High resolution ultrasound in subclinical diabetic neuropathy: A potential screening tool

Abstract

Introduction

Detection of subclinical neuropathy can aid in triage, timely intervention and dedicated care to reduce disease progression and morbidity. High resolution sonography has emerged as a promising technique for evaluation of peripheral nerves. The aim of the present study was to assess the utility of high resolution sonography in screening diabetic patients for subclinical neuropathy.

Methods

A total of 70 adult patients with type 2 diabetes mellitus and 30 controls were enrolled; those with clinical features of neuropathy constituted the diabetic polyneuropathy group and those without symptoms/normal nerve conduction the non-diabetic polyneuropathy group. After institutional ethical committee approval and informed consent, high resolution sonography was performed by two musculoskeletal radiologists. Nerves studied were median (elbow and wrist), ulnar (cubital tunnel and Guyon’s canal), common peroneal (fibular head) and posterior tibial nerve (medial malleolus).The size (cross sectional area), shape, echogenicity and morphology of nerve were assessed and compared between the groups.

Results

The mean cross sectional area of all nerves was significantly higher both in diabetic polyneuropathy and non-diabetic polyneuropathy group compared to controls (p value < .001). Common peroneal nerve cross sectional area of 4.5 mm² had the highest sensitivity (93%) and specificity (86%) for detecting nerve changes in the non-diabetic polyneuropathy group. The nerves were more rounded, hypoechoic and had an altered morphology in both study groups.

Conclusion

Presence of sonographic nerve changes in asymptomatic diabetics depicted that morphological alterations in nerves precede clinical symptoms. High resolution sonography detected nerve changes with a good accuracy, and thus, can be a potential screening tool for detection of subclinical diabetic polyneuropathy.

Introduction

Diabetic polyneuropathy (DPN) is the most common form of neuropathy and is responsible for 50–75% of non-traumatic amputations.^1–4 It accounts for more hospitalisations than all other diabetic complications combined. Painful DPN is often difficult to treat and is associated with reduced quality of life, poor sleep, depression and anxiety.⁵ Neuropathy management including adequate podiatric care in these patients adds to the economic burden of the national health system. Therefore, early detection of nerve dysfunction is important, to triage patients for referral and to initiate appropriate and timely management for these patients and thus prevent the major complications.

DPN presents primarily with characteristic clinical symptoms which include numbness, pins and needle sensation, tingling sensation and signs such as loss of ankle reflex and vibration perception; diagnosis being confirmed by nerve conduction studies (NCSs).

In recent years the focus has shifted to early detection of diabetic neuropathy in the subclinical stage. Various screening methods being used in asymptomatic diabetics include tuning fork test, monofilament testing and NCSs. The tuning fork and monofilament tests have good specificity and accuracy; however, their sensitivity is low.⁶ NCSs are time-consuming, painful, slightly invasive and often not well tolerated for repeated evaluations.^7,8

High resolution sonography (HRUS) has emerged as a promising technique for evaluation of peripheral nerve disorders as it provides good morphological evaluation of the nerves. It has been shown to have a higher sensitivity than MRI (93% versus 67%) and an equivalent specificity (86%) for detection of multifocal peripheral nerve disorders.⁹ Though there are studies on sonographic evaluation of nerves in patients with diabetic neuropathy, its role in subclinical assessment has remained largely unexplored.

The aim of this study was to assess the utility of HRUS, a non-invasive readily available modality, in screening diabetic patients for subclinical neuropathy.

Methods

A prospective case–control study was conducted at a tertiary care centre in New Delhi over a one-year period. A total of 70 adult patients with type 2 diabetes mellitus (as per American Diabetes Association criteria 2015) and 30 age- and sex-matched controls were enrolled.¹⁰ All patients provided informed written consent. Institution ethical committee approval was obtained.

