Metabolic Factors, Lifestyle Habits, and Possible Polyneuropathy in Early Type 2 Diabetes: A Nationwide Study of 5,249 Patients in the Danish Centre for Strategic Research in Type 2 Diabetes (DD2) Cohort

Diabetic polyneuropathy (DPN) affects 25–50% of patients with type 2 diabetes (1,2). DPN increases the risk of falls, foot ulcers, and lower-extremity amputations (1,2), and up to 38% of patients experience neuropathic pain (1,3,4). Current preventive measures for DPN are mainly limited to strict glycemic control, which exerts a limited effect against DPN risk in type 2 diabetes patients (5).

Increasing evidence supports an association between the degree of obesity and risk of DPN in type 2 diabetes (6–10), but the exact biological mechanisms remain unclear. Visceral fat accumulation associates with metabolic dysfunction, e.g., low-grade inflammation, insulin resistance, and dyslipidemia. Some individuals who are obese by BMI criteria (BMI >30 kg/m²) can be metabolically healthy (11), which can possibly be attributed to less visceral fat distribution (12). Accordingly, central obesity has been shown to be a stronger predictor of some diabetes complications than—and independently of—BMI (13). However, it is not known whether central obesity associates with DPN independently of BMI in type 2 diabetes. Results are mixed and conflicting from studies examining other possible DPN risk factors in diabetes populations, such as metabolic syndrome–related factors, including dyslipidemia, hypertension, and hyperglycemia as well as low-grade inflammation and lifestyle habits like smoking and physical activity (3,7–9,14,15). Large-scale studies on DPN in type 2 diabetes patients are scarce, and existing DPN studies have often included patients with long-standing diabetes rather than newly diagnosed diabetes where the potential to prevent complications may be largest. Specifically, little knowledge is available on risk factors that may underlie the presence of neuropathic pain in type 2 diabetes (16).

In a recent cross-sectional analysis of the nationwide Danish Centre for Strategic Research in Type 2 Diabetes (DD2) cohort, we investigated the association of DPN and painful DPN with mental health comorbidities and with a few selected patient characteristics including age, sex, and BMI, assessed concurrently with DPN (17,18). In the current study, we linked the DD2 cohort with a range of other medical databases to comprehensively investigate the association of various metabolic and lifestyle factors at diabetes diagnosis with occurrence of DPN and neuropathic pain at a median of 2.8 years later. We examined the hypothesis that central obesity markers are strongly—and independently of general obesity—associated with DPN (12,13,19). We also investigated whether higher levels or rates of a range of other metabolic and lifestyle risk factors would associate with higher DPN prevalence. Finally, we explored the hypothesis that distinct metabolic factors associate with painful DPN compared with nonpainful DPN.

The DD2 cohort is a nationwide cohort of individuals with newly or recently diagnosed type 2 diabetes (median diagnosed diabetes duration at enrollment time 1.3 years [interquartile range (IQR) 0.3–2.9]) enrolled from hospital specialist outpatient clinics and from general practitioners’ offices in Denmark since November 2010. The enrollment process, implementation, logistics, DD2 biobank, and characteristics of this cohort have previously been described (18). Briefly, interview and clinical examination data for each patient are recorded at the DD2 enrollment date, and fasting blood and urine samples are obtained and stored in the DD2 biobank. The unique civil personal registration number assigned to all Danish citizens links the DD2 cohort to other Danish health registries, including a complete hospital contact history from the Danish National Patient Registry, filled drug prescriptions from the Danish National Health Service Prescription Database, and information on vital status and migration from the Danish Civil Registration System. For a subcohort of DD2 patients (69%), additional detailed clinical data can be retrieved from the nationwide quality-of-care database, the Danish Diabetes Database for Adults (DDDA) (18).

In June 2016, a median of 2.8 years (IQR 1.8–3.7) after the DD2 enrollment date, a detailed questionnaire on neuropathy and pain was sent out to all 6,726 living DD2 participants enrolled from November 2010 to February 2016 (17). The questionnaire included the 15-item Michigan Neuropathy Screening Instrument questionnaire (MNSIq), the 7-item Douleur Neuropathique en 4 Questions (DN4) questionnaire, questions about pain location (e.g., whether the person experienced pain in both feet), anthropometric data, and lifestyle factors.

