-cell survival and function rather than insulin resistance (20). In contrast, most of the environmental risk factors are linked to insulin resistance directly or via their obesogenic effects (8,9,11–17,35,40–42).

Our findings underscore the importance of obtaining a detailed family history in clinical practice, as the information could inform clinical recommendations for individuals with prediabetes and PH or SH. Timely diabetes prevention interventions would be desirable for such individuals. Furthermore, the increased diabetes risk conferred by biparental history could be integrated into risk models to improve early detection and targeted preventive efforts. Additionally, a case may be made for more intensive lifestyle (or earlier pharmacological) interventions in people with biparental history than for those at lower risk. However, the merits and efficacy of such a stepped intervention approach based on the number of parents affected have not been formally assessed in clinical trials.

Our study has several strengths and limitations. The prospective design, lengthy (21-year) follow-up, and rigorous ascertainment of diabetes outcomes are some of the strengths of our study. Additionally, the inclusion of participants from different racial and ethnic groups and with different educational and socioeconomic background, PRS, and treatment exposure (ILI, metformin, placebo) enabled a rigorous assessment of the interactions of these factors with our key findings regarding family history. Nonetheless, because the inclusion criteria for the overall DPP study included prioritization of higher BMI and high diabetes risk factor burden, the present study population cannot be considered representative of the general society. Thus, our findings are best interpreted as reflecting the impact of family history on diabetes outcomes in people already at considerable risk for type 2 diabetes. The self-reported nature of family history is a limitation. Another limitation is the relatively small number (n = 226) of DPP participants who gave an SH, which might explain the weaker effect observed in that subgroup. Moreover, as we did not verify the stated biological relationships, adopted participants or those with half-siblings or adopted siblings could have been inadvertently included in our analysis. Thus, our findings might have underestimated the effects of SH. Furthermore, the PRS data analyzed in the present study did not include additional type 2 diabetes–related gene variants discovered after the initial genotyping of DPP participants (19,20).

In conclusion, in the multiethnic DPP/DPPOS population with prevalent prediabetes, PH increased risk of progression to diabetes, an effect that was independent of exposure to DPP treatment and only partially explained by PRS. Thus, unknown genetic, nonheritable familial, environmental, and possible epigenetic factors contribute substantially to risk of progression from prediabetes to diabetes.

This article contains supplementary material online at https://doi.org/10.2337/figshare.29621537.

Funding Statement

S.D.-J. is supported, in part, by National Institutes of Health (NIH) grants R01 DK128129, R01 DK067269, and U01 DK048411. E.J.K. is supported by grant U01DK048489, L.D. by U01DK048489, P.S. by U01DK048443, and J.B.E.-T., in part, by grant K23 HL153774 from the National Institutes of Health. W.C.K. worked on this article as a paid consultant with the Biostatistics Center at The George Washington University supported by NIH grant U01DK048489. S.E.K. is supported by U01DK048413 and S.M., in part, by 2U01DK098246-06, 1U19AG078558-01, and 2U01DK094176-06 from the NIH. M.T. is supported by NIH grant U01DK048489. Research reported in this publication was supported by grants from the NIH under award nos. DK-55433 and DK-55564 and the NIDDK of the NIH under awards U01 DK048489, U01 DK048339, U01 DK048377, U01 DK048349, U01 DK048381, U01 DK048468, U01 DK048434, U01 DK048485, U01 DK048375, U01 DK048514, U01 DK048437, U01 DK048413, U01 DK048411, U01 DK048406, U01 DK048380, U01 DK048397, U01 DK048412, U01 DK048404, U01 DK048387, U01 DK048407, U01 DK048443, and U01 DK048400, through provision of funding during DPP and DPPOS to the clinical centers and the Coordinating Center for the design and conduct of the study and collection, management, analysis, and interpretation of data. Funding was also provided by the Eunice Kennedy Shriver National Institute of Child Health and Human Development, the National Institute on Aging, the National Eye Institute, the National Heart, Lung, and Blood Institute, the National Cancer Institute, the Office of Research on Women’s Health, the National Institute on Minority Health and Health Disparities, the Centers for Disease Control and Prevention, and the ADA. The Southwestern American Indian Centers were supported directly by the NIDDK, including its Intramural Research Program, and the Indian Health Service. The General Clinical Research Center Program, National Center for Research Resources, and U.S. Department of Veterans Affairs supported data collection at many of the clinical centers. Merck KGaA provided medication for DPPOS. DPP and DPPOS have also received donated materials, equipment, or medicines for concomitant conditions from Bristol-Myers Squibb, Parke-Davis, LifeScan, Health o meter, Hoechst Marion Roussel, Merck-Medco Managed Care, Merck & Co., Nike Sports Marketing, SlimFast Foods, and Quaker Oats. The Henry M. Jackson Foundation provided support services under subcontract with the Coordinating Center. The sponsor of this study was represented on the Steering Committee and played a part in study design, how the study was done, and publication.