Demographic information of age, sex, BMI, type and duration of disease was obtained. The neurological complaints such as numbness, burning feet, pins and needle sensation were recorded and a detailed systemic neurological examination was performed by an experienced clinician, which included testing for light touch, pin prick, vibration sense, knee and ankle jerks. Diabetic neuropathy was diagnosed, based on presence of any two of the following three features (as proposed by the Diabetic Neuropathy study group of Japan¹¹):

1. Subjective symptoms in bilateral lower limbs or feet

2. Loss of or decreased ankle jerk

3. Decreased vibration perception, assessed using a C128 tuning fork and bilaterally measured at the medial malleoli.

Relevant laboratory investigations such as blood glucose levels, HbA1c, lipid profile, vitamin B12 levels and thyroid profile were obtained.

NCSs were performed in all diabetic patients by a physician using a Dantec workstation (Dantec Dynamics, DK-2740, Skovlunde, Denmark) and at a standard room temperature of 25°C. Motor studies were obtained in median, ulnar, common peroneal and posterior tibial nerves and sensory studies were done in median, ulnar and sural nerves. Presence of an abnormality (reduced nerve conduction velocity, altered amplitude or increased latency) in two or more nerves in electrophysiological testing was also grouped as DPN.

Patients diagnosed with other conditions known to cause peripheral neuropathy like leprosy, vitamin B₁₂ deficiency and those taking drugs which improve/mask/aggravate the normal course of neuropathy like pyridoxine, hydralazine, isoniazid, etc. were excluded from the study. Patients with trauma/surgery in the area of the nerve, those with entrapment neuropathies, radiculopathies/disc disease affecting the nerves to be studied were also excluded.

Among the 70 diabetic patients, 41 subjects who met the inclusion criteria were categorised as DPN group while the rest of 29 subjects were categorised as non-DPN group. The control group comprised 30 age- and sex-matched healthy volunteers.

Nerve sonography was performed on all subjects by a radiologist with more than 10 years’ experience in musculoskeletal imaging. Another radiologist experienced in musculoskeletal imaging also recorded the cross sectional area (CSA) of all the nerves from the stored cine clips. They were blinded to the group status of the subjects. Ultrasound of the peripheral nerves was performed using a GE Logic P6 Pro Colour Doppler (Wipro GE Ltd) or a Philips HD7XE (Philips, India Ltd) with high resolution linear transducers of frequency range 7–12 MHz. The nerves examined were median and ulnar in the upper limb and common peroneal and posterior tibial in the lower limb. To ensure standardisation, predefined sites were chosen for assessment of nerves based on previous studies. These sites were identical in the control and study groups. To avoid anisotropic artefact, the transducer was kept perpendicular to the surface. Appropriate depth setting was employed for individual target nerves with precise location of focal zone at the level of examined nerve and gain settings were adjusted to optimise image contrast.

Median nerve examination was performed with the patient supine and arm by the side. The median nerve was traced from above elbow level, where it was seen to lie medial to the brachial artery, down to the carpal tunnel at the wrist. The measurements were recorded at two locations, 5 cm above the elbow and at the inlet of the carpal tunnel.

Ulnar nerve examination at the wrist was done with the patient in the same position and with hand supinated. The nerve was evaluated in transverse plane at the Guyon’s canal level, where it was seen to lie lateral to the pisiform and medial to the ulnar artery. With the patient’s arm positioned across the chest, the transducer was placed in the transverse plane over the bony protuberance of the olecranon and medial epicondyle and the ulnar nerve was examined in the cubital tunnel between these two points.

For evaluation of the common peroneal nerve the patient was rolled on the contralateral side, the transducer was placed in coronal plane with the fibular head in view, then its proximal end was rotated posteriorly to bring the biceps femoris tendon into view. From this position the transducer was moved posteriorly and the hypoechoic common peroneal nerve was seen in long axis. Rotating the transducer through 90° allowed the nerve to be visualised in transverse.

