An interventional study of LLM-supported pain neuroscience education and Expert-feedback pain neuroscience education in Pain, Chronic Pain and Health Professions Education, sponsored by İsmail Palalı. Completed at 1 site in Turkey (Türkiye). Open to participants aged 18 Years and older, including healthy volunteers. Per ClinicalTrials.gov, last updated 2026-11-09.
Sponsored by İsmail Palalı · Not applicable, Interventional, and Other
Physiotherapists need to explain pain to patients in a way that is scientifically accurate, understandable, and non-threatening. Pain neuroscience education (PNE) improves students' pain knowledge, but it is unclear whether this translates into retained ability to explain pain to patients. This randomized controlled trial tested whether adding structured feedback from a large language model (LLM; GPT-5.2 Instant, OpenAI, via the ChatGPT interface) to a standardized four-week PNE programme improves undergraduate physiotherapy students' pain-explanation competence compared with the same programme using standardized expert-developed (non-AI) feedback. The primary outcome was Pain Explanation Competence (PEC) on a previously unseen clinical transfer scenario, assessed without AI access 12 weeks after the intervention by three assessors blinded to group allocation.
Design: Prospective, single-centre, two-arm, parallel-group, assessor-blinded randomized controlled educational trial conducted at the Department of Physiotherapy and Rehabilitation, Faculty of Health Sciences, Harran University, Şanlıurfa, Türkiye.
Randomization: After baseline assessment, participants were allocated 1:1 using a computer-generated sequence with variable block sizes of four and six, generated by a researcher not involved in recruitment, intervention delivery, or outcome assessment. Allocation was concealed using sequentially numbered, opaque, sealed envelopes.
Intervention: Both groups received an identical four-week PNE programme (one supervised \~45-min session per week; \~180 min total) covering (1) pain, nociception and tissue damage; (2) multidimensional modulation of pain; (3) pain-related beliefs, clinical language and metaphors; (4) person-centred pain explanation and contextual adaptation. Each session included two training scenarios (eight in total) following the same five-step sequence: independent initial explanation, allocated feedback, critical appraisal of the feedback, independent revision, and an adaptation/reflection task. The experimental contrast was the feedback source: adaptive LLM feedback using a standardized prompt (seven domains; the model was instructed not to provide a complete answer) versus standardized expert-developed written feedback addressing the same seven domains.
Assessments: Baseline (T0), immediately after the intervention (T1), and 12 weeks after intervention completion (T2). At each time point participants completed three written fictional clinical scenarios not used in training, without access to AI or feedback. Responses were de-identified and independently scored by three assessors blinded to allocation, using the study-specific PEC rubric (six domains, each 1-5; total 6-30). The rubric and scenarios were content-validated by an eight-member expert panel and pilot-tested in 20 students not included in the trial.
Blinding: Participants and the session facilitator could not be blinded; outcome assessors were blinded.
Sample size: Assuming d = 0.57, two-sided alpha = 0.05 and 80% power, about 100 completers were required; the recruitment target was 120 to allow for attrition.
Statistical analysis: The primary outcome (T2 PEC on the transfer scenario) was compared between groups using a linear model adjusted for baseline PEC. Longitudinal PEC, RNPQ-TR and KNAP-TR were analysed using linear mixed-effects models (group, time, group×time; random intercept for participant). The six PEC domains at T2 were compared with Holm correction. Unsafe explanations were analysed with a binomial GEE using rater-level ratings clustered by participant. Prespecified sensitivity analyses adjusted for age, prior formal PNE and prior generative-AI use, and addressed external AI use (contamination) and intervention adherence.
Note: This study was registered retrospectively after completion.
3,227 studies on the registry are indexed under Chronic Pain; 729 are open to participants now.
This study's enrollment of 120 is above the median of 60 across 2,388 interventional studies indexed under Chronic Pain.
Browse Chronic Pain studies →İsmail Palalı is the lead sponsor of 2 studies on the registry; none are open to participants now.
Compared within Chronic Pain, this study's primary condition as ClinicalTrials.gg classifies it; the registry's first condition term is Pain. The counts are of registry records indexed under each term, refreshed daily. How we classify →
Go through the criteria one by one — yes, no or not sure — and take the list to your doctor or the study team. Kept in this browser only.
Exclusion Criteria:
Underlined terms in the criteria open these too. Explained by clinicaltrials.gg, not part of the registry record. The study team decides who can take part.
Before you join, the study team explains what taking part involves — its purpose, what happens, the risks and possible benefits — and answers your questions, so you can decide freely. Consent is usually recorded on a form (the informed consent form, ICF) signed by you or, where the law allows, by someone legally authorised to decide for you.
You can keep asking questions during the study, you are told of new information that matters, and you can choose to leave at any time.
Informed consent
Before you join, the study team explains what taking part involves — its purpose, what happens, the risks and possible benefits — and answers your questions, so you can decide freely. Consent is usually recorded on a form (the informed consent form, ICF) signed by you or, where the law allows, by someone legally authorised to decide for you.
You can keep asking questions during the study, you are told of new information that matters, and you can choose to leave at any time.
Explained by clinicaltrials.gg, not part of the registry record. The study team decides who can take part.