Our main study population consisted of the 5,249 (78%) DD2 patients, who provided information on both polyneuropathy and neuropathic pain status from the questionnaire survey (Supplementary Fig. 1). Characteristics of all 5,514 DD2 patients who returned a fully or partly filled questionnaire have previously been described (17).

The MNSIq tool was developed to screen for and identify DPN (20,21). We used the validated cutoff score of ≥4 (specificity 92%, sensitivity 40%) to assess possible DPN (21,22). Neuropathic pain was evaluated according to NeuroPPIC (neuropathic pain phenotyping by international consensus for genetic studies) (23), and the updated NeuPSIG (Neuropathic Pain Special Interest Group) neuropathic pain grading system (24) defining possible neuropathic pain as 1) pain with neuropathic characteristics, 2) an anatomically plausible pain distribution (here, pain in both feet), and 3) a history of a relevant underlying somatosensory lesion or disease (here, diabetes) (24). We used the DN4 questionnaire (25), which has specifically been validated in DPN (specificity and sensitivity 84%) (26), and defined neuropathic pain as the presence of pain in both feet together with a DN4 score of ≥3. It was emphasized in the questionnaire that the DN4 questions specifically related to pain in the feet and should only be answered if there was pain in both feet. Thus, DPN was defined as MNSIq ≥4; painful DPN as MNSIq ≥4 + pain in both feet + DN4 ≥3, and nonpainful DPN as MNSIq ≥4 in combination with either DN4 <3 or no pain in the feet (Supplementary Fig. 2A).

We used information on BMI (weight in kilograms divided by the square of height in meters) as a measure of general obesity at three different time points: at 20 years of age (based on recall at the time of DD2 enrollment), at time of DD2 enrollment (subcohort: based on DDDA data, i.e., recorded as part of routine clinical diabetes care [18]), and at time of the questionnaire survey in 2016 (based on self-reported data). We then calculated changes in BMI from age 20 years to the time of questionnaire in 2016.

Waist and hip circumference were measured as part of the DD2 enrollment process and were used to assess central obesity with three different measures: waist circumference, waist-to-hip ratio, and waist-to-height ratio (13).

The timeline of obesity measures, nonobesity metabolic risk factors, lifestyle factors (see below), and DPN status is shown in Fig. 1. For further definitions and categories, see Supplementary Table 1.

Information on other patient characteristics and metabolic and lifestyle factors at time of DD2 enrollment (from here on referred to as “baseline”) was extracted from the DD2 cohort data and linked health registers. Patient characteristics and metabolic factors of particular interest that were available for the entire population included C-peptide level (insulin resistance surrogate), low-grade inflammation assessed by hs-CRP (excluding hs-CRP values ≥10 mg/L to exclude patients with ongoing infections) (27), physical activity (days per week with >30 min of physical activity), alcohol consumption (≤ or >21/14 units per week for males/females—the recommended safe dose in 2010 when the DD2 was initiated), and diabetes duration (at the time of questionnaire survey [2016]). Metabolic risk factors available for the subcohort linked via the DDDA included hemoglobin A_1c (HbA_1c), lipid levels (total cholesterol, LDL cholesterol, HDL cholesterol, and triglycerides), smoking habits (current, former [>6 months ago], and never), and blood pressure.

Some patients may have had normal blood pressure, HbA_1c, and lipid levels at baseline due to relevant treatment. Thus, we also retrieved information on antihypertensive, glucose-lowering, and lipid-lowering drug usage within 1 year prior to baseline.

In contrast to the focus of our previous publication based solely on data at the time of the questionnaire survey (17), our main focus was the lifestyle and metabolic risk factor profile in patients around the time of type 2 diabetes diagnosis, i.e., at baseline. For smoking and physical activity, however, we also used follow-up data from the neuropathy questionnaire a median of 2.8 years later to assess the role of risk factor changes.

