Risk factors for neuropathic pain in diabetes mellitus

1. Introduction

According to the International Diabetes Federation, diabetes mellitus (DM) is estimated to affect around 415 million adults worldwide, roughly 8.8% of the adult population, with the figure projected to rise to over 600 million by 2040.²⁰ Regional prevalence varies from 3.2% in Africa to 12.9% in North America. Diabetes mellitus is associated with a number of chronic sequelae and around 50% of people with DM go on to develop polyneuropathy.³⁵ This condition has a variety of clinical manifestations, which are grouped into positive symptoms including dysesthesia (abnormal sense of touch), tingling and itching, and negative symptoms including numbness, muscle weakness, and trouble with balance. Up to 25% of people with diabetic neuropathy (DN) also develop neuropathic pain (NP).³⁹ Neuropathic pain is defined by the International Association for The Study of Pain as “pain directly caused by a lesion or disease affecting the somatosensory system.”^12,22 Symptoms of painful diabetic neuropathy (PDN) include those described above for nonpainful DN with additional “burning,” “electric shocks,” “stabbing,” and “pins and needles” symptoms all being described. Painful diabetic neuropathy is associated with increased distress and poor quality of life compared with nonpainful DN, DM, and the general population³⁸ including depression, anxiety, and sleep disturbance.¹⁵ In addition, an association has been described with reduced productivity and employability at work compared with nonpainful DN.³⁷ The combination of these factors places a large economic burden on patients and health care services,¹⁰ a situation likely to grow steadily worse with the aforementioned projected rise in DM prevalence. This situation is further exacerbated by the fact that 13% of patients with PDN do not report their symptoms to primary care, and 39% of patients with PDN have never received treatment.⁸ Even for those patients who do attend primary and secondary care for their diabetes, pain is not a symptom that is always included in clinical assessments. Furthermore, not all patients with DN develop PDN, and the reasons for this are unclear.

Understanding the risk factors for PDN will go some way to resolving this and will also help to inform management and prevention of this painful condition by health care services. Any factor that increases the risk of DM or DN is likely to be a risk factor for PDN. However, it is the specific nature and magnitude of the risk that remains unclear and is the focus of this topical review.

2. Risk factors

There have been relatively few published studies examining risk factors specifically for PDN in DM. Clinical, environmental, and genetic factors have been shown to be predictive of developing DM and some of these have also been implicated in the development of DN, including age,¹¹ body mass index,^25,28 hypertension,¹³ smoking, and waist circumference²⁸ (Fig. 1). Given the likely overlap of risk factors between DM and DN, it seems reasonable to hypothesize that some of these factors will also influence the development of PDN.

Figure 1.

Schematic of the process from diabetes mellitus to diabetic neuropathy and finally painful diabetic neuropathy. Both diabetes mellitus and diabetic neuropathy have their own set of risk factors, both of which could provide important clues as to the risk factors that contribute to painful diabetic neuropathy.

We conducted a literature search using relevant key words and terms aiming to identify a wide range of studies that investigated risk factors for PDN and to include all the important studies (Table 1). A number of limitations can be identified with these studies as a whole. Most of these studies are cross sectional in nature and therefore unable to establish temporal relationship between patient characteristics/factors and PDN. Some studies report only univariate analysis and are therefore unable to assess intervariable relationships and to identify confounding between variables.^{2,6,8,9,17,23,30} In addition, it is not always clear in the methods and statistical analyses whether PDN or nonpainful DN is being analysed and what control group the PDN subjects are being compared with. In some studies, those in the control group are diabetic participants with nonpainful neuropathy^30,38,41 and in others they are diabetic participants without neuropathy of any form.^{1,2,6,8,11,17} In other studies, it is not possible to determine the nature of the control group from the description of the methods. There was considerable heterogeneity in PDN case ascertainment, with only 6 studies using a validated NP screening questionnaire (the DN4^6,7,17,21,38 or the Leeds assessment of neuropathic symptoms and signs¹¹) with the remainder using nonvalidated questionnaires or clinical examinations. This makes it difficult to assess the sensitivity and specificity of each study to identify PDN cases and to make direct comparisons between studies as effect size estimations and associations are likely to be different. Despite these limitations, some potential risk factors have emerged, including environment, clinical, lifestyle, and genetic factors.

