Intermittent fasting and chronic pain: Protocol for a systematic review of randomised controlled trials.

Chronic pain, defined as pain that lasts or recurs for 3 months or more, 1 is a leading public health problem in the USA and globally. In 2023, 24% of US adults (over 80 million people) experienced chronic pain, 2 and 8% experienced high-impact chronic pain that substantially limited work or activities of daily life. 2 Chronic pain disproportionately affects older adults, women, individuals with lower socioeconomic status, and those living in rural areas, underscoring its role as both a clinical and health equity issue. 3 Globally, chronic pain is similarly a widespread condition. Conservative estimates suggest that approximately 20% of the world’s population, or over 1.5 billion people, live with chronic pain. 4 Pain is one of the most common reasons individuals seek medical care and is associated with substantial disability, reduced quality of life, and increased risk of co-occurring conditions such as depression and anxiety. 4 Standard of care for chronic pain varies by specific condition but commonly includes pharmacologic therapies (eg, non-steroidal anti-inflammatory drugs, antidepressants, anticonvulsants and opioids), physical therapy and, in some cases, surgery. Despite the availability of these approaches, many patients experience incomplete relief, adverse effects or functional limitations, highlighting a persistent unmet need for safe, accessible and scalable interventions. 5 6 The ongoing opioid epidemic further underscores the need for effective non-pharmacologic approaches to pain management. Over the past two decades, opioid prescribing for chronic non-cancer pain increased substantially, contributing to widespread misuse, opioid use disorder and overdose deaths. 7 In 2022 alone, more than 80 000 deaths in the US involved opioids, reflecting the continued magnitude of this crisis. 8 Although prescribing rates have declined in recent years, many patients with chronic pain remain exposed to opioids or face challenges related to tapering and inadequate alternative treatments. 9 In response, clinical guidelines, including those from the United States Centers for Disease Control and Prevention (CDC), now emphasise maximising non-opioid and non-pharmacologic therapies for chronic pain. 9 These shifts highlight the urgent need to identify safe, accessible and scalable interventions that can reduce reliance on opioids while improving pain-related outcomes. In recent years, lifestyle-based approaches, including dietary interventions, have gained increasing attention as potential adjunctive or alternative strategies for chronic pain management. Diet may influence pain through multiple pathways, including systemic inflammation, oxidative stress, metabolic dysfunction and gut-brain signalling. 10 – 12 Intermittent fasting (IF), defined as eating patterns that cycle between periods of fasting and normal food intake, has emerged as a particularly promising dietary approach. Common IF regimens include time-restricted eating, alternate-day fasting and periodic fasting protocols (eg, 5:2 diets). Research in healthy individuals without chronic pain suggests IF can increase patient-reported physical and mental health and decrease inflammation, fatigue and body pain. 13 14 Recent narrative reviews have highlighted the potential utility of IF in the management of chronic pain conditions, highlighting the known cellular and biochemical changes induced by fasting, such as alterations in insulin, cytokines, neurotransmitters, endogenous opioids, brain-derived neurotrophic factor and others. 15 16 Therefore, we propose a systematic review on IF for adults with chronic pain. To the best of our knowledge, no other recent systematic review exists on the topic. A preliminary search of PubMed/MEDLINE, PROSPERO, Epistemonikos, and Open Science Framework (OSF) was conducted and no current or ongoing systematic or scoping reviews specifically evaluating the effect of IF in adults with chronic pain were identified.

