Global Disability Burdens of Diabetes-Related Lower-Extremity Complications in 1990 and 2016

The International Diabetes Federation estimated that 1 in every 11 adults (451 million) had diabetes in 2017, and 1 in 10 (693 million) will have diabetes by 2045 (1). People with diabetes are at high risk of developing a range of complications, including cardiovascular, kidney, eye, and lower-extremity complications (2,3). Arguably, the most disabling of these are the lower-extremity complications of peripheral neuropathy, foot ulceration, and amputation (4,5).

Diabetes-related lower-extremity complications (DRLECs) typically first present as neuropathy (4,5). Neuropathy is the critical risk factor for developing foot ulceration, and foot ulceration is the critical risk factor for foot infection and amputation (4,5). It is estimated that up to 50% of people with diabetes have neuropathy (6,7), up to 34% will develop a foot ulcer in their lifetime (5), and ∼50% of those ulcers will become infected (8–10) and 20% amputated (5,10,11). DRLECs also account for up to 80% of global lower-extremity amputations, are a leading cause of hospitalization (12–14), and result in significant reductions in quality of life (4,15). Thus, DRLECs appear to cause considerable global disability burden.

While several narrative reviews have reported basic global epidemiology estimates for some DRLECs (4–6,16,17), only one article has reported a robust global prevalence estimate on the basis of systematic reviews (16). In this meta-analysis, the authors reported a pooled global foot ulcer prevalence of 4.6% in the diabetes population as well as reported foot ulcer prevalence estimates for five geographical regions and 33 countries (16). Otherwise, no article has systematically reported global, regional, or country prevalence estimates for the other individual DRLECs, including neuropathy and amputations.

Additionally, to our knowledge, only our group has reported an estimate of the global disability burden of DRLECs (18). In a brief published letter, we used data from the Global Burden of Diseases, Injuries, and Risk Factors Study (GBD) 2015 to report an overall estimate for the years lived with disability (YLDs) caused by aggregated DLRECs in 2015 (18). YLDs are an important population measure of disability burden (or nonfatal health loss), as they enable objective comparisons of disability burden of a condition across various populations or for various conditions within the same population (19). YLDs are calculated by multiplying the total numbers of people affected by a condition by the average severity of disability for that condition (disability weight) (19). However, in that letter, we did not report YLD estimates for individual DRLECs or estimates by sex, age, region, or country (18). Thus, robust data on the global, regional, and country prevalence and disability burden of DRLECs is still not accessible to decision makers. GBD presents a unique and robust opportunity to address this gap.

GBD is a multinational collaborative research project, which systematically integrates all available data to produce consistent, transparent, and up-to-date estimates on the disability and mortality burdens of most diseases and injuries affecting the world (19–23). Initiated in the early 1990s by the World Health Organization, and since 2010 independently funded by the Bill and Melinda Gates Foundation, it has grown into an international consortium of >3,600 researchers, with >200 articles published. GBD estimates are now arguably the most well-known, well-used, and well-cited burden of disease estimates and have had a profound impact on health policy and agenda setting throughout the world, especially as GBD has brought global attention to otherwise hidden or neglected diseases and injuries (19–23).

While estimates have been presented for diabetes in GBD capstone articles, the estimates for the disabling sequelae that make up the diabetes burden have not been described in detail, including DRLECs (19). Although it is apparent in these capstone articles that the global burden of diabetes is comparatively large and has markedly increased over recent decades (19), and it could be expected that the disease burden of DRLECs would have a commensurate increase, to date, global, regional, and national estimates and changes over time related to DRLECs remain hidden. Thus, this article uses GBD data with the primary aim to provide global prevalence and disability burden (YLDs) estimates for individual and overall DRLECs in 2016. Secondary aims are to further explore these 2016 estimates by sex, age, 21 regions, and 195 countries or territories and, over time, to compare these with 1990 estimates. Such estimates should enable health service providers and researchers to evaluate the comparative disease burden DRLECs have on their population of interest and, in turn, inform decisions about resource allocation to target cost-effective prevention and treatment at populations disproportionately affected.