Table 1.

List of studies conducted on predictors of painful diabetic neuropathy and their characteristics.

2.1. Demographic

Two nonmodifiable factors—age^{2,17,21,38,42} and sex^{1,11,17,21,41}—have been specifically associated with PDN, in addition to their known roles as risk factors for DM. Although these are of limited use to clinicians in terms of intervention, they could provide useful clues as to the underlying biological pathways involved and increased awareness of at-risk patients. In particular, the association of PDN with older age (>50 years) is likely to be related to the time it takes for nerve damage and painful symptoms to develop after the onset of DM and the decreased ability of the body to deal with this. Similarly, gender associations may indicate possible subtle differences in biology and psychosocial factors that affect the risk of PDN, something that requires further investigation. It is interesting to note that while 4 of the studies report greater risk in women,^1,11,17,21 1 study reports greater risk in men.⁴¹ This discrepancy in the latter study could be related to the limited statistical analysis, which did not adjust for potential confounding factors. Despite the prevalence of DM varying according to ethnicity, this has not been found as a risk factor so far for PDN.^{1,17,18,21,32,36} One study reported that South Asians were more likely to report painful symptoms than people in other ethnic groups in the absence of clinical neuropathy.¹ Another found an association with pain among people with DM residing in a Gulf state and Lebanon compared with Egypt, but did not analyse ethnic origin.²¹

2.2. Clinical

Clinical and physiological factors associated with PDN are important for clinicians and primary care as they may indicate possibilities for targeted treatment or primary prevention strategies. The clinical diagnosis of the type of DM and the duration since onset of the disease may be particularly relevant. Two studies found an association with DM type in multivariate analysis, with 1 identifying type 1 diabetes (T1D)²¹ and the other type 2 diabetes (T2D)¹ as conferring greater risk of PDN. Differences in case definition and study populations could have contributed to the heterogeneity in these results. A clearer consensus is apparent for DM duration with risk increasing over time since diagnosis.^{2,6,17,21,32,38} Severity of preceding neuropathy has been found to be associated with PDN, but associations with neuropathy duration and comparison with type of (peripheral or sensory) neuropathy have not been found.^9,11,33,36 Most studies included only 1 type of neuropathy in their analysis. A number of clinical factors and comorbidities have been found to be associated with PDN. These include poor glycaemic control and high HbA1c levels,^2,18,36 hypertension,^2,18 retinopathy,² nephropathy,^2,38 cardiovascular disease,^2,42,43 and glycosuria.¹⁸ However, as these conditions are all known complications of DM, it is uncertain from cross-sectional analysis whether these factors are contributing to PDN risk and onset, or simply coexisting factors, perhaps confounded by other factors or with shared aetiology. Biomarkers for the development of PDN can be exploited by providing preventative or diagnostic tests. In this respect, tumour necrosis factor alpha and inducible nitric oxide synthase expression,³⁰ triglycerides, and low high-density lipoprotein cholesterol³⁸ all show promising associations but require replication to be confident in their role in disease pathogenesis.

2.3. Lifestyle

Behavioural and social circumstances are important lifestyle aspects that patients can theoretically influence and act on, with greater or less practical difficulty. In particular, some physical characteristics known to be associated with DM and DN are also implicated in PDN. Body mass index has been clearly linked to PDN, particularly in the form of obesity (≥30 kg/m²),^21,33,38 while in another study, weight was reported independently of height and found to be significantly associated with PDN, although this was attenuated in multivariate analysis.⁴² A related study also found a positive correlation with increased waist circumference and high levels of physical activity and risk of PDN.⁴³ Despite being included in the analyses in most of the studies,^{1,2,4,8,17,18,21,32,33,38,42,43} smoking and alcohol consumption have not been specifically associated with PDN. Psychological factors have also been widely reported in the context of general chronic pain, but its relationship with PDN is less clear. Increased depression, anxiety, enjoyment of life, and social relationships are associated with PDN, but without prospective studies and longitudinal analysis, the temporal relationship cannot be established.^7,15

