Mental Health Effects of Long-Term Fasting: Brain Changes, Symptoms, and Treatment Options

A narrative search was carried out in PubMed/MEDLINE, Scopus, and PsycINFO for articles published from 1 January 2010 to 30 June 2025. The string (“prolonged fasting” OR “intermittent fasting” OR “alternate-day fasting” OR “time-restricted feeding”) AND (“psychology” OR “psychiatry” OR “mood” OR “cognition” OR “stress”) was applied to titles, abstracts, and keywords. Reference lists of retrieved papers and recent systematic reviews were screened manually to capture additional studies.

We included studies meeting the following a priori criteria. Population (humans): Adults (≥18 years); adolescent samples were considered only when psychological/psychiatric outcomes were explicitly assessed. Interventions/exposures: Fasting paradigms, namely, time-restricted eating (TRE), alternate-day fasting (ADF), periodic/prolonged fasting, and religious/ritual fasting. Comparators: Any (active control, usual diet, parallel control, or within-subject pre–post). Human outcomes: Validated psychological/psychiatric endpoints (e.g., HDRS-17, PHQ-9, GAD-7, STAI-S, BAI; clinician-assessed relapse/recurrence) and/or stress-physiology/autonomic markers with a clear mental-health relevance (e.g., morning salivary cortisol, high-frequency heart-rate variability [HF-HRV]). Preclinical inclusion: In vivo studies directly pertinent to neurobiological pathways of mood, anxiety, or psychosis (e.g., BHB→NLRP3, BDNF/PGC-1α signaling, hypothalamic–pituitary–adrenal axis recalibration). Designs: Randomized controlled trials (RCTs), non-randomized trials, cohorts, case–control, pre–post studies, and case series (humans); in vivo experiments (preclinical). In vitro studies were excluded unless tightly anchored to converging in vivo evidence. Language/time window: English; January 2010–June 2025.

Exclusion codes (mapped to PRISMA boxes in Figure 1) were defined as follows: wrong outcomes—absence of validated psychological/psychiatric measures or mental-health-relevant physiological markers (studies reporting only metabolic/weight outcomes without mental-health endpoints); wrong population—purely pediatric samples; severe medical cohorts without mental-health outcomes; animal/in vitro work not mechanistically relevant; wrong intervention—calorie restriction without a fasting component or diet/supplement interventions unrelated to fasting; wrong design—editorials, narrative opinions without primary data, conference abstracts without a retrievable full text; reports not retrieved—full texts unobtainable after institutional access attempts and author contact; duplicates—multiple reports from the same cohort, in which case the most complete version was retained.

We searched PubMed/MEDLINE, Scopus, and PsycINFO for records published from January 2010 through June 2025. The search combined controlled vocabulary and free-text terms for fasting paradigms and mental-health outcomes (e.g., “prolonged fasting” OR “intermittent fasting” OR “time-restricted eating” OR “alternate-day fasting” OR “religious fasting”) AND (psycholog* OR psychiatr* OR mood OR anxiety OR stress OR depression). Core MeSH/keywords included fasting; intermittent fasting; time-restricted eating/feeding; alternate-day fasting; religious fasting (Ramadan, Eastern Orthodox); and mental-health terms (psychology, psychiatry, mood, depression, anxiety, stress, cognition, HPA axis, BDNF). Full Boolean strings and database-specific filters are reported in. We performed backward and forward citation chasing of included records. “Gray literature” was limited to reference-list screening and trial-registry checks; non-peer-reviewed materials (preprints, theses, abstracts without retrievable full texts) were excluded. Duplicates were removed using reference-manager de-duplication and manual verification. Table S1 (Supplementary Materials)

Titles/abstracts and then full texts were screened in duplicate by two reviewers working independently; disagreements were resolved by a third reviewer. For each eligible study we extracted, on pilot-tested forms, the design, population/sample, clinical status/diagnosis, fasting protocol (type, timing, duration), comparator, psychometric instruments, follow-up window, key psychological/psychiatric outcomes, and a brief design-level risk-of-bias note. These fields are summarized for human studies in Table 1 (Summary of human evidence linking prolonged fasting to key psychological outcomes) and correspond to the columns reported in Table 2. Preclinical data items included species/strain, fasting paradigm, mechanistic readouts, and behavioral endpoints when applicable. Data integrity and consistency checks were performed prior to synthesis, and reasons for exclusion at the full-text stage were recorded and are summarized in Figure 1 and detailed in Table S1.