This study was a secondary analysis of GBD 2016 data obtained from the Institute for Health Metrics and Evaluation (IHME), Seattle, Washington. IHME provided prevalence and YLD estimates for diabetes and DRLEC-related sequelae by sex, 20 age-groups, 21 regions, and 195 countries or territories for the years 2016 and 1990. Prevalence and YLDs were provided as counts, rates, and age-standardized rates with 95% uncertainty intervals (UIs). Results of this study were reported according to the Guidelines for Accurate and Transparent Health Estimates Reporting statement (24) (Supplementary Table 1).

Since 1990, GBD has been estimating global disease burden by quantifying the health loss from premature mortality and nonfatal disability (19–23). GBD attributes the health loss to hundreds of causes (diseases and injuries that cause death or disability), which are made up of thousands of sequelae (complications of those causes), using a mutually exclusive and exhaustive list of causes organized into a hierarchy of levels (19). Level 1 causes consist of three broad, overarching cause groups, such as noncommunicable diseases (19). These are disaggregated into >20 level 2 cause groups, such as diabetes, urogenital, blood, and endocrine diseases (19). These are further disaggregated into >100 level 3 causes, such as diabetes, and again into >200 level 4 causes, such as chronic kidney disease (CKD) due to diabetes (19). Finally, these level 4 causes are disaggregated into >2,000 sequelae, such as the DRLEC-related sequelae of diabetic neuropathy and amputation with treatment (19). Thus, the prevalence and YLD estimates for all the sequelae of a particular level 4 cause should sum exactly to the prevalence and YLD estimate of that level 4 cause, respectively, and so on up the hierarchy of levels.

GBD is now remeasured and republished at least every 3 years. For each new iteration, GBD provides new estimates not only for more recent years but also for all previous years as improvements in GBD methodology are implemented. These new estimates for previous years supersede any previously published result estimates for those years, including 1990 estimates. GBD 2016 estimated global prevalence, incidence, and YLDs for 328 causes and 2,982 sequelae, 21 regions, and 195 countries or territories during 1990–2016 (19,25). The extensive GBD 2016 methodology has been published in detail elsewhere (19,25). For the purpose of this article, we summarize the methodology used in the prevalence and YLD estimations for diabetes and its eight mutually exclusive sequelae: uncomplicated diabetes, moderate vision loss due to diabetes, severe vision loss due to diabetes, blindness due to diabetes, diabetic neuropathy, diabetic foot due to neuropathy, diabetic neuropathy and amputation with treatment, and diabetic neuropathy and amputation without treatment (19).

As a cause in GBD 2016, diabetes was defined as individuals with a fasting plasma glucose (FPG) >126 mg/dL (7 mmol/L) or being on treatment for diabetes (19). DRLECs were defined as the four mutually exclusive DRLEC-related sequelae of diabetes: 1) diabetic neuropathy, defined as patients with diabetes with diagnosed neuropathy with no current foot ulcer or previous major amputation (19); 2) diabetic foot due to neuropathy, defined as patients with diabetes with a current foot ulcer (19); 3) diabetic neuropathy and amputation without treatment, defined as patients with diabetes with previous major leg amputations (above or below the knee) without a prosthetic limb to ambulate (and instead require crutches or similar aid to move around); and 4) diabetic neuropathy and amputation with treatment, defined as patients with previous major leg amputations (above or below the knee) with a prosthetic limb to ambulate (19). Amputation was broken down into with and without treatment (i.e., with and without prosthesis) to more accurately estimate disability burden on the basis of studies that individuals with major amputation without a prosthesis had markedly worse quality of life and disability burdens than those with a prosthesis (19,26–28). Patients with multiple DRLEC-related sequelae were assigned the most severe of their DRLECs as their only DRLEC so as to be counted only once. Full case definitions and diagnostic criteria with associated ICD-9 and ICD-10 codes are presented in Supplementary Table 2. Shortened terms are used to denote these four sequelae for the remainder of this article as neuropathy, foot ulcer, amputation with prosthesis, and amputation without prosthesis, respectively.