Descriptive data were median (IQR) for continuous variables and proportions (n [%]) for categorical variables. We examined the proportion of overall DPN, nonpainful DPN, and painful DPN. We then calculated prevalence ratios (PRs) with 95% CIs of DPN associated with obesity measures and other metabolic and lifestyle factors using log-binomial and Poisson (with robust error variance) regression models (28). Continuous risk factors were investigated both as categorical and continuous variables. All PRs were adjusted for age, biological sex, and diabetes duration. We did not make further adjustments in our main analysis because the obesity measures and other metabolic and lifestyle factors may act as intermediates and clusters in the same incompletely understood pathophysiological pathways. The associations between obesity measures and DPN were also evaluated with the use of restricted cubic spline regressions with five knots (29). To elaborate on the associations of central obesity measures with DPN independently of BMI, we additionally adjusted for BMI in these models. The analyses of change of physical activity level were stratified according to baseline physical activity level.

Next, we restricted the cohort to those with DPN (MNSIq ≥4) and calculated the adjusted PR of painful DPN for each risk factor under study.

In sensitivity analyses, we restricted the population to individuals with a short registered diabetes duration (<1 year and <0.5 years) at DD2 enrollment. Since having diabetes and neuropathic pain (DN4 ≥3) in both feet may be considered to fulfill the NeuroPPIC/NeuPSIG criteria for possible painful DPN (23,24) despite a MNSIq score <4, we did a sensitivity analysis including these patients in the painful DPN group, as we did in our previous publication (17) (Supplementary Fig. 2B). In another sensitivity analysis, we excluded patients with alcohol overconsumption because peripheral neuropathy may result from DPN, alcoholic polyneuropathy, or a mixture in these patients. Because central obesity may precede other metabolic risk factors, we performed additional analyses in which we adjusted the DPN analyses for waist-to-hip ratio. We also did analyses controlling for HbA_1c level.

Finally, we examined presence of the metabolic syndrome as one entity as a risk factor for DPN and painful DPN (definition in Supplementary Table 1).

We tested for effect measure modification by sex using the likelihood ratio test.

The Danish National Committee on Health Research Ethics (record number S-20100082) and the Danish Data Protection Agency (record number 2008‐58‐0035) approved the DD2 study. All DD2 patients volunteered to participate in the DD2 study and gave written informed consent.

We included 5,249 patients, of whom 938 (17.9%) had DPN, including 386 (7.4%) with painful DPN (17). Median age was 65 years (IQR 57–72), 42% were female, and median diabetes duration was 4.6 years (IQR 3.5–5.7) at DPN assessment (Supplementary Table 2).

Higher BMI at baseline and at the questionnaire date, as well as larger waist circumference, waist-to-hip ratio, and waist-to-height ratio were all associated with a higher DPN prevalence (Fig. 2). Similar results were observed when the obesity measures were analyzed as continuous variables (Supplementary Table 3). The magnitude of the association with DPN for 1 SD increase was similar for the general and central obesity measures (Supplementary Table 3). Spline regression analyses yielded approximate linear relations with DPN for general and central obesity measures, except for a J-shaped association observed with BMI at age 20 years and with BMI change since age 20 years (Supplementary Fig. 3).

When we additionally adjusted central obesity for BMI, all central obesity measures remained positively associated with DPN (Supplementary Fig. 4). For example, for a given BMI, DPN prevalence increased by a factor of 1.86 (95% CI 1.32; 2.61) for individuals with a waist circumference of ≥102/88 cm (male/female) vs. <94/80 cm. In these analyses, BMI persistently associated with DPN.

No statistically significant effect measure modification by sex was found; however, PRs tended to be higher among males for central obesity measures (Supplementary Table 4).

Figure 3 shows risk estimates for baseline nonobesity metabolic risk factors and lifestyle habits. Metabolic factors markedly associated with DPN included low HDL cholesterol, high triglycerides, low-grade inflammation, higher C-peptide, and higher HbA_1c (Supplementary Table 5). Antihypertensive drug treatment was associated with DPN but not systolic and diastolic blood pressure. Finally, insulin treatment was associated with DPN (Supplementary Table 6). No statistically significant effect measure modification by sex was found.

