The prevalence of obesity has rapidly increased among both the general population and patients with diabetes.^1-3 Since the 1970s, the percentage of individuals with obesity has increased from 3.2% to 10.8% among male adults and from 6.4% to 14.9% among female adults.¹ The situation is worse among patients with type 2 diabetes mellitus (T2DM), with 85.2% of them being overweight or obese.² Patients with type 1 diabetes mellitus (T1DM) are also experiencing an upsurge in the prevalence of overweight or obesity.³ A cohort of children with T1DM showed a sharp increase in both overweight patients, from 28.6% to 47%, and obese patients, from 3.4% to 22.7%, over a period of 18 years.³ On the other hand, the proportion of underweight adults in the general population has decreased, from 13.8% and 14.6% to 8.8% and 9.7% among male and female, respectively.¹ Underweight children accounted for 5.7% of children with newly diagnosed T1DM in Poland from 2006 to 2017.⁴ However, despite the decreasing burden of underweight people, this population composes a considerable portion of total residents in the poorest regions of the world.¹

Large-scale prospective studies on body mass index (BMI) and all-cause mortality in the general and diabetic populations found J- or U-shaped relationships, indicating increased mortality risks in both underweight and obese subjects.^5-7 While BMI values corresponding to the nadir of the mortality risk curve were within the range of normal weight in some cohort studies,^5,6 several meta-analyses have demonstrated a protective effect of being overweight or mildly obese, which is referred to as the “obesity paradox.”^8-10 Prior reports on BMI and mortality among patients with T1DM are limited, and many of the existing studies omitted the assumption of non-linear association, drawing a hasty conclusion of a non-significant or inverse relationship.^11,12 Inadequate evidence eventually leads to uncertainty about consistent weight management in patients with T1DM.

To our knowledge, no previous systematic review has been performed on the relationship between BMI and mortality risk among patients with T1DM. Therefore, this meta-analysis aimed to evaluate the risk of all-cause mortality in each BMI category within a population with T1DM.

Data search and selection

English-language literature published before July 25, 2022, was searched in PubMed, Embase, and the Cochrane Library databases using the search terms described in Supplementary Table 1. References from the identified articles were manually searched. The following criteria were required for inclusion in this systematic review: (1) a prospective or retrospective cohort study; (2) reported results for patients with T1DM; (3) included analysis of the hazard ratio (HR) for all-cause mortality based on BMI; and (4) included data for each BMI category, including underweight, normal-weight, and obese individuals. Studies that recorded BMI only as a continuous variable or had no data for baseline characteristics were excluded.

Two independent investigators (HNJ and YKC) reviewed the articles to determine whether they satisfied the inclusion criteria. Disagreements were resolved through consensus. We conducted the present meta-analyses according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guideline (Appendices 1, 2). We used the Population, Intervention, Comparison, and Outcome (PICO) protocol to formulate research questions and define the scope of the study (Appendix 3).

Data extraction and quality assessment

The primary outcome was all-cause mortality. Extracted information included the author’s name, years of enrollment and publication, study location, study name or dataset, study design, criteria for T1DM, exclusion criteria, number of participants, follow-up duration, BMI categories, adjusted covariates, ratio of male to female, and other baseline data such as age and diabetes duration. For each BMI category, we also gathered data on the number of patients and events, along with adjusted HR and 95% confidence interval (CI). The Newcastle-Ottawa Scale was used to evaluate the quality of individual studies (Supplementary Table 2).¹³

Data integration and statistical analysis

Following the criteria of the World Health Organization, BMI categories were as follows: underweight <18.5 kg/m², normal weight 18.5 to <25 kg/m², overweight 25 to <30 kg/m², and obesity ≥30 kg/m².14 The data of studies using other categories were incorporated into the above scheme based on a difference <2 kg/m².

The pooled estimates of HRs and 95% CIs for each BMI category were calculated with reference to the normal-weight group. A random-effects model was used to combine the studies, and the outcomes were visualized using forest plots. Heterogeneity between studies was assessed with I² statistics and the Cochrane Q test.¹⁵ I² values of 25%, 50%, and 75% were considered to indicate low, moderate, and high heterogeneity, respectively.¹⁵ A Q-statistic P<0.1 implied the presence of heterogeneity.¹⁵

The R statistical software version 4.2.0 (The R Foundation for Statistical Computing) was used for statistical analyses. Statistical significance for the summary estimate was indicated by a two-tailed P<0.05.

Study selection

From 3,910 citations identified through a literature search, three studies involving 23,407 patients with T1DM were eligible for the meta-analysis (Fig. 1).^16-18 Table 1 summarizes the outlines and baseline characteristics of the selected studies. The research conducted by Conway et al.¹⁶ was the Pittsburgh Epidemiology of Diabetes Complications Study (EDC). Studies by Vestberg et al.¹⁷ and Dahlström et al.¹⁸ were based on data from the Swedish National Diabetes Registry (NDR) and the Finnish Diabetic Nephropathy (FinnDiane) study, respectively.