This review will be conducted in accordance with published Cochrane Guidance and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) checklist. 17 18 The protocol was prospectively registered in PROSPERO (CRD #420261395061) in May 2026 and plans include completion of the review by July 2027. Deviations from the protocol, with rationale, will be documented and reported. The literature search will include the PubMed/MEDLINE, Embase (Ovid) and Web of Science (Core Collection) databases. The search string will initially be built for PubMed but will then be translated using the Polyglot Search Translator. 19 The search will be translated into Portuguese by a Portuguese language speaker and the LILACS (VHL) database will be searched with both English and Portuguese search terms (see online supplemental material for full search strategies). No language or date restrictions will be placed on any search. To search the grey literature, up to 300 results from Google Scholar will be retrieved using a WebCrawler strategy. 20 We will also search the reference lists of all included studies for other possibly relevant studies and report the results following the TARCIS guidelines. 21 ClinicalTrials.gov will be searched for relevant research in progress or recently completed, and the PROSPERO, Open Science Forum, and MedRXiv databases will be searched for protocols for other relevant studies or reviews, completed or ongoing. The literature searches will be repeated before the completed review is submitted for publication, to ensure they have captured all relevant literature. Inclusion criteria for this review are as follows: Only studies with adult human subjects will be included (18 years and older). If a mixed population (eg, 16 years or older), 80% or more of the population must be 18 years of age or older to be included. If outcomes are reported by age groups, we will use outcomes reported for the 18 and older cohort. Diagnosis: participants will have chronic pain (as defined by pain that lasts or recurs for 3 months or more). Potential chronic pain conditions include but are not limited to musculoskeletal pain (eg, low back pain, neck pain), neuropathic pain, headache (tension-type, migraine or cluster), endometriosis, dysmenorrhoea, fibromyalgia, rheumatoid arthritis and others. Studies will include an IF intervention such as time-restricted eating, alternate day fasting, the 5:2 diet or fasting for Ramadan (29–30 day sunrise to sunset religious fast). IF for the purposes of this review is defined as a pattern of eating that involves regular, structured periods of little to no calorie intake. We recognise that Ramadan fasting differs from other IF paradigms because it is a religious practice, typically involves abstinence from both food and fluids from dawn to sunset, and may be accompanied by changes in sleep, meal timing, social routines and spiritual practices. In addition, individuals with some chronic pain conditions may be exempt from fasting due to known exacerbations in symptoms caused by fasting. Therefore, Ramadan fasting studies in chronic pain populations may be uncommon. Ramadan fasting studies will not be excluded a priori, but will be included only if they otherwise meet the review’s eligibility criteria. Included studies will be randomised, controlled trials with a group that receives a fasting intervention as described above, and a group that receives any kind of comparison (eg, placebo, waitlist, health education, standard-of-care or active (eg, a transcutaneous nerve stimulation device such as the CEPHALY)). Studies must include self-reported measures of either function or pain (including but not limited to pain intensity, pain duration, pain interference or pain-related disability) at baseline and at least one additional post-baseline measurement. Measures that assess pain-related cognitions such as pain catastrophising, pain acceptance, or fear of pain will also be eligible for inclusion. Composite scores that include pain severity and impact (eg, the MIDAS scale) are also eligible. Pain and function-related outcomes will be analysed separately, based on recommendations for core outcome sets in chronic pain research. 22 After the literature search all identified citations will be uploaded into the PICOportal software and deduplicated. 23 Study screening will be performed by two authors, independently, blinded and in duplicate. Screening will first be conducted on a title/abstract level and subsequently reviewed on a full-text level with reasons for exclusion noted. Screening will be done independently and in duplicate. In case of disagreement in which consensus cannot be reached, a senior researcher will serve as an adjudicator. Studies of interest in languages other than English will be initially translated using DeepL (an advanced neural machine translation tool; https://DeepL.com ), and essential key details confirmed, when possible, with fluent speakers of that language. Data extraction will include study characteristics, including authors, publication year, study design, sample size, pain diagnosis, fasting intervention description, pain measure values (at baseline and postintervention), pain condition and duration of pain condition. If studies report rescue medication use by treatment group, for example, non-steroidal anti-inflammatory drugs (NSAIDs) or triptans for migraine treatment, this data will also be extracted as long as it is reported at baseline and study end. Data extraction will be performed independently by two reviewers using a standardised, piloted form in the PICOPortal software with discrepancies resolved through discussion or consultation with a senior researcher. If any studies require clarification of findings or appear to be lacking relevant information, authors will be contacted up to three times to request clarification and/or data if possible. A narrative synthesis of the included studies will be conducted. If data are homogeneous enough and appropriate for pooling, effect estimates will be pooled in a meta-analysis using a random-effects inverse-variance model. Data will be extracted from each study including means, standard deviations (SDs), confidence intervals (CIs), and