In GBD, prevalence was estimated first for causes and then split into sequelae. For all included causes, GBD initially performed systematic reviews to identify all available data sources from published articles and unpublished registry data before 31 December 2016 (19). All accessible information identified on disease occurrence, natural history, and any sequelae on the basis of severity for a cause, such as diabetes, then passed a strict set of inclusion criteria and adjustments to maximize the comparability of the data, despite whether different collection methods or case definitions were used in the included studies over time. For diabetes, data from 806 prevalence articles were extracted and processed, including time, age, and sex splitting. A series of adjustments were then made if required; this involved crosswalking data inputs from studies with less desirable methods to data inputs from optimally conducted studies. Crosswalks among several different case definitions with different thresholds of FPG or glycated hemoglobin levels for diabetes were used on the basis of available data with individual records of the actual measurements. The adjusted input data were entered into a validated Bayesian meta-regression tool, DisMod-MR 2.1, to produce adjusted pooled prevalence estimates for diabetes for each age-sex-country-year combination (19). To predict estimates for countries with sparse or no available data, GBD uses a “wrapper” code that bases country estimates on the regional estimates from the geographical parent region for that country. Furthermore, the coefficients for the parent geographical region were also used to take advantage of the predictive power of the covariates in a meta-regression. In the case of diabetes, a range of covariates was systematically tested, and two covariates were chosen to further inform estimates for data-sparse countries on the basis of statistical fit (Akaike information criterion and adjusted R²): prevalence of obesity per country and lag-distributed income per capita (19).

For DRLEC-related sequelae, GBD included 89 prevalence articles for neuropathy, 43 for foot ulcer, and 12 for amputation. Separate DisMod-MR 2.1 models were used to produce estimates of the proportion of the diabetes population affected by each sequela. These proportion estimates were multiplied by the estimated prevalence of diabetes to determine prevalence estimates for each sequela after a series of adjustments (19). Crosswalks were again performed to adjust for any studies using different DRLEC diagnostic criteria to that used as the case definition by GBD for each DRLEC (see Supplementary Table 2). A series of “squeeze” adjustments were then made to ensure first that the sum of all complicated sequelae did not exceed 90% of the diabetes prevalence and, second, that the sum of foot ulcer and amputation did not exceed 90% of the total neuropathy prevalence (19). Finally, the amputation prevalence was split into those with and without prosthesis on the basis of a standardized scaled country-specific health system access covariate developed by GBD and the assumption that at least 10% of individuals with amputations will not receive a prosthetic. This was supported by data from a pivotal historical study that reported the proportion of prosthetic treatment for major amputations (26) and since supported by multiple contemporary studies (29–31). A population-weighted average was then performed to obtain a regional estimate for the proportion of individuals with amputations who receive a prosthetic. This adjusted regional proportion was then used as the proportion of amputation with prosthesis for each country and a conservative CI of ±50% applied to adjust for any uncertainty of these prosthesis proportion assumptions (19).

Adjusted prevalence estimates were then produced using DisMod-MR 2.1 for each DRLEC for each age-sex-country-year combination on the basis of the GBD reference population, a standard global reference population structure used to facilitate age-standardization (32,33). For countries with sparse or no DRLEC data, the aforementioned wrapper code was again used to adjust estimates on the basis of the estimates of geographical parent region. All GBD 2016 estimates have been estimated with uncertainty. Uncertainty comes from many sources, including heterogeneity in the empirical data that are available and identified through systematic reviews and uncertainty in the indirect estimation models used to make predictions for populations with few or no data. Because the empirical basis for estimating prevalence is much weaker for some DRLEC sequelae than for others, uncertainty varies substantially between and within sequelae across regions and countries.

The disability burden resulting from a sequela or cause within a population is measured using YLDs (19). YLDs are calculated by multiplying the prevalence estimate of a sequela by a disability weight for the same sequela (19). The disability weight represents the average severity of nonfatal health loss associated with the particular sequela on a scale of 0 (equivalent to perfect health) to 1 (equivalent to death) (19,34). For example, uncomplicated diabetes has a comparatively low disability weight of 0.049, whereas blindness due to diabetes has a comparatively high disability weight of 0.187 (19). Disability weights were determined from a series of multinational, household-based, or web-based surveys, “which used paired comparison questions in which respondents considered two hypothetical individuals with different health states and specified which person they deemed healthier” (34). Data from these surveys were analyzed and rescaled to disability weight units between 0 and 1 (34). Lay descriptions of the four DRLEC-related sequelae used to frame the comparison questions can be found in Supplementary Table 2. The final stage in the estimation of YLDs is a microsimulation that adjusts for comorbidity. It is performed by simulating 40,000 individuals for each age-sex-country-year combination. On the basis of the prevalence estimates of sequelae, these simulated individuals end up having none to multiple disease sequelae. The disability weight for each individual is estimated on the basis of the sequelae they acquired, and the disability weight is then attributed to each sequela. The disability weight determined for neuropathy was 0.133 (95% UI 0.089–0.187) (19). The disability weights for the other three sequelae were calculated as a combination of two health states: neuropathy and either foot ulcer (0.02 [0.01–0.034]) or amputation with prosthesis (0.039 [0.023–0.059]) or amputation without prosthesis (0.173 [0.118–0.24]) (19).