Lower physical activity level and a current smoker status or former smoker status at baseline were clearly associated with DPN. Of note, continued smoking compared with smoking cessation between baseline and questionnaire date was also associated with DPN (adjusted PR [aPR] 1.24 [95% CI 0.80; 1.92]), yet with limited statistical precision. In contrast, no clear association was observed with change in physical activity level in the main analyses (Fig. 3). However, in the analyses stratified by baseline activity, DPN prevalence was low in patients whose activity level was low at baseline but had increased at DPN assessment time (aPR 0.67 [95% CI 0.52; 0.85]). Conversely, DPN prevalence was high in patients whose activity was high at baseline but had decreased at DPN assessment (aPR 1.20 [95% CI 1.00; 1.46]) (Supplementary Table 5). No statistically significant effect measure modification by sex was found.

Among DPN patients, we did internal analyses of risk factors associated with neuropathic pain (i.e., factors associated with painful vs. nonpainful DPN) shown in Table 1 and Supplementary Tables 7–10. Several risk factors appeared to associate with increased prevalence of painful DPN, yet often with limited statistical precision. These included central obesity (waist circumference and waist-to-hip ratio), high systolic blood pressure, high total cholesterol, high LDL cholesterol, and high triglycerides. In analyses of continuous data (Supplementary Tables 8 and 9), some of these factors reached statistical significance, e.g., total cholesterol (unit = 0.5 mmol/L, aPR 1.07 [95% CI 1.01; 1.13]). Neuropathic pain was associated with alcohol overconsumption at baseline and with smoking at baseline (current smoking, aPR 1.17 [95% CI 0.90; 1.51]) and at the questionnaire date (current smoking, aPR 1.29 [95% CI 1.03; 1.62]) (n = 608 vs. n = 938, respectively), with the latter in agreement with our previous study using a slightly different painful DPN definition (17). Increased physical activity from baseline to questionnaire date was associated with lower painful DPN prevalence (aPR 0.82 [95% CI 0.67; 0.99]).

Of the 5,249 patients, 130 (2.5%) had pain in both feet and DN4 ≥3 but MNSIq <4 (17). For the majority of risk factors, these patients were more similar to patients without DPN than to patients with painful DPN. Including these 130 patients in the painful DPN group yielded an overall prevalence of painful DPN of 10% as reported in our previous study (17). This inclusion marginally reduced most risk estimates for DPN and for neuropathic pain but did not change our conclusions (data not shown).

Restricting the cohort to individuals with diabetes duration ≤1 year and ≤0.5 year generally supported the main analyses but with lower precision (data not shown). Excluding DPN patients with alcohol overconsumption from the analyses did not change conclusions (data not shown).

Of note, most risk factor–DPN associations were robust to further adjustment for central obesity and for hyperglycemia (Supplementary Tables 11 and 12). For example, the strong association of high C-peptide with DPN (≥1,550 pmol/L, aPR 1.72 [95% CI 1.43; 2.07]) attenuated only moderately when we adjusted both for central obesity as a potential cause of insulin resistance and for HbA_1c as a potential effect of insulin resistance: aPR 1.49 (95% CI 1.19; 1.85).

The metabolic syndrome associated strongly with DPN (aPR 2.19 [95% CI 1.58; 3.02]) but not with painful DPN (aPR 1.09 [95% CI 0.70; 1.70) (Supplementary Tables 5 and 7).

This is the largest study to date to investigate in detail multiple obesity measures and several metabolic and lifestyle factors with both DPN and painful DPN in patients with early type 2 diabetes. We found that both general and central obesity are strongly—and independently of each other—associated with DPN prevalence. Other metabolic and lifestyle factors that clearly associated with DPN prevalence included low-grade inflammation, hypertriglyceridemia, hyperglycemia, high C-peptide levels, low HDL levels, antihypertensive drug use, tobacco smoking, and low physical activity. These findings suggest that controlling metabolic factors through weight loss and medications as well as lifestyle interventions, including smoking cessation and increased physical activity, may potentially reduce DPN risk.

