Demographic and lifestyle information.

The study records demographic information, including age, gender, weight, height, and smoking and alcohol consumption.

Characterisation of pain – pain location, intensity, quality, analgesics.

We assess the presence and duration of pain, including dysaesthesia. Participants are asked to indicate where they feel pain on their body using a list of body sites and a body map. They identify all areas where they have experienced pain over the past three months and mark the location that bothers them the most.

Pain intensity is measured using different scoring systems such as the Chronic Pain Grade [43] over the past three months and the Brief Pain Inventory [44]. An additional item asks about the average pain experienced over the past seven days. Pain descriptors are characterised by the DN4 (Douleur Neuropathique en 4 questions) [45], and the Neuropathic Pain Symptom Inventory [46]. These questionnaires use an 11-point numerical scale to quantify severity of pain and symptoms.

The Michigan Neuropathy Screening Instrument is a screening tool for diabetic neuropathy and asks participants about symptoms related to diabetic neuropathy, such as pain, numbness, tingling, and loss of sensation in the feet and legs [47].

We record details of pain medication (e.g., paracetamol, gabapentin), dose, analgesic relief, and adherence to medication.

Pain interference, quality of life and psychological variables.

The Patient-Reported Outcomes Measurement Information System (PROMIS) questionnaires are used to assess several psychological and psychosocial variables [48]. These include depression, anxiety, sleep disturbance, fatigue and pain interference. Other psychological questionnaires include Ten-Item Personality Inventory [49], and assessments of Pain Related Worrying [50] and State Optimism Measure [51].

The EQ-5D [52] measures quality of life with a visual analogue scale and five items evaluating the impact of pain on the ability of the participant to perform everyday tasks.

Microneurography.

Microneurography records single-unit activity from peripheral sensory nerves in awake participants by inserting a microelectrode into a peripheral nerve [54]. We typically use the superficial peroneal nerve on the dorsum of the foot because it is superficial, purely sensory, anatomically relevant to distal symmetrical polyneuropathy, and generally well tolerated. It is also the most distal nerve available, which provides an advantage for length-dependent (‘dying-back’) neuropathies that show a distal-to-proximal gradient. In older participants (>70 years) or those with severe neuropathy where the superficial peroneal nerve is denervated or inaccessible, alternative sites (superficial radial nerve or more proximal peroneal sites) will be used.

C-fibres are characterised by the phenomenon of activity-dependent slowing of conduction velocity in response to repetitive transcutaneous electrical stimuli, and natural stimulation in a series of experiments [24].

The core protocol will consist of:

Identification of C-fibres at 0.25 Hz stimulation. After identification, stimulation is paused for three minutes to allow baseline recovery, followed by six minutes at 0.25 Hz. The rate is then increased to 2 Hz for three minutes and followed by a further six minutes at 0.25 Hz. The pause and 0.25 Hz blocks may be adjusted for participant tolerability, but the 2 Hz block will not be shortened.

Where possible this will be followed by the extended protocol.

1. Stimulation at 0.125 Hz, 0.25 Hz, 0.5 Hz and 1 Hz for at least two minutes at each frequency.
2. Paired-pulse testing to assess recovery cycles by delivering a test stimulus at progressively shorter intervals after a conditioning stimulus.
3. Mechanical and thermal natural stimulation of the receptive field using von Frey hairs and thermal probes (40°C, 45°C, 48°C and 20°C) to determine transduction thresholds and sensitisation to mechanical and thermal stimulation.

We aim to record from up to three independent receptive fields per participant, but a single stable recording suffices for inclusion. All C-fibres will undergo the core protocol; the extended protocol is optional and recorded whenever feasible.

C-fibre subtypes are classified according to their response to an increase in electrical stimulation frequency to 2 Hz. This response, known as activity-dependent slowing of conduction velocity, is a characteristic feature of C-fibres in which conduction latency increases (i.e., slows) following repetitive stimulation. Using this method, five distinct populations of C-fibres can be identified: type 1A (“mechano-sensitive”) nociceptors, type 1B (“mechano-insensitive”) nociceptors, cold thermoreceptors, low-threshold mechanoreceptors, and cutaneous sympathetic efferent fibres [24,25].

There are two main reasons for distinguishing between a core and an extended recording protocol. The first relates to participant tolerability. Some individuals, particularly older participants, find prolonged recordings difficult due to musculoskeletal discomfort, difficulty keeping the limb still, challenges in following instructions, or urinary urgency. To minimise discomfort, participants are seated in a supportive chair with pillows to cushion pressure points, and are offered comfort breaks at the start of the recording. The second reason concerns recording stability. C-fibre signals may be lost during longer sessions because of small movements of the electrode or limb, or loss of contact during natural stimulation. As a result, extended recordings cannot always be completed. To prevent ascertainment bias, data from the extended protocol will therefore be treated as secondary or exploratory outcomes.