Baseline characteristics

All included studies were performed with a prospective cohort design and included adults with T1DM from Western countries.^16-18 The diabetes duration of participants ranged from 19 to 24 years. The mean or median follow-up duration was dissimilar among the studies, with the shortest duration of 8.5 years reported by Vestberg et al.¹⁷ and the longest duration of 18.2 years reported by Conway et al.¹⁶ Patients were recruited around the year 2000 by Vestberg et al.¹⁷ and Dahlström et al.,¹⁸ while Conway et al.¹⁶ commenced enrollment 10 to 15 years earlier. The baseline age was approximately 10 years younger in the study by Conway et al.¹⁶ in comparison with the others.^17,18 Vestberg et al.¹⁷ and Dahlström et al.¹⁸ excluded patients with previous myocardial infarction and unknown renal status, respectively.

Conway et al.¹⁶ classified risk factors into the following three groups: complications, biological risk factors, and socioeconomic and lifestyle risk factors. The risk factor groups were analyzed separately with different adjusted variables, and the complication model was included in this meta-analysis. Following exclusion of non-responders, 465 of 655 study participants were included in the complications group. The lowest BMI category was defined as <20 kg/m² by Conway et al.,¹⁶ which was classified into the underweight group for pooled analysis. Meanwhile, Vestberg et al.¹⁷ divided patients with obesity into two subgroups of BMI 30–35 kg/m² and BMI >35 kg/m². Considering the conflicting trends for mortality risk between these two subgroups (HR, 0.86 vs. 1.47),¹⁷ patients in each subgroup were regarded as sourced from separate studies when calculating the pooled risk.

Mortality risk according to BMI

The BMI categories with the lowest and highest mortality rates in the study by Conway et al.¹⁶ were the normal-weight and obese groups, respectively. On the contrary, the overweight and underweight groups exhibited the smallest and greatest risk, respectively, in the other studies.^17,18 The pooled analysis demonstrated that underweight increased the risk of all-cause mortality by 3.4 times compared to normal weight (95% CI, 1.67 to 6.85) (Table 2, Fig. 2). The overweight and obese groups showed a tendency toward lower (HR, 0.90; 95% CI, 0.66 to 1.22) and higher (HR, 1.36; 95% CI, 0.86 to 2.15) mortality risks, respectively, compared to the normal-weight group, although the difference was not statistically significant. The results of the I² statistics and Cochrane Q test suggested between-study heterogeneity.

A meta-analysis of longitudinal cohort studies evaluating mortality risks among BMI categories in patients with T1DM demonstrated a significantly increased risk of mortality in the underweight group compared to the normal-weight group. Meanwhile, the included studies showed heterogeneous results for the overweight and obese groups across the studies, resulting in non-significant differences in the pooled analysis.

Underweight as a risk factor for higher mortality has been supported by numerous studies.^9,19 An analysis of individual-level data from 239 prospective studies with healthy people showed that the underweight group, with a BMI range of 15 to 18.5 kg/m², displayed a 51% greater risk of mortality than the reference group with a BMI range of 20 to 25 kg/m². This risk was between those in obese groups of grade 1 (BMI, 30 to 35 kg/m²) and grade 2 (BMI, 35 to 40 kg/m²).¹⁹ This was in agreement with a meta-analysis of patients with T2DM, which showed decreased risk of all-cause mortality in individuals with BMI up to 28 kg/m².⁹ Correspondingly, weight loss >10% in patients with T2DM led to 2.27 times higher risk of all-cause mortality compared to weight maintenance.²⁰

The current analysis indicated that the increased mortality risk in the underweight group, as shown by previous research enrolling the general population or patients with T2DM, also applies to patients with T1DM. The lack of significantly increased mortality risk of the underweight group in the study by Conway et al.¹⁶ may have resulted from the low discriminatory power due to the lowest number of total deaths among the included studies. One possible explanation for the higher mortality from underweight in patients with diabetes is a hyperglycemia-induced exacerbation of protein catabolism and oxidative muscle damage, leading to restricted mobility and frailty.²¹ It is also possible that other comorbidities in underweight individuals with diabetes may be linked to increased mortality, such as malignancy, gastrointestinal motility disorders, infection, stroke, dementia, or periodontal disease.²²

Although there is a clear link between severe obesity and a higher mortality risk, a conclusion has not been reached on the risk among overweight or mildly obese subjects. Several large-scale prospective studies have reported increased mortality in the overweight and obese general population.¹⁹ Flegal et al.⁸ reported in a meta-analysis that people with BMI >35 kg/m² had a significantly higher all-cause mortality relative to the normal-weight group. However, the mortality risk was not increased in individuals with BMI 30 to 35 kg/m², and overweight subjects exhibited fewer deaths.⁸ Meanwhile, studies in patients with T2DM generally revealed a reduced mortality risk among overweight patients compared to those with normal weights.^9,10 On the other hand, obese individuals, referring to those with BMIs of >30 kg/m², showed conflicting results for mortality among meta-analyses.^9,10