p-values. The R software 24 and meta 25 and metafor packages 26 will be used to meta-analyse results using random effects due to anticipated clinical and methodological heterogeneity among studies. The effect of IF on pain levels will be expressed as the standardised mean difference (SMD) using Hedges’ g. When available, we will analyse the change score (post-intervention score minus pre-intervention score) and the SD of the change, between the two groups. If this is not reported, we will impute this value using established methods from the Cochrane Handbook if there is sufficient data to do so. If the change scores cannot be imputed, we will do the following: in a mean difference (MD) meta-analysis, we will use the post-intervention mean and SD for studies lacking change data, in combination with change data from studies which provide it. In a SMD meta-analysis, we will analyse studies using change scores and post-intervention scores separately, following recommendations in the Cochrane Handbook. The meta-analytic plan is as follows: Combining all studies with a continuous outcome measure of pain, we will conduct an SMD meta-analysis using Hedges’ g. We will separate active controls (eg, CEPHALY device) from passive (no treatment or waitlist comparison) in separate meta-analyses. Using all studies which report a post-randomisation by-group proportion of rescue medication use, we will specify this proportion of medication use as the outcome in a MD meta-analysis. Only including studies which use the Numerical Rating Scale (NRS) and which report post-randomisation rescue medication use, we will conduct an MD meta-analysis. Using the studies from analysis #3, we will calculate a QPAC 1.5 -adjusted NRS score, accounting for medication use, following methods from Sridhar 2025. 27 28 These results will be compared with the results from analysis #3. Following published guidance to improve interpretability, 29 using all studies with a continuous outcome measure of pain, we will convert all outcome measures to a 1–10 scale and analyse using MD meta-analysis (Method 2i from Thorlund et al , Converting to natural units of a familiar instrument). I 2 will be used as a measure of statistical heterogeneity, 30 and a funnel plot (Egger’s test) used to assess visually and statistically when there are 10 or more studies reporting a particular outcome. 31 We will follow published guidance and use p<0.10 as statistical evidence of small study effects/publication bias. 32 If the funnel plot identifies potential publication bias, Duval and Tweedie’s trim-and-fill method will be applied to correct the effect size. 33 Risk of bias will be assessed with the Cochrane Risk of Bias 2.0 (CROB) tool for randomised, controlled studies. 34 A study will be identified as being at high risk of bias if (1) any domain on the CROB2 is rated as ‘high risk of bias’ or (2) two or more domains are rated as ‘some concerns’. We will use the INveStigating ProblEmatic Clinical Trials in Systematic Reviews (INSPECT-SR) tool to assess the trustworthiness of the included randomised controlled trials. 35 We will also conduct a leave-one-out sensitivity analysis to iteratively repeat the main meta-analysis excluding one study at a time, to evaluate its impact on the overall effect size and assess the robustness of our meta-analytic findings. If sufficient data are available, we will estimate the within-group correlation (r) between pre-intervention and post-intervention scores. Because change scores assume a reasonable positive correlation between repeated measures within individuals, low or negative correlations (eg, r<0.5) may reflect measurement error, data inconsistencies, or implausible variance structures that can distort effect estimates. We will therefore conduct sensitivity analyses excluding such studies, as recommended in the Cochrane Collaboration guidance. 18 For meta-analysis #4 above, we will hold the weight (w) constant (at 1.5) and vary the correlation r (trying 0 and 0.16) in our calculation of the adjusted SD, following established methods. 27 For meta-analysis #4 above, we will hold the correlation r constant and vary w (trying 1.0 and 2.0). 27 A priori subgroups will include the following ( Table 1 ): Type of pain condition: inflammatory, neuropathic, nociplastic/centralised. Duration of pain condition. As a continuous variable, for example, time in months or years Pain intensity Dichotomise studies into two groups NRS scores 0–3 (low baseline pain). NRS scores ≥4 (medium/high baseline pain). Inflammatory pain would see largest effect of treatment Other types of pain would have smaller effects We will evaluate the credibility of any observed subgroup effects using the Instrument to assess the Credibility of Effect Modification Analyses (ICEMAN) tool. 36 The Grading of Recommendations, Assessment, Development and Evaluation (GRADE) tool will be used to assess the quality of evidence for each meta-analysed outcome. 37 The studies in each meta-analysed outcome will be assessed for serious concerns about risk of bias, imprecision, indirectness, inconsistency, or publication bias. For each meta-analysis, if ≥55% of participants come from studies rated at high risk of bias with the CROB2 tool, a test for subgroup differences will be conducted, comparing the low and high risk of bias studies, using p<0.10 as a threshold. 38 If there is no statistically significant difference between them, the results for all studies will be presented, and we will not rate down for risk of bias. If there is a significant difference between them, we will present the results for (1) all studies, rating down for risk of bias and (2) the low risk of bias studies only, without rating down for bias. If there is a significant difference between the low and high risk of bias studies, but the high risk of bias studies have an effect estimate biased towards the null value, we will not rate down. In alignment with the Transparency and Openness Promotion (TOP) guidelines 39 and the Findable, Accessible, Interoperable, and Reusable (FAIR) principles, 40 the data and code for this meta-analysis will be made available on the Open Science Framework platform after publication.