2.4. Genetics

Numerous published studies have found that both T1D and T2D have a heritable component^3,19,29,31 and genetic studies have been conducted in DN,^24,34 although heritability studies have yet to be conducted. A heritable component to PDN has been hinted at in 1 study, which found that 56% of participants also reported a family member with the condition.¹⁴ Two PDN genome-wide association studies have been conducted, both in the same Scottish diabetic cohort, but with slightly differing phenotype definitions. The first used a positive monofilament test combined with a prescription history of at least 1 from 5 recommended NP medicines. This yielded an association at chromosome 8p21.3 near the gene GFRA2 and estimated the narrow-sense heritability (proportion of phenotypic variance explained by additive genetic variance) to be 11%.²⁷ The second used the same definition but without the monofilament test and aimed to capture associations that may have been missed in the previous study, using a less specific cohort. This found sex-specific associations at chromosome 1p35.1 in the ZSCAN20/TLR12P gene regions in females and chromosome 8p23.1 next to HMGB1P46 in males.²⁶ The narrow-sense heritability was 30% in males and 14.7% in females. In both of these studies, controls were defined as patients who had not previously been prescribed any of the 5 NP medicines or other medicines which are predominantly prescribed for other conditions but are also known to be prescribed for NP. Two separate candidate-gene studies have been conducted and have reported associations with PDN. The first was in the sodium channel gene SCN9A, which is expressed in dorsal root ganglia, using a combination of numerical rating scales and clinical examinations to compare PDN with healthy controls.²³ The second was in the opioid receptor gene OPRM1 using a visual analogue scale for pain intensity and a grading system for pain occurrence during daily activities.⁵ All 4 of these studies require independent replication, and further studies are required to establish the extent to which genetics contribute specific risk of PDN and the mechanism of this contribution.

3. Conclusions

Despite the limited number of studies reporting specific predictors for PDN, clear similarities are emerging with the known general risk factors for both DM and DN. These include clinical factors such as diabetes type and duration and lifestyle factors such as body mass index and waist circumference, some of which are not easily or not at all modifiable. Although further work is needed, this suggests that PDN is a manifestation of longer-lasting and more severe diabetes and certainly that it requires specific testing and diagnosis in routine diabetic care. However, there are likely to be factors, among those with DM (with and without DN) that confer a greater risk of PDN, and these require further exploration. It should be noted that while the published literature (and this topical review) has mainly explored NP arising from DN, NP can also arise in the diabetic population through other causes, for example, sciatic neuralgia and carpal tunnel syndrome. The influence of diabetes on the development of NP in these conditions is an area that requires further investigation. There is clear evidence to suggest PDN has a negative impact on quality of life; however, the extent to which the reverse is true—bidirectional aetiology—is currently unknown. Future studies need to be conducted in a longitudinal manner or as clinical trials to establish the temporal relationship between variable and disease, particularly with respect to identifying specific PDN risk factors in T1D and T2D patients. The previous point can be further strengthened by running Mendelian randomization studies, something that has been used in DM.¹⁶ Mendelian randomization studies establish causal relationship by comparing 2 groups of individuals with and without a genetic marker known to influence the variable being studied. As genotype assignment is random and not subject to confounding typically found in epidemiological studies, a higher prevalence of disease in the group with the marker implies causality. However, we would first need clearer evidence to identify genetic factors associated with PDN. Finally, greater clarity is needed in specifying whether painful or nonpainful DN is being analysed. This can be enhanced by forming a consensus on PDN phenotype definition, to enable studies to be more comparable with one another. This is something that has been addressed for NP generally and could also be applied to PDN.⁴⁰ This would make replication of results more likely and brings the added potential of being able to perform meta-analyses in the future. All these limitations will be addressed in the DOLORisk study (http://dolorisk.eu/), a European consortium which aims to identify risk factors for NP.

Conflict of interest statement

B. H. Smith is a member of the DOLORisk consortium which is funded by the European Commission Horizon 2020 (ID: 633491) and is partly supported by this grant. He has received, on behalf of his institution, occasional lecture and consultancy fees from Pfizer Ltd, Napp Pharmaceuticals, Grunenthal and Eli Lilly. H. L. Hébert and A. Veluchamy are supported by DOLORisk. N. Torrance has no conflicts of interest to declare.