Renal impairment is one of appropriate predictors of future diabetic peripheral neuropathy: a hospital-based 6-year follow-up study

The relationship between renal impairment and diabetic peripheral neuropathy (DPN) remains inconclusive. We aim to investigate the risk factors for the occurrence of DPN in Taiwanese adults with type 2 diabetes mellitus (T2DM) and focus on renal impairment. A hospital-based study was conducted from 2013 to 2019 and 552 Taiwanese people who had T2DM without DPN at baseline were enrolled. DPN was diagnosed using the Michigan Neuropathy Screening Instrument. Potential risk factors were recorded, including patient’s sociodemographic factors, current medication usage and biochemical markers. As of 2019, 73 developed DPN and 479 had no DPN. The cumulative incidence during the 6-year period was 13.22%. Multivariable logistic regression analysis revealed that lower estimated glomerular filtration rate (eGFR) (odds ratio [OR] 0.98, p = 0.005), advanced age (OR 1.06, p = 0.001), increased body weight (OR 1.04, p = 0.018), duration of DM (OR 1.05, p = 0.036) and male gender (OR 3.69, p = 0.011) were significantly associated with future DPN. In addition, patients with T2DM under the age of 65 with higher serum creatinine concentration (OR 8.91, p = 0.005) and higher baseline HbA1C (OR 1.71, p < 0.001) revealed significantly associated with future DPN. In conclusion, this is the first large scaled hospital-based study with long term follow-up to investigate risk factors for DPN in Taiwanese. Lower eGFR and higher serum creatinine concentration, particularly in people under the age of 65, are predictors of future DPN in Taiwanese people with T2DM. Other predictors included advanced age, increased body weight, duration of DM, male gender for all ages and HbA1c in enrolled patients under the age of 65. Our study not only confirms the association between renal impairment and future DPN but also provides a commonly available assessment to predict the future DPN.

The global burden of diabetes mellitus (DM) has increased enormously in recent decades and will continue to soar in the next few decades. In fact, the global incidence of diabetes has increased by 102.9% from 11.3 million in 1990 to 22.9 million in 2017. Consequently, the prevalence of the complications resulting from type 2 diabetes (T2DM) is likely to rise 1 .
DPN is the most common complication, and its lifetime prevalence is up to 50% in adults with T2DM 2 . DPN is associated with a wide range of clinical manifestations, of which distal sensory neuropathy is predominant. This manifestation contributes to numerous disabling morbidities, such as diabetic foot ulceration, impaired balance, and distressing neuropathic pain, which are often difficult to treat. Furthermore, DPN is the most common cause of non-traumatic lower-limb amputations in most high-income countries 3 . The current study focuses on distal and symmetric polyneuropathy.