For posterior tibial nerve assessment, the patient was seated with the plantar surface of the foot rolled internally or the patient was supine with the foot rotated slightly laterally. The transducer was positioned transversely behind the medial malleolus to view the tibialis posterior flexor digitorum longus tendon, the posterior tibial vessels and the tibial nerve. Measurements were recorded at midpoint of medial malleolus level.

All nerve measurements were recorded in the transverse planes. For measuring the CSA, the transducer was placed perpendicular to the nerve with minimal probe pressure, and measurement performed from inner border of echogenic epineurium for consistency using the tracing method. Multiple other qualitative and quantitative parameters were evaluated and recorded as stated below:

Qualitative parameters:

• Shape: round/oval

• Echogenicity (subjective assessment): normal/hypoechoic

• Nerve morphology (based on fascicular pattern): normal/altered

Quantitative parameters:

• CSA (mm²)

• Percentage hypoechoic area of nerve (using ImageJ software)

With the use of ImageJ software, the amount of the hypoechoic area falling below the threshold echo intensity was calculated as detailed by Watanabe et al.¹² ImageJ is a public domain Java-based image processing program developed at the National Institutes of Health and the Laboratory for Optical and Computational Instrumentation (LOCI, University of Wisconsin). It can calculate area and pixel value statistics of user-defined selections. Refer to the flowchart for this evaluation process:

All the above parameters were recorded on predefined pro formas and compared among the three study groups. Data were analysed using the SPSS software package, version 20.0 (SPSS Inc., Chicago, Illinois, USA). Quantitative data were expressed using range, mean, SD and median, whereas qualitative data were expressed as frequency and percentage. Quantitative parameters between the three groups were compared by one way ANOVA followed by Tukey’s test and other quantitative parameters between two groups by unpaired t-test. All qualitative parameters were analysed using Chi square test/Fisher’s exact test. Pearson correlation between quantitative parameters was obtained. Receiver operating characteristic (ROC) curves were obtained for determining sensitivity and specificity and cut off value. P value <0.05 was considered statistically significant.

Results

The baseline demographic data of the three study groups are presented in Table 1. The majority of patients in all three groups were 40–60 years old and there was no significant difference between age and sex among the three groups (p = 0.080).

Table 1.

Comparison of demographic characteristics among various study groups.

Variable	DPN group	Non-DPN group	Control group	p-Value*
N	41	29	30
Agea (years)	53.43 ± 7.70	48.72 ± 8.46	51.26  ± 9.62	.080
Sex, male/female (%)	49/51	38/62	53/47	.078
Heighta (cm)	158.4 ± 7.48	151.9 ± 8.95	156.1 ± 9.48	.009
Weighta (kg)	66.95 ± 10.05	62.18 ± 9.81	61.5 ± 5.99	.023
BMIa (kg/m2)	26.72 ± 3.92	27.13 ± 4.47	25.46 ± 2.50	.198
Duration of diseasea (years)	9.6 ± 4.3	6.5 ± 2.58	NA	.002
HBA1ca	8.9 ± 1.50	8.6 ± 1.64	NA	.499
Fasting sugar (mg %)	161 ± 54.26	150.41 ± 41.41	87 ± 8.58	<.001
Post prandial sugar (mg %)	250 ± 60.62	246 ± 54.11	112 ± 6.85	<.001

BMI: body mass index; DPN: diabetic polyneuropathy; HBA_1c: glycated hemoglobin (A1c).

^aValues depict mean +/- standard deviation; *DPN versus controls; p value <0.05 taken as significant.

The BMI of the diabetic population was marginally higher than that of control participants; the HbA1c value was also slightly higher in patients with DPN than those without DPN but these differences were not statistically significant.

Nerves were easily visualised at all sites after a short learning curve and the average scanning duration was 10–15 minutes per patient.