Similar to previous mortality studies with other populations, studies of patients with T1DM exhibited heterogeneous mortality risks in the overweight or obese groups. Notably, the study by Conway et al.¹⁶ displayed a particularly higher mortality risk in the overweight and obese groups compared to the other two studies.^17,18 This may have resulted from a younger baseline age of around 10 years in the study by Conway et al.¹⁶ Overweight and obese people were found to have a higher risk of death at a younger age in numerous studies.^10,19 Several cardiovascular risk factors like dyslipidemia or hypertension accumulate with aging, independent of excess weight.¹⁰ Therefore, the adverse effects of obesity on mortality may be more obvious in younger individuals with fewer risk factors.^10,23 Furthermore, loss of muscle mass, a possible cause of underweight-related mortality, is infrequent at a young age.²¹ Meanwhile, patients were enrolled in the 1980s in the study by Conway et al.¹⁶ compared to in the 2000s in the other studies.^17,18 Considering a remarkable decrease from 1980 to 2000 in deaths caused by cardiovascular disorders,²⁴ which were the most common cause of death in overweight and obese patients with T1DM,¹⁸ underdeveloped treatments may have contributed to greater mortality in the excess weight groups in the study by Conway et al.¹⁶ Additionally, Conway et al.¹⁶ demonstrated a survival benefit of BMI increase during the first 10 years, supporting the benefits of a certain degree of weight gain.^17,18 Differences in follow-up and diabetes duration among the studies may also have played a role in the discordant mortality risks, though their connections are yet to be elucidated.

All the studies selected for this meta-analysis consisted of Western populations.^16-18 Initially, this systematic review included a study by Lee et al.,²⁵ which investigated the BMI-stratified mortality risk of patients receiving insulin treatment using Korean National Health Insurance Service datasets. Although the study by Lee et al.²⁵ presented separate mortality data for patients with T1DM, it was excluded because there were no baseline data limited to patients with T1DM and it was considerably heterogeneous compared to the other included studies—for example, due to the use of different BMI categories and much shorter follow-up duration.^16-18,25 Moreover, T1DM was operationally defined by Lee et al.²⁶ based on the International Classification of Disease 10th revision (ICD-10) code system. This definition might unintentionally involve many patients with T2DM having a low insulin secretion ability. Nevertheless, Lee et al.²⁵ also described the highest mortality risk in the underweight group, and the pooled analysis including the data of Lee et al.²⁵ showed a 2.83 times greater risk in the underweight group than in the normal-weight group (Supplementary Table 3, Supplementary Fig. 1A). The mortality risk in the overweight and obese groups was not significantly different from the normal-weight group in the pooled data (Supplementary Fig. 1B and C). Considering the greater fat composition and risk for metabolic disorders in Asians compared to the Western population with similar BMI values,²⁷ a prospective mortality study of Asian patients with T1DM is required.

A principal limitation of this meta-analysis was the substantial level of heterogeneity among the selected articles. This was attributable to the different investigation years, study designs, definitions of variables, and population characteristics. Even though sensitivity analysis is required to reduce the effects of confounding factors, this was not performed due to the small number of included studies. Alternatively, a random-effects model was used to calculate the pooled estimate, considering variability. Descriptive analysis was also performed as thoroughly as possible based on differences between the studies. Second, the only anthropometric measure used in this study was BMI, which does not discriminate between adiposity and muscularity.²⁸ Third, the impact of weight change over the follow-up period could not be examined due to inadequate data. Last, outcomes other than all-cause mortality, such as cardiovascular diseases, could not be evaluated. Despite these limitations, our report is the first systematic review of the relationship between BMI and mortality in patients with T1DM. The limitations of this study stem primarily from the small number of available longitudinal studies on T1DM, which supports the need for more research in this area.

In conclusion, underweight patients with T1DM have a significantly higher mortality risk than their normal-weight counterparts. Overweight and obese patients showed heterogeneous risks among cohort studies, but more recent reports suggested a reduced risk in overweight patients. Further exploration is needed to determine the associations between mortality and various anthropometric measurements that directly reflect body composition in patients with T1DM.

Supplementary materials can be found online at https://doi.org/10.7570/jomes22061.

The authors declare no conflict of interest.

This study was supported by the 2022 JOMES Research Grant (grant no. KSSO-J-2022001) of the Korean Society for the Study of Obesity. The authors thank the Asan Medical Library (University of Ulsan College of Medicine, Seoul, Republic of Korea) for providing the search terms for the systematic review and eWorldediting (www.eworldediting.com) for the English language review.

Study concept and design: YKC; acquisition of data: HNJ and YKC; analysis and interpretation of data: HNJ, CHJ, and YKC; drafting of the manuscript: HNJ and YKC; critical revision of the manuscript: CHJ and YKC; statistical analysis: SK and YKC; obtained funding: YKC; and study supervision: CHJ and YKC.