Even within this large study, statistical precision was limited for the analyses of painful DPN. While metabolic syndrome metrics (central obesity, increased lipids, and hypertension) seemed to associate with neuropathic pain, these results generally did not reach statistical significance and we observed no clear association of neuropathic pain with the metabolic syndrome. Notably, we found clear evidence that high alcohol intake, tobacco smoking, and failure to increase activity after diabetes diagnosis associated with higher prevalence of painful DPN. These results are extremely important, as all three risk factors are modifiable without requiring medications.

Compared with previous studies of obesity (6,8,9,17,30), we additionally found that central fat distribution is associated with DPN independent of BMI. In line with this observation, we found increased DPN prevalence with metabolic factors closely associated with central obesity, including low-grade inflammation and C-peptide as a marker of insulin resistance (12,31). On the contrary, other studies have linked insulin deficiency with DPN (32,33). A recent study by Zaharia et al. (33) suggested that a diabetes phenotype with less obesity, higher HbA_1c values, and more insulin deficiency but without autoimmunity (severe insulin-deficient diabetes) (n = 28) associated with increased DPN risk compared with four other clusters, paradoxically including also that of insulin deficiency with coexisting autoimmunity. Further studies are thus needed to clarify the exact association between insulin levels and DPN. Although associated with DPN, insulin resistance, HbA_1c, and low-grade inflammation did not seem to associate with neuropathic pain in our study, whereas Doupis et al. (15) previously reported an association of increased CRP in painful versus nonpainful DPN. Also, studies of diabetic animal populations and of human populations without diabetes have suggested a role of inflammatory markers in painful polyneuropathy (34). Future studies of painful DPN in type 2 diabetes should investigate a broader range of inflammatory markers in addition to hs-CRP.

The finding that lower HDL cholesterol associates with DPN was also recently reported in the Anglo-Danish-Dutch study of Intensive Treatment of Diabetes in Primary Care (ADDITION) cohort with screen-detected type 2 diabetes (8) and in most studies of long-standing diabetes (3,9,14). In contrast to our results, lower LDL cholesterol also predicted DPN in the ADDITION cohort (8) and was associated with peripheral nerve damage in another type 2 diabetes study (35). The authors speculated whether this might result from statin treatment (8) or impaired peripheral nerve regeneration with insufficient cholesterol supply (35).

Previous smaller studies of patients with type 2 diabetes have not found an association of hypertension and DPN (6,8,9,14,36). Well-controlled blood pressure in our cohort may have masked an association between DPN and hypertension, which might be supported by our higher DPN prevalence seen among patients on antihypertensive drugs.

Current, former, and continued smoking versus cessation associated with both DPN and neuropathic pain. The latter may possibly reflect detrimental effects from higher cumulative smoke exposure (dose and time), whereas cessation could possibly reverse smoke-induced nerve damage. A recent meta-analysis reported only low-grade evidence for smoking as a DPN risk factor (37); however, previous studies often merged former and never smokers as reference group, which may have masked any association (37). Since peripheral arterial disease is smoking related and may also cause pain and positive answers to some of the MNSIq questions, we cannot exclude that we overestimate the associations with smoking due to a mix of DPN with peripheral arterial disease. Thus, a post hoc analysis showed that smoking was strongly associated with MNSIq question 12, “Do your legs hurt when you walk?” with aPRs of 1.84 (95% CI 1.59; 2.13) for current and 1.44 (95% CI 1.26; 1.65) for former smokers (data not shown). On the other hand, post hoc restriction of our DPN analyses to patients without a prior history of peripheral arterial disease, or alternatively adjusting for peripheral arterial disease in the model, only reduced the associations of smoking with DPN and painful DPN very slightly (data not shown). Finally, we cannot exclude that reverse causation, i.e., tobacco smoking as self-medication for pain, may explain part of the association with painful DPN.

The fact that high alcohol intake at baseline (defined as >21/14 units for males/females) but not as defined at the later questionnaire (>14/7 units for males/females) (17) associated with neuropathic pain may suggest a dose-response relationship.

New and promising research advocates physical exercise for preventing and treating DPN (38). This is in line with our observation that DPN associated with low baseline physical activity and that increased physical activity after baseline related to reduced DPN prevalence among those with low baseline level and with reduced neuropathic pain prevalence. We cannot exclude the possibility of reverse causation, i.e., that absence of DPN symptoms may have led to more physical activity. Future intervention studies are needed to clarify any preventive or therapeutic role of physical activity against DPN.