We transformed all data provided by IHME into tables and figures to display the global prevalence and YLDs of individual and overall DRLECs in 2016 and 1990 and the composition by sex, 20 age-groups, 21 regions, and 195 countries or territories. Prevalence and YLDs were presented in counts, all-age rates, and age-standardized rates with 95% UIs. Overall DRLEC prevalence and YLDs were calculated by summing mean estimates of the four DRLEC-related sequelae. Mean percentage changes between 1990 and 2016 were calculated for all outcomes in counts, all-age rates, and age-standardized rates. Change in counts reflects the combined effects of population growth, population aging, and epidemiological change (19). Population aging and epidemiological change explain the mean percentage change in all-age rates, while the change in age-standardized rates reflects epidemiological change that is not due to aging or population growth (19). Male-to-female ratios were also calculated across age-groups. Estimates were considered significantly different if their 95% UIs did not overlap.

Table 1 displays global prevalence and YLD estimates for diabetes, individual, and overall DRLECs in 2016, and percentage changes from 1990 to 2016. In 2016, there were an estimated 131.0 million people (1.77% of the global population) affected by overall DRLECs, including 105.6 million (95% UI 85.5–128) with neuropathy only, 18.6 million (15.0–22.9) with foot ulcers, 4.3 million (3.7–4.9) with amputation without prosthesis, and 2.5 million (2.1–3.0) with amputation with prosthesis. The age-standardized prevalence rates showed that 1,848 per 100,000 were affected, including 1,480 (1,201–1,783) with neuropathy only, 270 (217–332) with foot ulcers, 60.6 (52.6–69.2) with amputation without prosthesis, and 36.7 (30.2–43.9) with amputation with prosthesis.

An estimated 16.8 million YLDs (2.07% of global YLDs) were caused by DRLECs in 2016, including 12.9 million (95% UI 8.3–18.8) from neuropathy only, 2.5 million (1.7–3.6) from foot ulcers, 1.1 million (0.7–1.4) from amputation without prosthesis, and 0.4 million (0.3–0.5) from amputation with prosthesis. The age-standardized YLD rates for DRLECs was 237 per 100,000, including 180 (116–263) from neuropathy only, 36.3 (23.8–51.4) from foot ulcers, 15.1 (10.6–20.4) from amputation without prosthesis, and 5.4 (3.6–7.6) from amputation with prosthesis.

Compared with 1990 estimates, counts of prevalence and YLDs in 2016 increased significantly by between 110.4% (neuropathy) and 140.3% (amputation without prosthesis). Age-standardized rates of prevalence and YLDs also increased by between 14.9% (neuropathy) and 31.5% (amputation without prosthesis) from 1990 to 2016, although statistical significance of these increases is unclear because of overlapping 95% UIs.

Figure 1 shows sex and age composition in YLD counts (Fig. 1A) and rates (Fig. 1B) in 1990 and 2016. In 2016, male-to-female ratios for age-standardized YLD rates were 1.11 for overall DRLECs, 0.96 for neuropathy, 1.56 for amputations (both with and without prosthesis), and 1.93 for foot ulcers. See Supplementary Table 3 for male-to-female ratios for YLD counts and rates in each age-group. Age-groups accounting for most YLD counts of individual DRLECs were neuropathy, 50–54 years in males and 55–59 years in females; foot ulcers, 60–64 years in both sexes; amputation without prosthesis, 60–64 years in both sexes; and amputation with prosthesis, 65–69 years in both sexes (Supplementary Table 3). In summary, the 50- to 69-year-old age-group accounted for most (47.8%) of the overall DRLEC YLD counts in 2016, while the >70-year-old age-group had higher YLD rates. YLD rates in 2016 increased in nearly all age-groups compared with 1990 (Supplementary Table 3).