Nerve size

Mean CSA of median nerve in the DPN group was 10.32 ± 2.61 mm² at wrist and 8.71 ± 2.11 mm² at elbow. In the non-DPN group it was 9.73 ± 2.81 and 8.41 ± 2.51 mm² at wrist and elbow, respectively, while in the control group it was 5.81 ± 1.13 and 5.21 ± 1.12 mm². There was a significant increase in the median nerve CSA in both the DPN group and non-DPN group in comparison with that of controls (p < 0.001) (Figures 1 to 3). Though the median nerve CSAs were higher in the DPN group compared to the non-DPN group at both sites, this difference was not statistically significant.

Figure 1.

Box-and-whisker plot shows comparison of CSA of the median nerve at the wrist among various study groups. Bottom and top edges of box represent 25th and 75th percentiles, horizontal line represents the median, and error bars delimit extent of 10th and 90th percentiles. Statistical significant difference was observed in the non-DPN versus control group and DPN versus control group. There was no significant difference between non-DPN and DPN group. P < 0.5 taken as significant. DPN: diabetic polyneuropathy.

Figure 2.

Sonogram depicting increased CSA, hypoechogenicity and altered morphology of median nerve (at wrist) in a 42-year-old female with a seven-year history of type 2 diabetes and symptoms of diabetic neuropathy (DPN group) (a); 47-year-old female with a three-year history of type 2 diabetes without symptoms of diabetic neuropathy (non-DPN group) (b); compared to a 49-year-old female control (c). CSA: cross sectional area.

Figure 3.

Sonogram depicting an enlarged, rounded posterior tibial nerve with altered morphology in a 51-year-old female with a seven-year history of type 2 diabetes and symptoms of diabetic neuropathy (DPN group) (a); 53-year-old female with a five-year history of type 2 diabetes mellitus without symptoms of diabetic neuropathy (non-DPN group) (b); compared to a 48-year-old female control (c). CSA: cross sectional area.

CSAs of ulnar nerve was also significantly higher in the DPN group and non-DPN group compared to controls both at elbow (5.92 ± 2.12, 4.84 ± 1.41 mm² versus 3.11 ± 0.81 mm²) and at Guyon’s canal level (4.51 ± 1.61, 3.91 ± 1.41 mm² versus 2.41 ± 0.51 mm²).

Similarly in the lower limb the mean CSAs of common peroneal nerve and posterior tibial nerve were significantly increased in the DPN group compared to the controls (8.13 ± 2.71 mm² versus 3.71 ± 1.21 mm² in common peroneal and 6.31 ± 3.71 mm² versus 3.01 ± 0.61 mm² in posterior tibial nerve). In the non-DPN group also the CSAs of all these nerves were higher than that of the controls (p value <0.001). Interestingly, CSAs in the DPN group were not significantly higher than the non-DPN group (Table 2).

Table 2.

Comparison of cross sectional area (CSA) of the examined nerves among the three study groups.

	DPN group	Non-DPN group	Control group
	N = 41	N = 29	N = 30
Nerve sites	Mean ± SD	Mean ± SD	Mean ± SD	p-Value
Median
Elbow	8.71 ± 2.11	8.41 ± 2.51	5.21 ± 1.12	<.001*,€.763a
Wrist	10.32 ± 2.61	9.73 ± 2.81	5.81 ± 1.13	<.001*,€.552a
Ulnar
Elbow	5.92 ± 2.12	4.84 ± 1.41	3.11 ± 0.81	<.001*,€.021a
Wrist	4.51 ± 1.61	3.91 ± 1.41	2.41 ± 0.51	<.001*,€.215a
Common peroneal
Fibular head	8.13 ± 2.71	7.72 ± .026	3.71± 1.21	<.001*,€.777a
Posterior tibial
Medial malleolus	6.31 ± 3.71	4.92 ± 1.71	3.01 ± 0.61	<.05*,€.066a

DPN: diabetic polyneuropathy.

*DPN versus controls; ^€Non-DPN versus controls; ^aDPN versus non-DPN; p value <0.05 taken as significant.