Methods
Study design and participants. This is a hospital-based, prospective, observational study. Between January 2013 and October 2013, patients over 18 years old with prevalent or newly diagnosed T2DM were eligible for inclusion. The diagnosis of T2DM were based on the criteria of American Diabetes Association (ADA). Data were obtained from patient's medical records, laboratory examinations, questionnaires and anthropometric measurements at the time of enrollment. Exclusion criteria were as follows: patients having type 1 DM or gestational diabetes, patients had DPN at baseline and whose did not complete the questionnaires or blood sample test at baseline or during the following 6 years. Finally, 552 participants were enrolled in our study.
Participants have been followed observationally via clinical follow-up examination and questionnaires. The blood sample test was performed at least once a year. Our study consequently carried out to 2019-6 years after the trial baseline.
Each of the participants was diagnosed by endocrinologists in the outpatient units at a tertiary medical center in middle Taiwan, which serves approximately 6600 outpatients and 1400 inpatients per day and mainly Han-Chinese population. Before drawn for analysis, the patients' information was anonymized by computer system, and the researchers were blinded to these data. The study was approved by the Institutional Review Board of Taichung Veterans General Hospital (CG18082B-1). All participants volunteered for the current studies, and provided written informed consent prior to enrolment. Besides, all the methods were performed in accordance with relevant guidelines and regulations.
Biochemical data. Laboratory examination were administrated during endocrinological follow-up. Blood samples were obtained in the morning after an overnight fasting period from the antecubital vein. Fasting plasma glucose (FPG; using standard enzymatic methods), glycated hemoglobin (HbA1c; using high-performance liquid chromatography), serum creatinine concentration and plasma lipid profiles (using standard enzymatic methods), including total cholesterol (TC), HDL, LDL, and TG. For lipid profile, we defined the following cut-off points of pathologic values according to the National Cholesterol Education Program (NCEP) Adult Treatment Panel (ATP) III 14 : TG: > 150 mg/dL; HDL: < 50 mg/dL in female and < 40 mg/dL in male. EGFR was estimated by the six-variable Modification of Diet in Renal Disease (MDRD) equation as the following equation: For the patients who had received lipid-lowering drugs, their baseline lipid profiles were defined as the mean plasma lipid values (including TC, HDL, LDL and TG) in the 5 years prior to drug prescription and the follow-up lipid profiles were defined as the mean lipid values in the 5 years after medication prescription. For the individuals never received lipid-lowering drugs, the baseline lipid profiles were circumscribed as the 5-year mean plasma lipid values prior to study enrollment, from 2008 to 2012, and the follow-up lipid profiles were defined as the mean lipid values within the 5 years after study entry, from 2013 to 2017. Assessment of diabetic peripheral neuropathy. All of the included patients received assessment of DPN at the time of enrollment and received second assessment after 6-year follow-up by the same trained and certificated care-management nurse to minimize the inter-rater reliability. DPN was evaluated based on the www.nature.com/scientificreports/ second component of MNSI. Physical appearance of feet, ulceration, ankle deep tendon reflexes, and the perception of light touch (using Semmes-Weinstein 5.07 10-g monofilament) and distal vibration (using 128-Hz tuning fork) were investigated. As previous validated studies in adults 15 , individuals whose MNSI examination (MNSIE) score > 2 were diagnosed with DPN.
Assessment of renal function. We evaluated participants' baseline renal function with serum creatinine concentration and eGFR in 2013. The eGFR was estimated by MDRD equation which contains elements as serum creatinine, age and gender (a constant in the equation). Because the relationship between serum creatinine, age and eGFR is hyperbolic, we establish a model that do not adjust the serum creatinine, age and gender for eGFR in multivariable logistic regression analysis (Table 3) to statistic the interference between baseline eGFR and the occurrence of DPN. Besides, it is well-established that serum creatinine had multiple limitations to represent the true renal function and age is an important factor among these 16 . Furthermore, renal function declines with advancing age. Recent research reported that serum creatinine concentration was not enough to represent a screening test for renal impairment in people aged 65 and above 17 . Because of there were high percentage (14.4% ~ 17% varied by gender) of people aged 65 and above had a serum creatinine concentration above the laboratory reported upper reference limit of normal 18 . Serum creatinine might lead to marked under-investigation and under-recognition of renal failure in this population. Thus we stratified the serum creatinine concentration by age group into age≧65 and age < 65. Each groups were carried out the multivariable logistic regression analysis (Table 4).
In addition, some previous studies revealed medications, baseline glycemic control, and comorbidities might bring about renal impairment 17,19 . Thus we conduct multivariable analysis for renal function which account for these three categories: baseline HbA1c and variables in mediation and comorbidities category that statistical significance level as P value less than 0.1 (P < 0.1) in addition to confounders which had shown a significant correlation. If P value > 0.1, we consider it might play a minor role in incident DPN pathogenesis. Because the relationship between hypertension and antihypertensive drugs is hyperbolic, we choose antihypertensive drugs instead of hypertension. Thus antihypertensive drugs of medication category, cerebrovascular disease of comorbidities category and baseline HbA1c enter the multivariable regression models for baseline eGFR and serum creatinine concentration (Tables 3, 4).

Statistical methods. Descriptive statistics were presented as the mean values ± standard deviation (SD)
and as the numbers with percentages. We used Fisher's exact test or chi-squared test to analyze categorical variables, while the analyses of continuous variables were conducted using independent t-test or paired t-test.
Multivariable logistic regression analyses were carried out to explore the effect of each identified independent variable on DPN. The multivariable regression models included all the confounders and the variables that had shown a significant correlation, and the adjusted odds ratios (OR) with 95% confidence interval (CI) were calculated between the comparison groups. The statistical significance level chosen was P value less than 0.05 (P < 0.05), and all tests were two-sided. All the data were analyzed using statistical package SAS version 9.4 for Windows.