Figure 2 shows broad variation in age-standardized YLD rates across the 21 regions in 1990 and 2016. In both 1990 and 2016, the highest YLD rates of overall DRLECs were in North Africa and Middle East, Central Latin America, Oceania, and Caribbean, while the lowest rates were in East Asia, Western Sub-Saharan Africa, Australasia, and Western Europe. The largest increases from 1990 to 2016 were in Southern Sub-Saharan Africa, South Asia, and Southeast Asia, whereas a decrease was only observed in the High-Income Asia Pacific region.

For individual DRLECs in 2016, highest YLD rates of neuropathy were in North Africa and Middle East (493.9 [95% UI 318.9–726.2] per 100,000), of foot ulcers in Oceania (103.8 [68.6–147.9]), of amputation without prosthesis in Southern Sub-Saharan Africa (44.4 [30.1–59.7]), and of amputation with prosthesis in High-Income North America (20.3 [12.3–28.8]). Figure 3 displays world maps of age-standardized YLD rates for overall DRLECs across 195 countries or territories in 1990 (Fig. 3A) and 2016 (Fig. 3B). The country with the highest overall DRLECs rate was the Marshall Islands (1,096.4 per 100,000), while the lowest was Switzerland (97.1). World maps for prevalence and YLDs of individual DRLECs can be found in Supplementary Figs. 1 and 2, with specific data presented in Supplementary Tables 5 and 6.

We have presented detailed estimates on the disability burden caused by DRLECs for the first time. On the basis of GBD 2016 data, an estimated 131.0 million (1.77%) people worldwide had DRLECs in 2016, including 105.6 million with neuropathy only, 18.6 million with foot ulcers, and 6.8 million with amputations (with or without prosthesis). This resulted in 16.8 million YLDs (2.07% of all YLDs), including 12.9 million from neuropathy only, 2.5 million from foot ulcers, and 1.6 million from amputations. Overall, the DRLEC disability burden appeared to disproportionately affect males, the 50- to 69-year-old age-group, and those living in the regions of North Africa and Middle East, Central Latin America, Oceania, and Caribbean. Compared with 1990 estimates, the 2016 global estimates for age-standardized YLD rates increased by 16.7% for overall DRLECs, and for individual DRLECs, the increases ranged from 14.9% for neuropathy to 31.5% for amputation without prosthesis. Regional changes in age-standardized YLD rates of overall DRLECs also varied over time for different regions, from a 45.4% increase in Southern Sub-Saharan Africa to an 11.6% decrease in High-Income Asia Pacific.

To our knowledge, our previous brief letter was the first to highlight the contribution of DRLECs to the global burden of disability (18). In this article, we report that an estimated 16.8 million YLDs resulted from DRLECs on the basis of GBD 2016 data, a lower estimate than the 20.5 million in 2015 that was based on GBD 2015 data in our letter. This apparent decrease is most likely explained by lower prevalence estimates of diabetes and DRLECs in GBD 2016: Estimates decreased from 435 million (35) in 2015 to 383 million (19) in 2016 for diabetes prevalence, which resulted in a decrease from 159 million (18) to 131 million for DRLECs. Unlike previous GBD estimates (35), GBD 2016 excluded all articles using self-reported diabetes diagnosis, which are known to overestimate prevalence of diabetes (19). Thus, the remeasured GBD 2016 estimates reported in this article, for the new estimation year 2016 and the previous year 1990 using new data from annual updates, improved case definitions, and modeling, should be the more robust.

Because YLDs are a product of prevalence and disability weight, and disability weight has remained the same, it is also reassuring that our prevalence estimates appear to be similar to those reported in the very few other previous articles (5,6,16). We report that 131 million people with diabetes had DRLECs (i.e., neuropathy either alone or with foot ulcers or amputations), equaling 34% of diabetes population. Somewhat reassuringly, our 34% finding seems to plausibly fit within the neuropathy range, reported in another review, of 10–15% in patients with newly diagnosed diabetes and up to 50% in those with a >10-year diabetes duration (6). Additionally, we report that an estimated 18.6 million (4.8%) people with diabetes have a foot ulcer, which aligns very closely with a 4.6% global pooled prevalence estimated in a meta-analysis (16). Although studies have investigated diabetes-related major amputations, as per the combined definition of our amputation outcomes (with or without prosthesis), they did so using incidence, and thus, we are unable to compare our prevalence findings for amputations (17).