The area under ROC curve was greater for the median nerve at carpal tunnel and elbow, ulnar nerve at Guyon’s canal and elbow, common peroneal nerve at fibular head and posterior tibial nerve at the ankle. The optimal cut off value for diagnosis of DPN was 8.5 mm² for median nerve at the wrist and 3.5 mm² for ulnar nerve at Guyon’s canal in the upper limb. In the lower limb nerves, the relevant cut off was 5.5 mm² for common peroneal at fibular head and 4.5 mm² for posterior tibial at ankle. The common peroneal nerve with cut off of 5.5 mm² had the highest sensitivity (80%) and specificity (96%) for detecting nerve changes in the DPN group (Figure 4).

Figure 4.

The ROC curve analysis for the identification of DPN in symptomatic patients by measurement of CSA in cm² of the nerves as depicted. AUC: area under curve; CPN: common peroneal nerve; DPN: diabetic polyneuropathy; PTN: posterior tibial nerve; ULN: ulnar nerve.

Similarly the optimal cut off values in the non-DPN group (asymptomatic group) were 7.5 mm² for median nerve at the wrist and 3.5 mm² for ulnar nerve at Guyon’s canal, 4.5 mm² for common peroneal nerve at the fibular head and 3.5 mm² for posterior tibial nerve at the ankle. The common peroneal nerve with cut off of 4.5 mm² had the highest sensitivity (93%) and specificity (86%) for detecting nerve changes in the subclinical group (Figure 5).

Figure 5.

The ROC curve analysis for the identification of DPN in asymptomatic patients by measurement of CSA in cm² of the nerves as depicted. AUC: area under curve; CPN: common peroneal nerve; DPN: diabetic polyneuropathy; PTN: posterior tibial nerve; ULN: ulnar nerve.

Shape

It was observed that among the control group the nerves were oval at most of the sites though the percentages were variable. It was noted that there was rounding of nerves in diabetics in significantly higher number at most sites. In the median nerve as seen at the elbow, rounding was found in 80.5% cases in the DPN group, 68.9% in the non-DPN group and 50% in the control group (Table 4 and Figure 3). Similarly increased rounding was noted in the ulnar nerve at the wrist (78% versus 43%), common peroneal at fibular head (78% versus 13%) and posterior tibial (65% versus 23%) at the ankle level in DPN versus control group, respectively. Similar increased rounding, though in fewer cases, was observed in non-DPN group vis-à-vis controls.

Echogenicity

On visual evaluation nerves were hypoechoic in the DPN and non-DPN group compared to the control group. At the elbow, in 73.17% of individuals of the DPN group the ulna nerve looked hypoechoic, whereas in the non-DPN group the percentage was 68.97%, and just 63.34% in controls. At the wrist 26.67% of controls had hypoechoic nerve appearances. In DPN group and non-DPN group, the ulnar nerve at the wrist was hypoechoic in 43.91 and 24.14%, respectively. It was observed that nerves demonstrated hypoechogenicity in a higher percentage of cases of DPN and non-DPN individuals in comparison to the control group, though this difference was not statistically significant at most sites (Table 3). Only the median nerve at the wrist showed hypoechogenicity in a significantly higher number of patients in diabetics, i.e. 58% in DPN compared to 16.6% in the controls (p value <0.001) (Figure 2).

Table 3.

Comparison of nerve echogenicity (subjective and objective) among the three study groups.

	Percentage reduction in echogenicity (using ImageJ)				Percentage of patients with reduced nerve echogenicity (subjective)
Nerve sites	DPN groupN = 41	Non-DPN groupN = 29	Control groupN = 30	p-Value*	DPN groupN = 41	Non-DPN groupN = 29	Control groupN = 30	p-Value*
Median
Elbow	66.23	67.54	66.62	.899	46.35	41.38	30.00	.375
Wrist	80.34	82.10	75.71	.688	58.54	65.51	16.67	<.001
Ulnar
Elbow					73.17	68.97	63.34	.676
Wrist	76.57	70.65	72.56	.267	43.91	24.14	26.67	.151
Common peroneal
Fibular head	68.99	70.73	64.48	.217	24.40	24.14	20.00	.897
Posterior tibial
Medial malleolus	70.33	69.95	70.76	.983	43.91	41.38	33.34	.657

DPN: diabetic polyneuropathy.

