Loss of taste in possible viral infection: etiologies, pathogenesis, and the evidence for treatments
Description
Background. Loss or distortion of taste is a recognised accompaniment of viral infections - in the author's training and practice, of upper respiratory infections - and was reported very widely during the COVID-19 pandemic. Patients ask which infections do this, what is happening in the body, and whether treatment helps. Most early syntheses pooled smell and taste together and counted frequency only.
Objectives. For new-onset taste dysfunction attributed to a confirmed, probable or possible viral infection: (Q1) which viruses are associated, with what frequency, timing and recovery; (Q2) which pathogenesis mechanisms have been proposed, and how direct the evidence for each is; (Q3) what interventions have been studied, with what effects, doses, harms and special-population considerations.
Search methods. PubMed, Europe PMC (including its preprint index) and ClinicalTrials.gov were searched in full on 20 November 2026, with no date limit. The WHO ICTRP keyword search could not be run (form-only portal; logged as not run), and Embase, Scopus, Web of Science, CINAHL and Cochrane CENTRAL were not run (no licensed access); their ready strings are printed in an appendix, marked as not run. Citation chasing covered all studies included at that point and all competing reviews; studies chasing added were not then chased.
Selection criteria. Human studies of any design with a denominator, and case series of 10 or more, reporting taste separately or extractably from smell, in people with taste dysfunction attributed to a viral infection. Mechanistic human studies and an animal and in-vitro stratum informed Q2; any intervention started for the dysfunction informed Q3. Case reports of fewer than 10, narrative reviews, editorials and letters without original data were excluded from synthesis. Unretrieved records form an abstract-only stratum, never pooled and with no GRADE contribution.
Data collection and analysis. A team of large language model roles screened, extracted and assessed risk of bias in duplicate under the author's accountability, with every disagreement adjudicated by reading (title and abstract agreement 0.890, kappa 0.618). Proportions were pooled on the logit scale with random effects (REML, Hartung-Knapp), Q-based I squared and 95% prediction intervals. Certainty was rated with GRADE; risk of bias used RoB 2, ROBINS-I and JBI checklists.
Main results. The search retrieved 5,988 records; 3,501 were screened, and 363 were included after full-text assessment, citation chasing and a final eligibility-consistency step, representing 362 unique publications (one paper, Ohla 2022, appears under two record IDs, R01101 and R02974; neither contributed data to any pool). At the final step, 26 records with no primary data (narrative reviews, letters and one hypothesis paper; the class includes C8424) were formally excluded under the locked eligibility for primary studies and are listed in Appendix E; none had contributed data to any pool, so no pooled number changed. Every pooled study concerns SARS-CoV-2: in the searches run and the records retrieved and classified, no study of influenza or any other named virus mapped to any pool, so the quantitative answer to the general viral question is, in substance, a COVID-19 answer. Acute taste dysfunction (taste-specific, SARS-CoV-2) pooled at 37.7% (95% CI 32.9 to 42.7; 137 studies, 155,010 participants), with I squared 99.5% and a 95% prediction interval of 5.3 to 86.7%. By ascertainment: 47.4% (psychophysical testing, 15 studies) and 36.5% (self-report, 122 studies); by setting: 26.8% (hospitalised, 55) and 39.6% (community, 35); no interaction test was performed and no subgroup difference is established. Complete recovery pooled at 87.4% within 4 weeks, 85.2% at more than 4 to 12 weeks, and 87.8% beyond 12 weeks, each from a different study set, with near-scale-wide prediction intervals. Any recovery beyond 12 weeks pooled at 90.3% (5 studies). No study reported any recovery separably in the more-than-4-to-12-week window; that cell is empty. Any recovery within 4 weeks rests on a single study (63 of 81, 77.8%). For pathogenesis, all 214 studies carrying a mechanism code concern SARS-CoV-2, and 73.4% were graded speculative (proposed in discussion, not measured). Only 15 carried direct human evidence. The largest raw direct-human count belongs to the residual 'other mechanisms' category (5 studies, five unrelated tests). On coherence grounds the best-supported theories are systemic or local inflammation, olfactory-mediated misattribution, direct infection of taste tissue, and central gustatory pathway involvement; direct infection has one null direct study among three and no demonstration of infection of human taste buds in vivo, and the central theory rests on one genuinely direct imaging study. Six of the eleven pre-specified theories have no direct human evidence for their own mechanism. For treatments, only 23 studies tested an intervention and 8 had any comparator; 2 randomised trials reported a taste-outcome estimate, and a third (zinc) reported only a composite smell-and-taste outcome. Oral prednisolone may result in little to no difference in gustatory scores in the one low-risk-of-bias trial (low certainty). Photobiomodulation, tested in people with taste loss of at least 4 weeks, showed a large complete-restoration proportion in one 36-participant trial at some concerns of bias, which did not report adverse events. Zinc has no taste-specific effect estimate anywhere. Adverse events were reported in only 6 of the 23 treatment studies. GRADE certainty was very low for frequency, complete recovery at every window, time to recovery, the zinc composite outcome and all harms outcomes; low for any recovery beyond 12 weeks, prednisolone and photobiomodulation; and no evidence for several pre-specified cells.
Limitations. Five databases and the WHO ICTRP keyword search were not run; 245 reports were never retrieved in full text and 99 records remain awaiting classification; and all pooled evidence concerns SARS-CoV-2.
Authors' conclusions. On the evidence that can be pooled, this is a SARS-CoV-2 literature: taste dysfunction is common, usually recovering, and no more precise than the wide prediction intervals allow. The pathogenesis literature mostly proposes mechanisms rather than testing them. The treatment literature cannot support a recommendation for or against any intervention, and none has the comparative evidence a follow-on meta-analysis would need. Useful next studies are an adequately powered photobiomodulation trial with taste-specific outcomes, standardised recovery outcomes, and any eligible study of a non-SARS-CoV-2 virus.
Files
manuscript.pdf
Additional details
Related works
- Is supplemented by
- Software: https://github.com/rtmalikian/viral-taste-loss-systematic-review (URL)