Randomized controlled trials were appraised with the Cochrane RoB-2 tool, observational studies with the Newcastle–Ottawa Scale, and animal studies with the SYRCLE risk-of-bias tool. Narrative grades (low, moderate, or high risk of bias) were assigned to each study. Overall, randomized trials exhibited low-to-moderate risk of bias, mainly due to incomplete blinding and selective reporting; observational studies were moderate, with residual confounding and measurement variability; animal studies showed variable risks related to randomization and outcome assessment. These considerations informed the narrative grading of evidence strength.

Given heterogeneity in fasting regimens, psychiatric assessments, and endpoints—most relying on self-report instruments rather than clinician-administered interviews [23,24], a quantitative meta-analysis was not feasible. Results were therefore synthesized qualitatively, with particular attention to protocol duration, psychiatric stratification, and outcome directionality.

The database and manual search identified 3029 records. After removal of duplicates (n = 962), non-English (n = 96), and non–peer-reviewed items (n = 6), 1905 records were screened. Of these, 227 reports were sought for full-text retrieval, 129 were assessed for eligibility, and 87 studies (39 human; 48 preclinical) were included in the qualitative synthesis (Figure 1; Table S1).

The human evidence base comprised randomized controlled trials, non-randomized trials, cohort studies, pre–post investigations, and observational studies spanning time-restricted eating (TRE), supervised prolonged fasting programs (e.g., Buchinger regimens), and religious fasting. For each study, design, sample characteristics, fasting protocol (type, timing, duration), comparator, psychometric instruments, follow-up, and outcomes are summarized in Table 2. Across human trials, intervention durations typically ranged from 4 to 12 weeks for TRE and 4–21 days for supervised prolonged fasting programs. Sample sizes were predominantly small-to-moderate (≈20–100). Primary assessment tools included PHQ-9/BDI-II (depression), GAD-7/STAI-S (anxiety), PSS (perceived stress), and selected cognitive tests (executive/working-memory tasks).

Across metabolically healthy adults, TRE and supervised prolonged fasting were associated with small improvements in depressive symptomatology and perceived stress; effect magnitudes varied by baseline symptom burden and the degree of supervision. Supervised Buchinger-type programs consistently reported reductions in State-Trait Anxiety Inventory-State (STAI-S) scores, lower morning salivary cortisol, and increases in high-frequency heart-rate variability (HF-HRV). Religious fasting yielded mixed findings: Eastern Orthodox Great Lent was associated with modest reductions in Perceived Stress Scale and Beck Anxiety Inventory scores [25], whereas during Ramadan, a higher relapse risk in bipolar disorder was observed under specific clinical conditions. Severe adverse events were uncommon but documented: isolated psychotic or manic decompensations occurred in vulnerable individuals, and restrictive-eating phenotypes tended to worsen in the absence of clinical supervision.

Overall, most human studies reported favorable changes in mood and/or perceived stress under circadian-aligned TRE or supervised prolonged fasting, whereas neutral findings were more common in longer, unsupervised, or less structured regimens.

In vivo studies converged on several pathways plausibly linking fasting to affective and stress-related outcomes: ketone-mediated signaling (e.g., β-hydroxybutyrate effects on the NLRP3 inflammasome), neurotrophic modulation (brain-derived neurotrophic factor and PGC-1α-related mitochondrial programs), hypothalamic–pituitary–adrenal axis recalibration, gut–brain interactions influencing barrier integrity and microglial activation, and immuno-oxidative attenuation. Most preclinical evidence derives from rodent models (mice/rats) employing 24–72 h fasting or TRF paradigms. While these data elucidate plausible mechanisms (e.g., BHB→NLRP3, BDNF/PGC-1α), species differences, protocol heterogeneity, and behavioral readouts limit direct clinical extrapolation.