*DPN versus controls; p value <0.05 taken as significant.

On quantitative analysis of percentage hypoechoic area (using ImageJ software) no significant echogenicity difference was observed between the three groups. The mean percentage hypoechoic area within the ulnar nerve was found to be 72.56% in controls, 70.65% in the non-DPN group and 76.57% in the DPN group (Table 3).

Morphology

An altered nerve morphology was observed in a significantly higher percentage of patients for the median nerve at the wrist both in the DPN and non-DPN group compared to the control (<0.001) (Figures 2 and 3). No significant variation was seen at other sites between the three groups studied (Table 4).

Table 4.

Comparison of nerve shape and sonographic morphology among the various study groups.

	Altered morphology %				Round shape (%)
Nerve sites	DPN groupN = 41	Non-DPN groupN = 29	Control groupN = 30	p-Value*	DPN groupN = 41	Non-DPN groupN = 29	Control groupN = 30	p-Value*
Median
Elbow	39.02	51.72	33.33	.220	80.49	68.97	50.00	.025
Wrist	58.53	65.51	16.66	<.001	26.82	10.34	10.00	.093
Ulnar
Elbow	39.02	10.34	33.33	.918	63.41	58.62	36.66	.069
Wrist	14.63	31.03	10.00	.220	78.05	69.96	43.33	.009
Common peroneal
Fibular head	36.58	27.58	26.66	.062	78.04	55.17	13.33	.001
Posterior tibial
Medial malleolus	31.70	34.148	30.00	.983	65.85	62.06	23.33	.001

DPN: diabetic polyneuropathy.

*DPN versus controls; p value <0.05 taken as significant.

Interobserver agreement was excellent for CSA measurement for all nerves (ICC ≥ 0.9) (Figure 6).

Figure 6.

Interobserver reliability calculated with intraclass correlation coefficient of the CSA of various nerves. DPN: diabetic polyneuropathy; ICC: intraclass correlation coefficient.

Discussion

The pathophysiology of DPN is multifactorial and involves genetic, environmental, behavioural, metabolic, neurotrophic and vascular factors. Though the exact mechanism of DPN is uncertain it has been proposed that oxidative stress triggered by microangiopathy in the nerve and accumulation of hyperglycaemia-induced sorbitol in Schwann cells are the key pathological processes that induce demyelination and axonal nerve damage.^12,13 These changes are manifested as morphological alterations in nerves.

In the diagnosis of diabetic neuropathy, neurophysiological tests have a complementary role and provide information about dysfunction of affected nerves; however, they offer no insight into pathological/morphological changes. In recent years, technological advancement in transducer technology and state of art sonographic machines have made possible the evaluation of soft tissue structures such as nerves with good accuracy. Ultrasonography is an inexpensive, reproducible and comfortable technique, which has been proposed as an alternative method for detecting neuropathies.¹⁴ The present study was undertaken to assess the utility of HRUS as a non-invasive, readily available modality for screening diabetic patients for subclinical neuropathy.

We evaluated CSA, shape, echogenicity and nerve morphology. Most of the previous investigators have evaluated only the nerve size on sonography and none has simultaneously examined multiple morphological features of peripheral nerves in diabetic patients.

Our study demonstrated a pronounced increase in CSA for all the examined nerves in diabetic patients when compared to controls .This is in agreement with several previous studies.^7,12,14–21 Nearly all researchers except Hobson et al.²² reported an enlarged CSA on sonography in peripheral nerves of diabetic patients with DPN in comparison with a control population. However, it was observed that all the nerves in our study were smaller in size in controls as well as diabetic patients compared to Western and other Asian studies. A plausible explanation could be racial and ethnic differences among the study groups or possibly the lower BMI in our study population. The mean BMI of patients in our study was 26.72 ± 3.92 in the DPN group, 27.13 ± 4.47 in the non-DPN and 25.46 ± 2.50 in the control group, which were lower than the study groups of some authors; Kelle et al.¹⁷ reported a mean BMI of 30.22 ± 3.24 in controls and 30.43 ± 3.62 in the diabetic patients.