This study included patients with early diabetes, which is relevant when determining correlative features that might emerge as possible preventive measures. The median 2.8 years time frame from baseline cohort enrollment until DPN assessment implies that DPN cases were a mixture of new incident and preexisting prevalent DPN. Of note, our findings were robust in the subcohort with very short diabetes duration at baseline, who more likely had new incident DPN at later follow-up, thus more closely mimicking a longitudinal study. This strengthens our conclusions and suggests potential interventional measures.

Our study also has limitations. First, DPN and painful DPN assessment relied on the MNSIq and DN4 questionnaires and not on neurological examinations, nerve conduction studies, or validated small-fiber measures. This leads to a DPN and neuropathic pain assessment at the level of “possible” (22–24). However, both tools are validated (21,26). Although MNSIq sensitivity is rather low (40% in a study of long-standing type 1 diabetes patients [21]) for measures of relative risk, high specificity leads to unbiased results in comparative analyses and is thus more important than sensitivity (39). Overlap of pain-related questions in the MNSIq (i.e., question 2, “Do you ever have any burning pain in your legs and/or feet?”) with questions on pain in the DN4 may have mitigated our possibility of discrimination between nonpainful and painful DPN (17). This may have impeded our ability to identify risk factors specifically associated with painful DPN. Second, despite the time elapse since patient characteristics were determined at baseline until DPN assessment a median of 2.8 years later, our main analyses reflect a cross-sectional study design due to the unknown DPN and pain status at baseline. This leads to intrinsic uncertainty about temporal relationships and the possibility for reverse causality for some associations. Third, self-reported BMI and other factors may be subject to recall errors. There is evidence that self-reported anthropometric data are reasonably accurate and adequate for use in large epidemiological studies (40). Finally, it is a limitation that some variables were only available for a subcohort of patients.

In conclusion, these data provide evidence that DPN in early type 2 diabetes is closely associated with specific risk factors in addition to hyperglycemia, including metabolic syndrome factors, insulin resistance, and low-grade inflammation. Moreover, unhealthy lifestyle habits, including smoking and physical inactivity, are modifiable factors strongly associated with DPN. Pain occurrence in DPN may share some, but not all, of these modifiable risk factors. Future longitudinal studies should further investigate specific risk factors for DPN and painful DPN and the clinical effects of improving such factors.

Acknowledgments. The authors sincerely thank all DD2 study participants including the patients and the staff in private practices and outpatient hospital clinics. The authors sincerely thank statistician Sia Kromann Nicolaisen (Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark) and statistician Eskild Bendix Christiansen (Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark) for skillful assistance in the data management process.

Funding and Duality of Interest. DD2 is supported by the Danish Agency for Science (grant number 09-067009, 09-075724), the Danish Health and Medicines Authority, the Danish Diabetes Association, and an unrestricted donation from Novo Nordisk A/S. Project partners are listed on the website www.DD2.nu. Research reported in this publication is part of the International Diabetic Neuropathy Consortium research program, which is supported by a Novo Nordisk Foundation Challenge Programme grant (grant number NNF14OC0011633). The Department of Clinical Epidemiology, Aarhus University Hospital, receives funding for other studies from companies in the form of research grants to (and administered by) Aarhus University. None of these studies have any relation to the current study. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. D.H.C. designed the study, performed the statistical analyses, wrote the first draft, and contributed to the discussion. R.W.T., S.T.K., H.A., N.B.F., B.C.C., and T.S.J. designed the study, contributed to the discussion, and reviewed and edited the manuscript. S.S.G. researched the data, contributed to the discussion, and reviewed and edited the manuscript. L.B.C. performed statistical analyses, contributed to the discussion, and reviewed and edited the manuscript. J.S.N., H.B.-N., and E.L.F. contributed to the discussion and edited and reviewed the manuscript. All authors approved the final manuscript. D.H.C. 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 in abstract form at the 55th Annual Meeting of the European Association for the Study of Diabetes, 16–20 September 2019, Barcelona, Spain.