Microneurography measurements will include measures of:

Axonal properties

1. i. Baseline conduction velocity.
2. ii. Degree of activity-dependant slowing to several stimulation frequencies (e.g., 0.25 Hz to 2 Hz)
3. iii. Action potential morphology analysis of the extracellular recorded action potentials
4. iv. Threshold responses to thermal and mechanical stimulation

Nociceptor excitability

1. i. Spontaneous activity
2. ii. Double or multiple spikes
3. iii. Thermal and mechanical sensitisation defined as prolonged ongoing discharges outlasting stimulus duration or lowered thresholds in response to thermal and mechanical stimulation.

Axonal excitability (across all C-fibre types)

1. i. The velocity recovery cycle refers to the interval-dependent change in C-fibre conduction velocity after an action potential. It is used to measure refractoriness and excitability. It progresses through several distinct phases that depend on the time interval between the first (conditioning) action potential and the second (test) action potential. These phases include the refractory period, the supernormal period, and the late subnormal period.

Nerve conduction studies.

Nerve conduction studies are performed to confirm the presence of a distal symmetrical polyneuropathy, following the recommendations of the American Academy of Neurology and the American Association of Electrodiagnostic Medicine [55]. The initial assessment includes sural sensory and peroneal motor nerve conduction studies in one lower limb. If both results are normal, no further testing is required to exclude large-fibre distal symmetrical polyneuropathy.

If either study is abnormal, additional tests are undertaken as appropriate. These may include assessments of the ipsilateral tibial motor nerve; the contralateral sural sensory, peroneal motor, or tibial motor nerves; or the median sensory and ulnar sensory and motor nerves in one upper limb.

Electrodiagnostic confirmation of a length-dependent distal symmetrical polyneuropathy is defined as an abnormality in any attribute of nerve conduction in at least two separate nerves, one of which must be the sural nerve. Variables such as skin temperature, age, and height are taken into account when interpreting results.

Axonal excitability.

In selected cases, axonal excitability will be assessed using an electrophysiological technique known as threshold tracking [23]. Nerve excitability reflects the biophysical properties of myelinated axons and their membrane potential. These measurements complement conventional nerve conduction studies, which primarily assess action potential amplitude and latency but provide limited information about underlying membrane dynamics. In contrast, threshold tracking is sensitive to changes in membrane potential at the site of stimulation.

Key parameters include refractoriness, supernormality, the strength–duration time constant, and threshold electrotonus. Recordings will be obtained from the motor and sensory divisions of the median nerve in accordance with published recommendations [56]. This approach also enables modelling of ion channel function.

Quantitative sensory testing.

Quantitative sensory testing is a psychophysical technique used to assess sensory perception in response to defined stimuli. It provides a quantitative measure of sensory function and generates a sensory profile that captures both loss and gain of function across multiple sensory modalities. Quantitative sensory testing will be performed according to a modified version [42] of the German Research Network on Neuropathic Pain (DFNS) protocol [20,57]. In this study population, most participants are expected to experience bilateral symptoms; therefore, testing will be performed unilaterally on the dorsum of the foot and/or hand.

Quantitative sensory testing data will be entered into the Equista software package (version 1.2.2., CASQUAR GmbH), developed by the German Research Network on Neuropathic Pain. Equista converts raw data into z-scores, normalising results for age, sex, and the body region tested [58]. A z-score of zero represents the population mean, while scores greater than +2 or less than –2 standard deviations indicate gain or loss of function, respectively.

Participants will be stratified into somatosensory phenotype groups based on their quantitative sensory testing results. Two complementary approaches will be used:

1. Cluster analysis – An unbiased algorithm identifies three distinct phenotypic groups: sensory loss, mechanical hyperalgesia, and thermal hyperalgesia. Individuals will be assigned deterministically to one of these groups [13,19].
2. Irritable versus non-irritable phenotype – Participants will be classified as having an irritable phenotype when detection thresholds are normal but there is evidence of sensory gain (mechanical or thermal hyperalgesia), or a non-irritable phenotype when there is either sensory loss or no evidence of sensory gain [59].

Skin biopsy.

Intra-epidermal nerve fibre density is a validated measure for assessing small-fibre pathology. Skin biopsy samples will be obtained in accordance with the published guidelines of the European Federation of Neurological Societies/Peripheral Nerve Society on the use of skin biopsy in the diagnosis of peripheral neuropathies [60]. The biopsy will be taken at the end of the clinical assessment, from a site located 10 cm proximal to the lateral malleolus, once all other relevant investigations have been completed.