Results
We recruited 681 participants who had T2DM at baseline in 2013. Of these, 116 (17%) who had DPN at baseline and 13 non-T2DM patients were excluded. Thus, 552 were deemed to be eligible to be included in the study. The participants' median age was 59.7 ± 10.7 years, and 60.1% were males. The mean duration of diabetes was 15.2 ± 6.9 years, and the mean level of HbA1c was 7.4 ± 1.3%. Table 1 summarizes their sociodemographic factors, diabetes-related factors, biochemical factors, comorbidities, and medication usage.
Patients' comorbidities at baseline revealed no significant differences between groups, but HTN (75.3% vs. 64.9%, p = 0.08) and CVD (26.0%% vs. 16.9%, p = 0.06) were more common at baseline in patients with incident DPN than in those without it. The DPN and non-DPN groups showed no significant differences in BMI, waist circumference, smoking status, fasting glucose levels, HbA1c levels, OHA and insulin usage, prescriptions of antihypertensive drugs and lipid-lowering drugs, DBP, urine albumin-creatinine ratio (UACR), pathologic high level of TG and LDL, cholesterol nor alanine aminotransferase levels.  Table 4.

Discussion
To our knowledge, this is the first large scaled observational study to investigate risk factors for DPN in a Taiwanese adult population. Using MNSIE for the diagnosis of DPN, we found that participants without DPN at baseline had a 13% cumulative incidence of DPN over the 6 years of follow-up (corresponding with an annual incidence of 2.204%) in a population where the duration of DM was as long as 15.2 ± 6.9 years. The incidence of DPN in our study is comparable with that of a previous longitudinal, large-scale, nationwide, population-based study in Taiwan (n = 37,375, annual incidence of 3.2%) 20 However, it was lower than that reported Western populations 21,22 . This discrepancy might be due to differences in the sample size, ethnicity of the study population (the prevalence of DPN is about 32.1% in the UK 23 and about 23.5% in Taiwan 24 ), diagnostic criteria, and measurement instruments. Apart from these, one of the crucial factors is the baseline duration of DM. One study well established that the prevalence of diabetic neuropathy increased from 8 to 42% in patients with T2DM when patients were monitored for 10 years 25 . Compared with the previous longitudinal study, patients had newly diagnosed DM with a cumulative incidence of 10% over the 13-year follow-up period and an annual incidence of 0.7% 9 . The relatively high cumulative incidence over our 6-year follow-up period might be attributable to the longer baseline duration of DM.
The association between renal function and incident DPN. In our study, baseline renal function was found to be an independent risk factor for DPN, including baseline eGFR (Table 3) and baseline serum creatinine concentration (Table 4), particularly in patients under the age of 65. This finding was inconsistent in patients aged 65 and above, which might be due to the decline of renal function in the aging process. This is Table 2. Risk factors of future DPN in multivariable logistic regression. OR odds ratio, CI confidence interval, SBP systolic blood pressure, eGFR estimated glomerular filtration rate, HDL-C high-density lipoprotein cholesterol. a Multivariable logistic regression was adjusted for all variables in Table 2. To date, the mechanisms of neurotoxicity in T2DM patients with renal impairment remains unclear, but they have been demonstrated in some studies 26,27 . Experimental evidence indicates that renal impairment result in alteration in membrane excitability which is induced by inhibition of the axonal Na + /K + pump. Consequently, it abolishes the direct contribution of the hyperpolarizing pump current to the membrane potential, leading to an accumulation of extracellular K + that causes depolarization 28 . Disruption of these various ionic gradients may affect the Na + /Ca 2+ exchanger, leading to increased levels of intracellular Ca 2+ and axonal loss 29 .
In addition, it is clear from previous research that impaired renal function results in microvascular endothelial dysfunction, even in the early stages of chronic kidney disease. Endothelial injury is caused by various factors, including inflammation, hypertension, diabetes-associated factors, and a uremic milieu 27,30 . Eventually, it leads to neuropathy due to impaired nerve blood flow, epineurial arterio-venous shunting, and reduced nerve oxygen tension 31 .
Other studies examining nephropathy as a risk factor for DPN have been inconclusive 13 . However, it is suggested that the selection of disease markers for renal impairment may be important (for example, eGFR or creatinine), and further investigation is needed. Based on the current study, we recommend that increased Table 4. The association between serum creatinine concentration and future DPN, stratified by age group into age < 65 and ≧ 65. OR odds ratio, CI confidence interval, SBP systolic blood pressure, HbA1c glycated hemoglobin, HDL-C high-density lipoprotein cholesterol. a Each group was adjusted for age, gender, height, weight, SBP, duration of diabetes, HbA1c, creatinine, HDL-C, cerebrovascular disease and antihypertensive drugs. www.nature.com/scientificreports/ serum creatinine concentration or lower baseline eGFR be used as an indicator to enhance the awareness of incident DPN.
