1. Harmonisation
We started by harmonising the study protocol across recruiting sites and defining a core protocol consisting of self-reported questionnaires for postal surveys, and an extended protocol for in-depth data collection with clinic visits. The 8 DOLORisk sites recruited over 2,000 participants between 2015 and 2019 following the extended protocol, and over 9,000 participants in population cohorts following the core protocol. We also used it as a basis to explore possible collaborations, from which we gained 2,000 samples with data comparable to DOLORisk. The DOLORisk method was applied to UK Biobank to collect more precise data on chronic pain, which will give new insights into the genetics of NP.
2. Genetics
We put together the largest cohort of people living with NP to date, adequately powered for genome-wide association studies. We identified association between NP and the loci encoding the genes KCNT2, LHX8 and SLC25A3.
We also identified variants in ion channel genes in participants with diabetic neuropathy (eg SCN9A encoding the voltage gated sodium channel NaV1.7) and some rare neuropathic disorders. We studied the functional impact of these variants and their links with phenotypes. We also found changes in gene expression in the skin which relate to painful (vs painless) diabetic neuropathy.
3. Central and peripheral physiological processes
Threshold tracking and QST clusters were unable to distinguish between participants with painful neuropathy and those with painless neuropathy. These techniques might be better suited to compare neuropathic and non-neuropathic states, or could be further investigated to track the evolution of nerve function in patients with neuropathy in a longitudinal study design.
Conditioned pain modulation tests showed that participants with painful diabetic neuropathy had more efficient pain inhibition mechanisms following a painful stimulus than the painless diabetic neuropathy group. EEG showed different connectivity patterns in the brain between the painful and painless groups.
4. Risk algorithms
Based on the psychological data collected in the common protocol, we built a risk algorithm for the development, maintenance and worsening of NP. The best fitting model is that designed for predicting the development of NP after 18 months, using quality of life, sleep disturbance, openness to new experiences, adverse childhood experiences and smoking as predictors. The main genetic component discovered in the population cohorts did not significantly improve the models’ predictive ability.
5. Dissemination
All DOLORisk peer-reviewed publications are available in open access. We shared our progress at national and international conferences, uploaded our results to public databases, and established links with many projects across Europe. Our experience in data harmonisation has been a useful lesson for other projects as well as grant providers, and DOLORisk has been cited many times as an example of how large international collaborations can advance research in their fields.