BHB binds to hydroxycarboxylic acid receptor-2 (HCA2) on microglia, inhibiting NLRP3-inflammasome activation and downstream IL-1β release—an anti-inflammatory action germane to affective disorders [27]. As a class I histone-deacetylase inhibitor, BHB increases histone H3 acetylation at promoters of neuronal resilience genes such as BDNF and Peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1α) (Figure S1), thereby enhancing synaptic plasticity [28,29]. Rodent hippocampal slices exposed to physiological ketone concentrations exhibit augmented GABAergic tone and reduced glutamatergic excitability, effects reproduced in vivo by 48 h fasting [30]. These shifts underpin the anticonvulsant and anxiolytic profile of ketogenic interventions. In humans, supervised 4–21-day Buchinger fasting programs have been associated with significant reductions in perceived tension/anxiety and morning salivary cortisol, alongside increases in high-frequency heart-rate variability, indicating strengthened parasympathetic control [31].

Energy scarcity activates AMP-activated protein kinase (AMPK) and sirtuin-1, leading to PGC-1α-driven mitochondrial biogenesis, enhanced mitophagy, and up-regulation of antioxidant enzymes including superoxide-dismutase-2 [32]. Fasting-induced autophagic flux clears dysfunctional organelles and misfolded proteins; in neuronal cultures, a single 24 h nutrient deprivation period reduces α-synuclein aggregates via ULK1 activation [33].

Acute fasts elevate cortisol and catecholamines, yet repeated cycles dampen basal ACTH and enhance vagal tone, suggesting a recalibrated hypothalamic–pituitary–adrenal axis (Figure 1) conducive to stress resilience [34]. Concomitantly, adiponectin—a metabolic hormone with AdipoR1-AMPK-mediated antidepressant effects—increases both peripherally and centrally, correlating with improvements in mood scores [35].

Time-restricted feeding paradigms remodel gut microbial communities, enriching butyrate-producing taxa such as Roseburia and Faecalibacterium. These metabolites fortify blood–brain barrier integrity and dampen microglial activation, thereby influencing anxiety-like behavior in mice [36,37,38]. A pilot crossover study reported parallel shifts in microbial α-diversity and reductions in perceived stress after two weeks of 16:8 fasting in healthy adults [39].

Meta-analytical evidence indicates that fasting interventions ≥4 weeks reduce high-sensitivity C-reactive protein, Tumor Necrosis Factor alpha (TNF-α), and Interleukin-6 (IL-6) by 10–30%, with concomitant down-regulation of NF-κB-responsive transcripts [40]. These changes mirror clinical improvements in depressive symptomatology and may constitute a mechanistic bridge linking metabolic and mood benefits.

Taken together, ketone-mediated signaling, mitochondrial rejuvenation, endocrine recalibration, microbiota reshaping, and immuno-oxidative dampening converge to form a neurobiological substrate through which prolonged fasting can influence cognition, emotion, and psychiatric vulnerability. Inter-individual variability in these cascades—driven by genetics, sex, circadian alignment, and psychiatric history—highlights the need for personalized protocols and mechanistic endpoints in future trials.

Emerging evidence positions prolonged or intermittent fasting (IF) as a double-edged sword in the affective domain. On one side, metabolic switching and the attendant rise in β-hydroxybutyrate (BHB) trigger neurotrophic, anti-inflammatory, and monoaminergic cascades that may convey antidepressant and anxiolytic benefits; on the other, caloric deprivation can destabilize mood in susceptible individuals, particularly when fasting is unsupervised or prolonged. Below, we synthesize data across preclinical models, mechanistic biomarker studies, and clinical trials, highlighting moderators that tilt the fasting–mood relationship toward resilience or risk.