The CSA values of controls in the present study were closer to those reported by Bathala et al.^23,24 in a study from Southern India on upper limb nerves in a normal population. In their study the mean CSA of the median nerve was 7.5 and 5 mm² at the wrist and forearm, respectively, while the ulnar nerve CSA was 3.8 mm² at the wrist. The slightly lower value of the mean CSA in ulnar nerve in our study could be due to difference in the site of measurement. In their study the ulnar nerve was evaluated at the wrist crease, whereas in our study it was evaluated more distally at the Guyon’s canal level (Table 1).

To obtain optimal possible cut off values of sonographic nerve CSAs for the diagnosis of DPN, ROC curves were generated. All nerves revealed a high AUC (>90%) suggesting that ultrasound has a good diagnostic accuracy for detection of DPN. Sonographic criteria for diagnosis of neuropathy have been proposed by several authors^16,19,25 based on CSA values on nerve sonography using ROC curves. However, only one¹⁶ has calculated the cut offs based on ROC curves in multiple nerves in a single study.

The cut off for median nerve at the wrist (8.5 mm²) in our study was close to those reported by Kang et al.¹⁶ and Watanabe et al.²⁵ (9.4 and 8 mm², respectively). Cut off values for the common peroneal were found to have a high sensitivity and specificity in our study (cut off of 5.5 mm² with 80% sensitivity and 96%). The fibular nerve has been described as being one of the first nerves affected in DPN, so this site could represent a helpful nerve ultrasound tool to recognise DPN.¹⁸

However, the cut off for the ulnar, common peroneal and posterior tibial nerve in the present study had a high sensitivity and specificity in our population, and were lower than those found in other study groups^12,16 (common peroneal nerve 11.5 mm²; posterior tibial nerve 14.5 mm²). A possible reason could be the demographic differences among our study population with those conducted in other countries as described above.

Often repetitive microtrauma/stress at the compression site may lead to nerve enlargement. The strength of our study is that assessment of nerves was done at both compression and non-compression sites, thus removing the potential bias. The nerves were enlarged at both these sites. Diabetes is a systemic disease and causes diffuse involvement of nerves and changes observed in multiple nerves, as in our study, increases the reliability of it being the causative factor.

Because the observed CSA in our study at compression sites (carpal tunnel, cubital tunnel and tarsal tunnel) was found to be higher in many patients without any symptoms, the use of these parameters for sonographic diagnosis of compressive neuropathy should be done with caution in the diabetic population. Possibly, a ratio of CSA of median nerve at carpal tunnel to that in the forearm should be adopted for diagnosis of compressive neuropathy like carpal tunnel syndrome in patients with diabetes. The increased nerve CSA in asymptomatic subjects in present studies can possibly explain the susceptibility of the diabetic population to nerve compression syndrome.

Most studies in the literature have assessed changes in diabetic patients symptomatic for neuropathy. Only a few studies are available^19,26 where a non-neuropathic diabetic group has been compared with normal controls for multiple nerves, as done in our study. It was noted that at most sites the mean CSA in the non-DPN group was significantly higher than our control population. The difference between the DPN and non-DPN group CSA at most sites had no statistical significance. As this patient group lacked relevant symptoms, changes in nerves in this group indicate that morphological nerve changes precede the symptoms. Ishibashi et al.²⁶ also stated that since CSA, hypoechoic area and maximum thickness of the nerve fascicle of diabetic patients precede diabetic neuropathy, these changes (as detected by HRUS) might play a hierarchical role in diagnosing diabetic neuropathy in peripheral nerves.