Standardized four-week PNE programme plus structured GPT-5.2 Instant feedback on students' pain explanations, followed by critical appraisal and independent revision. → Intervention: Behavioral: LLM-supported pain neuroscience education
Behavioral: LLM-supported pain neuroscience education
Identical PNE programme, scenarios, sequence and duration, with standardized expert-developed non-AI feedback addressing the same seven domains. → Intervention: Behavioral: Expert-feedback pain neuroscience education
Behavioral: Expert-feedback pain neuroscience education
Students write an independent pain explanation for a clinical scenario, submit it to GPT-5.2 Instant (OpenAI, ChatGPT interface) using a standardized prompt requesting feedback on seven domains (scientific accuracy, clarity, alignment with contemporary pain science, metaphors, misconception correction, threatening/nocebogenic language, patient adaptation) without providing a complete answer, critically appraise the feedback, revise independently, and complete an adaptation task. Delivered within a standardized four-week PNE programme (one supervised \~45-min session per week, two scenarios per session).
Students write an independent pain explanation for a clinical scenario, receive standardized expert-developed written feedback prepared before the trial addressing the same seven domains (key principles, common errors and reflective questions, without a complete answer), critically appraise the feedback, revise independently, and complete an adaptation task. Delivered within the identical four-week PNE programme (one supervised \~45-min session per week, two scenarios per session).
Pain Explanation Competence (PEC) on an unseen cross-context transfer scenario.
Study-specific analytic rubric with six domains (scientific accuracy; clarity/comprehensibility; alignment with contemporary pain science/biopsychosocial framework; misconception correction; safe, non-nocebogenic communication; patient/context adaptation), each scored 1-5. Total 6-30; higher = better. Mean of three blinded assessors.
Time frame: 12 weeks after completion of the 4-week intervention
Pain Explanation Competence (PEC) across all assessment scenarios
Mean total PEC score across the three written fictional clinical scenarios completed at each time point, scored with the study-specific analytic rubric (six domains, each 1-5; total range 6-30; higher scores indicate greater competence). Each response was independently scored by three allocation-blinded assessors and the mean rating was used. Analysed with a linear mixed-effects model (group, time, group×time).
Time frame: Baseline, immediately after the 4-week intervention, and 12 weeks after intervention completion
PEC domain scores on the transfer scenario
Separate scores for each of the six PEC domains (scientific accuracy; clarity and patient comprehensibility; alignment with contemporary pain science and the biopsychosocial framework; identification and correction of misconceptions; safe, non-threatening and non-nocebogenic communication; patient/context adaptation) on the unseen cross-context transfer scenario. Each domain is scored 1-5; higher scores indicate better performance. Holm correction applied across the six domains.
Time frame: 12 weeks after intervention completion
Metaphor competence
Quality of metaphor or analogy use in written pain explanations, scored 1-5 according to scientific appropriateness, conceptual correspondence, clarity, potential for misunderstanding, and adaptation to the patient context. Higher scores indicate more effective metaphor use. Scored by three allocation-blinded assessors; compared between groups adjusted for baseline.
Time frame: Baseline and 12 weeks after intervention completion
Proportion of unsafe pain explanations
Proportion of responses containing at least one predefined potentially unsafe communication element (scientifically false or misleading mechanisms; equating pain intensity with tissue damage; fragility- or danger-promoting language; fear-provoking statements about movement or loading; unsupported absolute reassurance; invalidating or judgmental language; potentially harmful recommendations). Classified by three allocation-blinded assessors; analysed with a binomial generalized estimating equation clustered by participant.
Time frame: 12 weeks after intervention completion
Pain neurophysiology knowledge (Revised Neurophysiology of Pain Questionnaire-Turkish, RNPQ-TR)
Turkish version of the Revised Neurophysiology of Pain Questionnaire (13 items; total score 0-13). Higher scores indicate greater pain-neurophysiology knowledge.
Time frame: Baseline, immediately after the 4-week intervention, and 12 weeks after intervention completion
Pain knowledge and attitudes (Knowledge and Attitudes of Pain Questionnaire-Turkish, KNAP-TR)
Turkish version of the Knowledge and Attitudes of Pain Questionnaire, assessing contemporary musculoskeletal pain knowledge and attitudes in healthcare students. Higher scores indicate greater alignment with contemporary pain science.
Time frame: Baseline, immediately after the 4-week intervention, and 12 weeks after intervention completion
External generative-AI use (contamination)
Self-reported independent use of generative AI outside supervised sessions for pain-science learning, creating pain explanations or metaphors, practising clinical scenarios, or receiving AI-generated feedback (yes/no). Reported as the proportion of participants in each group and examined in sensitivity analyses of the primary outcome.
Time frame: Immediately after the 4-week intervention and 12 weeks after intervention completion
Intervention fidelity
Proportion of participants completing each component of the scenario-based practice sequence (attendance, independent initial response, feedback phase, critical appraisal, independent revision, adaptation task) in each weekly session, together with actual session duration. In the LLM-supported group, completion of the standardized LLM interaction and any protocol or technical deviations were also recorded. Used to define the per-protocol sensitivity population (at least 75% intervention completion).
Time frame: During the 4-week intervention (weekly sessions)
Plan to share: No — De-identified data and the statistical analysis plan are available from the corresponding author on reasonable request, subject to institutional and ethical approval.
No publications or documents are linked to this record.
From the registry record's own update history. This site started tracking changes on Sep 25, 2026; for anything earlier, see the record history on ClinicalTrials.gov ↗
This study is completed, as verified in Nov 2026. You cannot join it, but the record below documents what was studied.
Get an email when the registry record changes — status, dates, results — or when someone posts here.
Sign in to followQuestions and observations about this study, from anyone following it. Not medical advice, and not a channel to the study team — their contact details are on the registry record.
Sign in to join the discussion. Reading takes no account; posting does. You choose a display name, and a pseudonym is the default.
Nothing here yet. If you are running this trial, taking part in it, or weighing whether to, this is the place to say so.
İsmail Palalı