Rodent models demonstrate that 24–48 h water-only fasts increase hippocampal BDNF and activate the CREB–PGC-1α axis, yielding behavioral antidepressant-like phenotypes in forced-swim and tail-suspension tests [41]. Translationally, association between circulating BHB and proBDNF mRNA in human peripheral mononuclear cells, correlating with mood elevation [42]. A 2022 randomized controlled trial (RCT) in 93 adults with mild-to-moderate major depressive disorder (MDD) reported that a 4-week 16:8 IF protocol reduced Hamilton Depression Rating Scale (HDRS-17) scores by −4.8 ± 1.3 points versus −1.9 ± 1.1 in an isocaloric three-meal control, accompanied by increased serum BDNF and lower high-sensitivity C-reactive protein [43]. These findings align with a recent network meta-analysis [44] showing small-to-moderate effect sizes (Hedges g = 0.29) for depressive symptom reduction across six IF randomized controlled trials (Figure 1). Representative randomized trials reported small-to-moderate symptom reductions on clinician-rated (e.g., HDRS-17) and self-report measures (e.g., STAI-S, PHQ-9), sometimes accompanied by BDNF shifts. Conversely, null effects on BDI-II/POMS at 12 months under less intensive TRE underscore heterogeneity by protocol duration, baseline symptoms, and assessment modality.

Fasting modulates hypothalamic GABA/glutamate ratios and attenuates NLRP3-inflammasome activity, mechanisms implicated in anxiolysis (Figure 1). In mice, a 24 h fast reduces anxiety-like behavior and improves memory via brain-wide reductions in caspase-1 activity (a key regulator of IL-1β maturation) [45]. Clinically, healthy adults undergoing a 10-day Buchinger water fast exhibited significant reductions in State-Trait Anxiety Inventory-State (STAI-S) scores [46]. However, anxiety outcomes across trials are heterogeneous, partly due to differing baseline anxiety levels and assessment tools.

Psychological responses to fasting vary markedly by sex, psychiatric history, and hormonal milieu. Women with pre-existing MDD or bipolar II disorder appear particularly vulnerable to mood destabilization during calorie restriction [47]. Fasting-induced hypoglycaemia, cortisol surges, and sleep fragmentation may jointly precipitate irritability, anergia, or mixed affective states. In an observational cohort of 1 214 Ramadan fasters, 7.4% of participants with a prior mood-disorder diagnosis experienced clinically significant symptom exacerbation [48].

Sleep disruption is a salient mediator: in a randomized pilot trial, early time-restricted eating shifted sleep timing in late sleepers; independently, selective REM-sleep suppression increases next-day negative affect [49]. Moreover, fasting can cue maladaptive cognitive–emotional schemas in individuals with dieting history, potentiating rumination and guilt, which are transdiagnostic risk factors for depression [50].

Collectively, data suggest a U-shaped relationship wherein moderate, circadian-aligned fasting confers modest affective benefits in metabolically healthy adults, while aggressive or prolonged restriction heightens risk in vulnerable groups. Screening for mood-disorder history, sleep quality, and stress load is therefore essential before prescribing fasting interventions. We recommend (i) starting with ≤14 h fasting windows, (ii) maintaining consistent bedtime to preserve REM sleep, (iii) monitoring glucose and cortisol in patients with affective disorders, and (iv) integrating cognitive–behavioral strategies to mitigate guilt-based cognitions. Large-scale, stratified RCTs with clinician-rated endpoints and ≥6-month follow-up are urgently needed to define safety boundaries.