Our estimate of 16.8 million YLDs from DRLECs equates to 59% of the diabetes YLDs (28.6 million). Interestingly, when compared with all 271 level 4 causes in GBD 2016, in terms of the proportion of overall YLDs (805.4 million) by each cause, diabetes (3.54%) was the eighth leading cause, made up of DRLECs (2.07%), diabetes-related vision loss (0.03%), and uncomplicated diabetes (1.43%) (19). If we further compare DRLECs (2.07%) with all level 4 causes, DRLECs would rank between the 10th leading cause (falls 2.35%) and the 11th leading cause (chronic obstructive pulmonary disease 2.06%) (19). Furthermore, DRLECs would rank as having a higher global disability burden than ischemic stroke (17th; 1.47%); ischemic heart disease (29th; 0.86%); and CKD due to diabetes (52nd; 0.46%), which is listed in GBD as a separate cause rather than a sequela of diabetes (19) (Supplementary Table 4). Thus, our estimate indicates that DRLECs make up a very large proportion of the diabetes disability burden and a comparatively large proportion of the overall global disability burden in 2016. In an aging global population, diseases largely affecting the elderly, such as diabetes, will become more prominent causes of YLDs. The increase in diabetes and the improved survival in people with diabetes through prevention of deaths from cardiovascular complications will most likely lead to further increases in prevalence and a need to prevent these disabling complications of diabetes, in particular DRLECs.

This comparatively large disability burden of DRLECs appeared to disproportionately affect males, and the largest proportion (47.8%) of this burden affected the 50- to 69-year-old population. These findings are not unique because diabetic foot ulcers and amputations have consistently been reported to be much higher in the male population, with median ages of disease onset in this same middle- to older-age bracket (16,36,37). The previous meta-analysis on diabetic foot ulcer prevalence reported a male-to-female ratio of 1.3 (16). A recent systematic review on diabetic amputation incidence reported a male-to-female ratio range between 1.5 and 3.0 (36,37) and attributed this to males having a higher prevalence of smoking, peripheral neuropathy, peripheral artery disease, and foot ulceration (36). We did not find this for neuropathy, although neuropathy did appear to affect males more so in middle-aged groups than females (38), but we did for foot ulceration. Similarly, increasing proportions of diabetes-related complications have been reported in the middle-aged population. Another review identified that the proportional contribution of 45- to 64-year-olds to all diabetes-related strokes increased from 20.1 to 31.4% and amputation from 32.4 to 52.8% during 1990 and 2010 (3). It could be suggested that the large proportional contribution of middle-aged groups may partly be attributed to the relatively large size of middle-aged groups compared with other age-groups, as suggested by the changing shapes of counts and rates over time shown in Fig. 1.

Our findings also show that almost all regions were faced with an increased burden of DRLECs. Along with the global epidemic of diabetes and interventions to prevent death and prolong life, more people were living with diabetes, and this increasingly complicated population is placing an extra challenge on health systems worldwide. The most affected regions found in this study were North Africa and Middle East, Central Latin America, and Oceania, perhaps not surprisingly, because these were the same regions most affected by diabetes (19); however, inadequate health care provision might be another underlying factor. An attempt to evaluate health care system performance in GBD is through a ratio of actual disease burden to expected disease burden. If the ratio is larger than 1 (actual burden exceeds expected burden), it is suggested that health care systems in those geographical locations are not performing to levels necessary to cope with the specific disease (19). In both Central Latin America and Oceania, actual YLDs from diabetes significantly outnumbered the expected by twofold, indicating that the quality of local health care is not meeting the needs of their diabetes population (19). Another attempt GBD has made to assess health care is through the establishment of the Health Access Quality Index, which assesses personal health care access and quality for regions and countries (39). GBD reports that Oceania scored 36.0 out of a 100 possible total score. Although North Africa and Middle East scored a higher 55.8, large between-country variation exists, ranging from 25.9 in Afghanistan to 85.6 in Lebanon (39). Interestingly, only High-Income Asia Pacific (Japan, South Korea, Singapore, and Brunei) saw reductions in disability burden from almost every DRLEC since 1990. Perhaps researchers could investigate the strategies used in this region that may have led to their apparent DRLEC success and, in turn, used in those regions most affected.