Participants will not undergo a skin biopsy if they decline the procedure, are receiving warfarin or other anticoagulant therapy, or have contraindications such as local skin infection or ulceration.

Genotyping.

This study uses DNA and RNA sequencing to investigate the genetic and molecular mechanisms that drive changes in C-fibre axonal properties and excitability. We will collect blood samples to extract DNA and analyse genetic variants that may influence C-fibre axonal properties and nociceptor excitability, focusing on genes involved in ion channel function and peripheral nerve signalling. We will extract RNA from blood and skin samples to characterise gene expression using high-throughput RNA sequencing, enabling us to identify transcriptional changes that modify C-fibre excitability and sensory transduction. Sequencing data will be processed through validated bioinformatic pipelines, with key findings verified using quantitative methods such as RT-PCR.

Together, these approaches will reveal how genetic and molecular factors shape C-fibre function by identifying variants and expression changes linked to altered axonal behaviour and nociceptor properties.

Patient stratification.

Patient study participant’s neuropathy [61] and pain [1] is graded according to published guidelines. The diagnosis of distal symmetrical polyneuropathy (DSP) is based on a detailed neurological assessment in combination with tests.

Download:

• PNG
  larger image
• TIFF
  original image

https://doi.org/10.1371/journal.pone.0349407.t001

Participants are then graded according to the presence or absence of neuropathic pain. Neuropathic pain classification is based on whether the pain present is neuropathic in nature and follows a distinct neuroanatomically plausible distribution.

Participants included in analysis will be divided into groups based on aetiology, neuropathy, and neuropathic pain grading [10].

1. i. Diabetic participants are divided into those with (possible/probable/confirmed) and without a distal symmetrical polyneuropathy
2. ii. Diabetic participants with a distal symmetrical polyneuropathy are further divided into those with and without neuropathic pain.
3. iii. Non-diabetic participants with distal symmetrical polyneuropathy are divided into those with and without neuropathic pain. This will include participants with small-fibre neuropathy, i.e., those participants with injury confined to the small fibres (Aδ and C-fibres.)
4. iv. Extreme pain phenotypes include:

Erythromelalgia and paroxysmal extreme pain disorder – episodic or persistent pain and erythema of the extremities or trunk that is exacerbated by warming and relieved by cooling;

Familial episodic pain syndrome – severe episodic pain localised to the trunk and limbs with no structural cause;

Insensitivity to pain – inability to feel nociceptive stimuli

Study hypotheses.

The primary hypothesis is that participants with neuropathic pain will exhibit altered axonal properties and excitability compared to those without neuropathic pain.

Secondary hypotheses we will test are:

1. 1) Somatosensory subgroups will match nociceptor properties. The planned outcomes are comparisons between neurophysiological and sensory phenotype variables, for example, mechanical hyperalgesia sensory profile will show lowered neurophysiological threshold to mechanical stimulation.
2. 2) Genetic variants will confer a unique nociceptive axonal profile and excitability signature. The planned outcome is the comparison of nociceptor fibre excitability and axonal properties between those participants with and without genetic variants.

Statistical plan.

Analyses will address the relationships between neuropathy, neuropathic pain, quantitative sensory phenotype and microneurography measures of nociceptor excitability and axonal properties. Analyses will be prespecified, scripted for reproducibility and conducted using appropriate statistical software.

Continuous variables will be summarised using mean (standard deviation) or median (interquartile range) depending on distribution; categorical variables will be summarised using counts and percentages. Point estimators will be reported alongside their respective 95% confidence interval. All hypothesis tests will be two-tailed. Where multiple comparisons are performed, p-values will be adjusted using methods appropriate to the comparison (for example Bonferroni or false discovery rate), and this will be reported alongside unadjusted p-values.

The primary hypothesis is that participants with neuropathic pain will exhibit altered axonal properties and excitability compared to those without neuropathic pain. The primary outcome will therefore be a comparison of microneurography measures of axonal excitability and axonal properties between participants with and without neuropathic pain.

Primary outcome comparisons.

Comparison of mean values and distributions of axonal excitability and axonal property measures at participant level for individual nociceptive fibres (for example, spontaneous activity, multiple spikes, mechanical and thermal sensitisation, conduction velocity, and activity-dependent slowing).

Planned primary analysis.

Between-group comparisons (neuropathic pain vs no neuropathic pain) will be conducted using appropriate non-parametric tests (Mann–Whitney U test) where data are not normally distributed, or independent-samples t-tests where normality is satisfied.

Data from individual nociceptive fibres will be analysed with explicit consideration of clustering within participants. Where appropriate, fibre-level measurements will be summarised at the participant level (for example, by averaging) to provide an independent observation per participant for primary analyses. In addition, mixed-effects models will be used, with participant included as a random effect, to account for within-participant variability. Multiplicity arising from multiple outcomes will be addressed using appropriate multiple-comparison adjustments (for example, Bonferroni or false discovery rate correction).