Fasting elicits a biphasic stress response. Acutely (<24 h), glycogen depletion activates the hypothalamic–pituitary–adrenal (HPA) axis, elevating ACTH and cortisol to mobilize glucose and free fatty acids (Figure 1), whereas with repeated cycles, homeostatic recalibration emerges, characterized by blunted basal cortisol, enhanced vagal tone, and lower heart-rate-variability indices of sympathetic drive [51]. In middle-aged mice, eight weeks of alternate-day fasting normalized diurnal corticosterone, up-regulated hippocampal glucocorticoid-receptor phosphorylation, and improved stress-coping in forced-swim and tail-suspension tests [52]. Translationally, in adults completing a 10-day Buchinger water fast, morning salivary cortisol decreased (≈18%) and high-frequency HRV increased, consistent with strengthened parasympathetic control and stress-system recalibration [53].

Structured religious fasts (e.g., those aligning feeding to circadian windows or imposing abstention from specific macronutrients) plausibly recruit similar stress-system adaptations, while differing in intent, oversight, and social context. At the immune–neuroendocrine interface, fasting paradigms comparable to religious regimens have been associated with 10–30% reductions in circulating IL-6 and TNF-α and with down-regulation of NF-κB–responsive transcripts, changes that track improvements in perceived stress and mood [54]. Mechanistically, ketosis emerging in longer fasts can exert ketone-mediated anti-inflammatory effects—β-hydroxybutyrate directly inhibits the NLRP3 inflammasome—thereby attenuating innate-immune activation relevant to stress responsivity and affective outcomes [55]. Human fasting programs with supervised protocols further illustrate feasibility and safety signals (cortisol reduction, HRV shifts), offering a translational bridge to religious contexts where adherence, sleep timing, and communal practices may modulate stress-related benefits [56]. Analyses of Ramadan studies should account for confounders such as altered sleep timing, caffeine intake, medication dosing schedules (e.g., lithium), and daily routine changes, which can independently influence mood or relapse risk.

Early-phase studies have begun testing fasting-mimicking diets (FMDs), alternate-day fasting, and 16:8 time-restricted eating (TRE) as adjuncts to standard care in major depressive disorder, post-traumatic stress disorder, and mild cognitive impairment. A randomized controlled pilot trial in adults with major depression (n = 20) showed that adding three monthly cycles of a fasting-mimicking diet to structured psychotherapy improved self-esteem and psychological quality of life versus psychotherapy alone, with comparable depressive-symptom reduction across groups [59]. An ongoing randomized controlled trial is comparing time-restricted eating with a Mediterranean-diet control for symptom reduction and quality of life in bipolar disorder, with rater-blinded assessments across 8–15 months. Despite these encouraging signals, fasting remains contraindicated or high-risk in patients with active psychosis, severe impulse-control disorders, borderline personality disorder with suicidality, and uncontrolled substance use.

Clinical checklist (monitoring and safety):

Implementing fasting interventions in psychiatric settings raises distinct ethical imperatives. Informed consent must articulate (i) uncertain long-term efficacy, (ii) potential for mood destabilization or metabolic derangement, and (iii) the right to withdraw without prejudice. High-risk conditions include active eating disorders (anorexia nervosa, bulimia nervosa, severe ARFID), unstable mood or psychotic disorders, and patients under active psychopharmacological treatment with narrow therapeutic windows or dosing-time sensitivity (e.g., lithium, sedative-hypnotics); in such cases, fasting should be avoided or undertaken only with close medical supervision. Contraindication screening should encompass active eating disorders, BMI < 18.5 kg m⁻², pregnancy, insulin-treated diabetes, cognitive impairment affecting decision-making capacity, and unstable cardiovascular or renal disease. A multidisciplinary team—psychiatrist, dietitian, internist/endocrinologist, and where available, clinical psychologist—should co-design the protocol and oversee monitoring.

Risk–benefit assessment ought to be continuous: laboratory review at baseline and week 4, structured mood and suicidality scales every fortnight, and caregiver check-ins for patients with limited insight. Ethics-committee oversight is advised for research applications, with rapid-response plans for adverse psychiatric events. Finally, fasting must remain a reversible, individualized adjunct rather than a one-size-fits-all replacement for evidence-based pharmacotherapy or psychotherapy.

No new data were created or analyzed in this study. Data sharing is therefore not applicable.