Other risk factors of future DPN. After adjustment for potential confounding factors, we also found that a higher risk of DPN was linked with increased age, body weight, duration of DM, and male gender. Our findings are consistent with most previous reports from cross-sectional studies and a meta-analysis of patients with T2DM in Western, Korean, and Taiwanese populations 5,6,32 . Concerning sugar control, previous studies indicated hyperglycemia as a risk factor for the development of DPN 5,8 , but we found no association between baseline HbA1c levels and incident DPN. This is likely explained by low levels of HbA1c at baseline (7.3 ± 1.2% in the no-DPN group and 7.6 ± 1.7% in the incident-DPN group) compared with the levels usually found in previous studies. These data possibly reflect better medication adherence among Taiwanese DM patients 33 compared with worldwide 34 . Our study also showed equally high numbers of hypoglycemic medication prescriptions in both groups. On the other hand, baseline HbA1c was found to be an independent risk factor for DPN in enrolled patients under the age of 65 (Table 4) but not in all ages. This finding might be attributed to the effect of age on the HbA1c. A possible explanation is that elderly individuals encounter physiologically decreased RBC count thus HbA1c is unsuitable for a marker of glycemic control in elderly 35 . In summary, the result implies us that baseline glycemic control might play a role in incident DPN pathogenesis in people under the age of 65 but further research is warranted.
In the current study, increased weight was independent risk factor of incident DPN, but no statistically significant associations with incident DPN were found for BMI and waist circumference. This is inconsistent with previous studies 5,9,10 but previous studies have not identified a consistent list of risk factors related to markers of obesity 10,12 . A possible explanation is that previous investigators did not adequately correct the reference cut-off values and the units for tests. This is not to say that markers of obesity may not be risk factors for DPN, but corrections must first be made for these characteristics in the cut-off values and the units 12 .
In terms of dyslipidemia, we found that serum lipid components had no statistically significant associations with the risk of DPN in T2DM. As stated above, these findings were consisted with some previous studies 36, 37 . In fact, accumulated evidence has shown a correlation between DPN and serum lipid profiles but has shown inconsistent results 38 . The possible underlying mechanisms of dyslipidemia leading to DPN are complex which may include insulin resistance, chronic inflammatory status, oxidative stress induced by elevated LDL, and demyelination 38 . Nevertheless, these mechanisms are mainly reported in preclinical studies [39][40][41] . It is well established that DPN is a multifactorial disease and our findings indicate that lipid metabolism may play a minor role in its pathogenesis.
The major strengths of the current study are its large sample size with long term follow-up, the unselected nature of participants, standardized data collection procedures, and inclusion of several potential risk factors at baseline. But despite these strengths, there are still plenty of limitations. First, our results might not apply to treatment-naïve cohorts of early-stage T2DM. A high proportion of medication prescription might have affected the cardiovascular risk factors. Furthermore, we did not use confirmatory tests such as nerve conduction studies or skin biopsy for DPN diagnosis. However, the diagnosis of DPN is principally a clinical one according to ADA recommendations, and the MNSIE is a sensitive, specific, validated clinical screening tool. Lastly, we included participants from a single hospital, which might limit the generalizability of the results.

Conclusion
Lower eGFR and higher serum creatinine concentration, particularly in people under the age of 65, are predictors of future DPN in Taiwanese people with T2DM. Other risk factors included advanced age, increased body weight, duration of DM, male gender for all ages and HbA1c in enrolled patients under the age of 65 which were compatible with most previous studies. These findings not only confirm the association between renal impairment and future DPN but also provides a commonly available assessment to predict the future DPN. Early detection of risk factors and control of the modifiable factors could enrich therapeutic strategies in clinical practice. Thus, we suggest that the therapeutic strategy for diabetes should provide early management of impaired renal function and prevent overweight. Also, these findings could provide useful information for researchers exploring the underlying mechanisms of DPN and inspire disease-modifying therapies in the future.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. www.nature.com/scientificreports/