This dual approach allows us to perform conservative primary analyses at the participant level by averaging, while also utilising more sophisticated modelling through a mixed-effects model that retains the detail of the individual fibre data while correcting for the dependence between fibres from the same participant. The mixed-effects model will bridge the participant (unit of inference) and individual fibre (unit of observation) without the statistical error of treating fibres from the same person as independent data points.

Multivariable logistic regression models will be used to examine the association between neuropathic pain status and microneurography-derived measures of axonal excitability and properties, adjusting for prespecified covariates (age, sex, diabetes status, neuropathy severity, and medication use).

Multicollinearity among predictors will be evaluated using variance inflation factors; when two predictors are highly correlated, the variable with greater variance or stronger biological relevance will be retained in the final model.

Secondary analyses.

Secondary hypotheses explore relationships between somatosensory phenotype (QST clusters), genetic variants and nociceptor excitability signatures.

1. Somatosensory subgroup correspondence

- Compare microneurography variables across QST-defined phenotypic groups (sensory loss; mechanical hyperalgesia; thermal hyperalgesia) using non-parametric ANOVA (Kruskal–Wallis) or one-way ANOVA as appropriate.

- If omnibus tests are significant, perform post-hoc comparisons with correction for multiple testing.

2. Genetic variant effects

- Compare microneurography measures between participants who carry rare variants in ion channel genes (NaV, TRPA1) and those who do not, stratified by aetiology (diabetic versus non-diabetic).

- Use regression models to estimate the effect of genetic variants on fibre excitability, adjusting for confounders. If sample sizes for specific variants are small, present descriptive statistics and cautionary interpretation.

3. Microneurography association analyses

- Spearman rank-order correlations will evaluate associations between microneurography measures and clinical metrics: spontaneous pain intensity, evoked pain intensity, DN4/NPSI descriptor scores, and quality-of-life measures.

- A binomial regression model will examine the association between microneurography variables and membership of somatosensory subgroups.

Dimensionality reduction and clustering

Given the high dimensionality of clinical, genetic, microneurographic and biochemical data, we will use dimensionality reduction and unsupervised clustering to identify latent patient subgroups.

- Apply principal component analysis to standardised variables across the domains of interest (genetics, demographics, microneurography, biochemistry). Retain components that explain an adequate cumulative variance (e.g., eigenvalues >1).

- Apply k-means, centroid and hierarchical clustering on retained principal components. Determine optimal cluster number using multiple metrics (gap statistic, silhouette width, elbow method).

- After cluster assignment, compare original variables across clusters using omnibus tests with post-hoc comparisons to characterise cluster profiles.

Sample size justification.

Sample size calculations are based on power to identify differences in QST cluster frequencies and to support clustering and multivariate analyses.

- Using pilot quantitative sensory testing data from DOLORisk, expected cluster frequencies for diabetic distal symmetrical polyneuropathy are 53.9% (sensory loss), 25.0% (mechanical hyperalgesia) and 15.8% (thermal hyperalgesia). For non-diabetic neuropathies the corresponding frequencies are 49.7%, 21.2% and 14.5%.

- For moderate effects (Cohen’s d ~ 0.5) in the quantitative sensory testing cluster analyses, a total of 130 diabetic and 148 non-diabetic participants provides approximately 80% power at α = 0.05.

- For k-means and hierarchical clustering applied after principal component analysis, a conservative estimate requiring 2^8 = 256 participants was used to ensure adequate representation across variable permutations.

- A multivariate regression model with age, sex, diabetes status, neuropathy severity, and medication use will require 8 degrees of freedom to represent all its variables (assuming a categorical variable coding for 4 categories of medication). Allowing for at least 15 event per variable (EPV) we will require a sample size of at least 120 participants for model development. As described above, variable selection and filtering based on multicollinearity will likely increase the EPV.

- Taking these calculations together (and accounting for anticipated success rates in microneurography recordings: minimum 60% for diabetic participants and 80% for non-diabetic participants), the recruitment target is 272 participants (124 diabetic and 148 non-diabetic).

Risk of bias and mitigation.

Potential sources of bias include misclassification of fibre type, sampling bias, outcome reporting bias and confounding.

Fibre classification bias will be minimised using pre-specified, reproducible criteria based on activity-dependent slowing of conduction velocity. Sampling bias will be reduced by recording from multiple receptive fields per participant and using mixed-effects models to account for within-participant variability. Confounders will be adjusted for in regression models. The protocol and statistical analysis plan will be pre-specified and published to minimise outcome reporting bias.