This enlargement probably indicates subclinical neuropathy and sonography can assess the prevalence of subclinical neuropathy within the diabetic population. However, a prospective longitudinal study is required to determine whether the morphological changes detected by HRUS predict the changes in the neurophysiological tests.

In the present study we assessed, both subjectively and objectively, the internal echogenicity of the peripheral nerves. On subjective assessment, the peripheral nerves were more hypoechoic in the DPN group compared to the control group; however, statistical significance could be established only in the median nerve (at wrist). A rationale for the decrease in echogenicity could possibly be attributed to the increased water content/increased sorbitol in the peripheral nerve of diabetic patients due to metabolic derangements associated with chronic hyperglycaemia. This leads to nerve swelling and also decreased echogenicity.¹² Interestingly, the nerves in asymptomatic diabetics were also more hypoechoic when compared to the control group.

We made an attempt to objectively assess the echogenicity of the nerves and did a quantitative assessment using the ImageJ method of computer quantification of echogenicity as proposed by Watanabe et al.¹² The nerves were found to be hypoechoic; however, statistical significance could not be established. Watanabe et al.²⁵ found a significant difference in the percentage hypoechoic area of the median nerve in patients with polyneuropathy in comparison to the normal controls. However, another study did not find any statistical difference between nerve echogenicity in DPN patients and the control population on using the same criteria.²⁶ These authors then arbitrarily selected <70% of the mean as a cut off and reported significantly decreased echogenicity in DPN patients compared to controls. They further stated that the technique and cut off values used by Watanabe and themselves are innovative and arbitrary and no rational definition of a hypoechoic area can be given on these parameters. Since it has been attempted by only a single author, it is not a validated technique. Albeit an interesting and innovative method which may open new doors towards objectively classifying echogenicity, more work is required in this context.

The shape of the nerve was found to be rounded in a significantly larger percentage of cases among diabetics (both DPN and non-DPN) as compared to non-diabetic subjects in the present study. This parameter is hitherto unexplored and none of the prior studies have compared the shape of peripheral nerves between diabetics and control population groups. It was interesting to note that rounding was significantly lower at the compression sites (carpal tunnel and cubital tunnel) possibly since these areas are tight compartments not permitting rounding despite increase in size.

It has been previously reported that altered nerve morphology on the account of changes in the fascicular size is one of the features of DPN which contributes to the hypoechogenicity and can act as a surrogate morphological marker for diabetic neuropathy severity.²⁶ The morphology of the nerve was seen to be altered in DPN patients and non-DPN patients in our study; some fascicles appeared larger while a few showed distortion. However, no consistent pattern could be observed in most of the nerves. A more detailed objective evaluation with higher frequency transducers (18 MHz or higher) may produce better results.

Limitations in our study include that we did not attempt the monofilament testing in these patients which is an important screening tool for DPN. Also the sample size is small and a larger study is warranted to validate these findings. The sural nerve is one of the earliest nerves affected in diabetic neuropathy and gives a subclinical sensory nerve deficit on NCS.²⁰ Thus, a drawback of the present study was not evaluating the sural nerve, though we did study the posterior tibial nerve at ankle which is comparable in length to the sural nerve and is likely to be similarly affected in this length dependent process.²⁰

In conclusion, patients with type 2 diabetes mellitus have sonographically detectable morphological alterations in nerves even prior to the onset of diabetic neuropathy and after development of neuropathy these changes progress in severity. High resolution ultrasound emerges as a practical tool for morphological analysis of peripheral nerves in the diabetic population and has the potential to be used as a screening tool for early diagnosis of DPN at the subclinical stage.

Contributors

Anupama Tandon, Tamanna Khullar and Siddharth Maheshwari: Substantial contributions to conception and design, or acquisition of data, or analysis and interpretation of data; drafting the article or revising it critically for important intellectual content; and final approval of the version to be published.

Shuchi Bhatt and Shiva Narang: Drafting the article or revising it critically for important intellectual content and final approval of the version to be published.