Several limitations should be taken into consideration when interpreting these findings. First, our estimates for diabetic neuropathy may be an overestimate. Unlike the other three outcomes, neuropathy can affect more than just the lower extremities. While the GBD definition of diabetic neuropathy included only peripheral neuropathy, this does mean that upper as well as lower extremities were included. However, peripheral neuropathy is known to overwhelmingly affect lower extremities, so any overestimate should be minimal (6). Conversely, if we chose to exclude neuropathy, it would have resulted in a significant underestimate of the overall burden (4,5). Second, we did not have data to include other diabetic complications of the lower extremity that also cause significant disability, such as peripheral artery disease, foot infections, and minor amputations (4,5). Thus, our overall DRLEC burden may in fact be an underestimate of the actual overall DRLECs, and there is a need to improve the way DRLECs are captured and categorized in future GBD iterations so that the overall DRLEC burdens are more comprehensively estimated. Third, we were reliant on GBD data, and caution is needed when interpreting the results, particularly in data-sparse countries with wide 95% UIs around estimates. While there has been significant investment in improving GBD global disease modeling approaches, there is still a need for more investment to improve vital registration and data collection in developing countries. The accuracy of DRLEC estimates depends on the availability of reliable and comparable information, and as better-quality evidence becomes available, future disease modeling at a global level will be more accurate and reliable. Fourth, disability weights, a crucial component of YLDs, were derived from models using survey data, and responses from these surveys were sensitive to details included in the lay descriptions (34). Further improvements on disability weight methodology may also improve the precision of estimates in the future. Fifth, some results, including the overall DRLEC prevalence and YLD estimates, were a simple summation of individual DRLECs by the authors, and 95% UIs for these findings were not estimated. Finally, diabetes is also associated with an increased risk of mortality. However, given the complex interrelationships between diabetes complications and other comorbidities causing mortality, estimating the impact of individual complications on mortality remains challenging; hence, we were unable to investigate the mortality burden of DRLECs.

Despite these limitations, results from this study add significant novel knowledge on the global burden of DRLECs. DRLECs appear to be a large contributor to the global disability burden, and this age-standardized burden has increased by 15–31% since 1990. Disability from DRLECs appeared to disproportionately affect males, the middle- to older-aged groups, and specific regions. It is recommended that these populations are targeted by DRLEC strategies shown to reduce these burdens. While general diabetes management strategies may help to reduce the DRLEC disability burden through prevention of complications, evidence suggests that the most significant DRLEC burden reductions were from DRLEC-specific strategies, such as the introduction of interdisciplinary foot care services adhering to evidence-based guidelines (4,5,11). Thus, in a similar manner to how CKD due to diabetes is treated as a separate cause by GBD and others, we also recommend that DRLECs should be monitored and addressed separately. This would allow ongoing interpretations of DRLECs and will particularly facilitate worldwide policy makers and clinicians to tackle DRLECs, a leading cause of global disability burden.

Acknowledgments. The authors acknowledge the IHME for providing the data for this article.

Funding. Funding sources for this work came from the Chinese Scholarship Council and Australian National Health and Medical Research Council.

The funding sources had no role in study design, data analysis, interpretation, or decision to submit for publication.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. Y.Z. contributed to data acquisition, analysis, and interpretation and drafted and critically reviewed the manuscript for intellectual content. P.A.L. contributed to the conception and design of the study, data analysis and interpretation, and drafting and critical review of the manuscript for intellectual content. S.M.M. contributed to the data analysis and interpretation and critical review of the manuscript for intellectual content. J.J.v.N. and D.G.A. contributed to the conception and design of the study and critical review of the manuscript for intellectual content. R.E.P. contributed to the conception and design of the study; data acquisition, analysis, and interpretation; and drafting and critical review of the manuscript for intellectual content. All authors reviewed and approved the final version of the manuscript. Y.Z. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented orally at the 8th International Symposium on the Diabetic Foot, The Hague, the Netherlands, 22–25 May 2019.