Skip to content
problem genome

One of 592 unsolved problems in the Problem Genome, a collection written up for student design and engineering teams.

health

sniffing jars, with or without the smell

Smell training is the standard advice for smell loss after COVID-19, but placebo-controlled trials disagree on whether it beats sniffing odour-free bottles, and patients tend to stop the months-long routine whether or not it seems to work

Source: Mogensen et al., “Effect of Olfactory Training in COVID‐19 Related Olfactory…” Open ↗source: primaryLast updated: 2026-11-12

Problem statement

Smell training (sniffing four strong odours, such as orange, lavender, clove and peppermint, twice a day for 12 weeks or longer) is the treatment most often recommended to people whose sense of smell has not come back after COVID-19. A Danish trial gave 65 such patients either the four essential oils or identical brown bottles of odour-free placebo oil, used the same way. After 12 weeks the essential-oil group "did not significantly improve their olfactory function compared to the control group," though it did report better smell-related quality of life. A Brazilian double-blind trial also concluded training was not superior to placebo, and a Canadian placebo-controlled pilot found no change on the smell test, only in how patients rated their own smell and in how many reported parosmia. But a second Canadian trial against odourless bottles found classical training significantly better than the control, and a 2026 meta-analysis of ten randomised trials (628 patients, searched to June 2025) found a small benefit overall (standardised mean difference 0.30, pooled from seven of the ten), and an Austrian trial reported in a 2026 preprint found training clearly better than no training. Six of the ten trials compared training with no training, which cannot separate the odours from the routine, the attention and the expectation; and with placebo bottles, blinding is hard for anyone who can still smell a little (the Danish authors judged it not possible; the Canadian pilot tried to keep it by not telling patients that the training worked through smelling odours). A second problem sits underneath: in an Austrian clinic, 31% of patients doing smell training stopped because they felt better and 51% stopped because they did not, after about five months on average. The unsolved problem is a home smell-training setup whose active ingredient can be tested honestly and that people keep using long enough to find out.

Why this matters

In the words of the 2026 meta-analysis, persistent chemosensory dysfunction "significantly impairs quality of life (QoL), affecting nutrition, safety, and psychosocial well-being, and is associated with increased depression and anxiety", and it removes a warning sense for smoke, gas leaks and spoiled food. The same review says about 10–20% of people with COVID-related smell loss develop persistent loss, and the Danish authors cite long-term prevalence of 2.9–8.3% two years after infection. With no proven drug, the advice patients get is mostly smell training, so whether it works, and for whom, decides what patients are told to spend three months or more doing.

What’s been tried and why it hasn’t worked

(1) Classical training against placebo. In the Danish trial (Rigshospitalet, Copenhagen; recruited June 2022–December 2023), 49 of 65 completed 12 weeks; neither group improved meaningfully on the Sniffin' Sticks TDI test, and the authors note "the increase in the olfactory function measured by TDI test is so minor that a few more patients would not have lifted the response." Patients with parosmia (distorted smell, present in 57% of the sample) did worse. The Brazilian double-blind trial (123 completers, followed 180 days; 84.5% overall reached normal smell) concluded "OT is not superior to PB" (placebo), with baseline smell scores the strongest predictor of recovery (abstract read only). The essential-oil group in the Danish trial did report a significant gain in smell-related quality of life, which the authors say could be attributed to taking part in an unblinded active treatment. Two Canadian placebo-controlled trials (the pilot read in full, the second at abstract level) split: a 50-person double-blind pilot found "an increase in subjective but not objective olfactory function" and less parosmia, while a 68-person three-arm trial found classical training improved smell significantly more than odourless propylene glycol. (2) Comparisons with no training. These mostly favour training: the ten-trial meta-analysis, whose smell-test result mixes both kinds of control (SMD 0.30, pooled from seven of the ten trials: three against placebo, four against no training; its quality-of-life result depended on one study) and the SMELL trial (70 randomised, preprint, per-protocol analysis, improvement versus natural course). The meta-analysis authors still call for caution because of "heterogeneity in OT protocols," from 4 to more than 24 weeks. (3) Measuring whether people actually train. The Danish trial used a training diary and got self-rated adherence of about 81.5% in both arms, but "the accuracy of these subjective assessments of adherence cannot be evaluated." The meta-analysis names poor adherence as "the primary challenge," and the Danish authors note that dropouts were younger and more recently infected, the patients most likely to benefit, and that some may have stopped because they recovered, which "could lead to an underestimation" of the effect. (4) Devices to make training easier. A "training ball" holding four odour tubes roughly doubled adequate adherence over jars (63% vs 30%, measured by a questionnaire) in 60 patients with post-infectious loss, and its users scored higher on the smell test at 12 weeks; scented nasal inserts that can be worn while moving cut dropout from 23.2% to 6.7% in 116 people with hyposmia (the nasal-insert trial read in full, the training ball at abstract level). Neither device logs use objectively, and neither was tested against an odour-free version. (5) Digital smell training. A UCL team with an industry partner (Hynt Labs) gave 18 UK adults aged 45–75 a digitally controlled scent-delivery device and a companion app for six months; the app recorded the user's intensity ratings after each session, and the researchers gave participants printed monthly adherence percentages at monthly meetings. The study was qualitative, about motivation and data visualisation; it "intentionally avoided analysing perceptual change data," had no control arm, and its authors say a clinical study of the approach "is needed." A 2026 report from an overlapping author group, on a six-month deployment in 18 UK households "with and without olfactory disorders", gives an app-logged figure: "Participants completed 74% of 5,600 potential sessions" (full text read: one of the two daily sessions counts as half a day, the base is the 15 of 18 who finished, and monthly adherence fell from 72.6% in the first month to 55.1% in the sixth); it reports no smell-test outcome and no control arm. A Johns Hopkins randomised trial of another device with a companion web app (ScentCare, NCT06733636, recruiting) uses Sudoku puzzles, not a sham scent, as its control.

What would unlock progress

Two things, which may be one device. First, a sham that patients cannot tell apart from the real training while it is going on, so the odour's contribution can be measured (the Danish authors could not blind because most participants still had some smell). Second, objective and low-burden evidence of use and progress at home: a kit that records when it is opened and sniffed, and gives a short self-test so patients see change before they give up (the Austrian clinic data suggest people stop on their own impression either way). The training-ball and nasal-insert trials show device design can change adherence, and digital smell-training devices already log sessions and self-rated intensity; what is still open is a validated home self-test (intensity ratings are not a smell test), a placebo-matched control (a digitally dosed device could in principle deliver sham sessions), and any controlled outcome data for a logging device. The parosmia finding suggests training may need to be matched to the type of loss, which is a further reason to record what each person actually did.

How we’d know it’s being solved

(a) Logged training sessions per week (device-recorded, not diary) and the share still training at 12 and 24 weeks; (b) a placebo-controlled comparison in which patients' guesses about their allocation are no better than chance; (c) change on a validated psychophysical test (Sniffin' Sticks TDI, minimal clinically important difference 5.5 points per the Danish power calculation), reported separately for people with and without parosmia. Beware the proxy: quality-of-life gains in an unblinded arm can reflect taking part, as the Danish authors note.

Where to start

Facility-free door: starting from the digital smell-training probe above (which logs sessions but has no validated self-test and no sham mode), prototype a smell-training kit that logs use (lid-open and sniff-duration sensing, or a phone app with a camera check) and includes a short at-home identification self-test with feedback; bench-test logging accuracy against observed use by volunteers. Design door: design an indistinguishable sham (matched bottles, carrier oil, smell of the bottle itself, routine) and a blinding check, and write the protocol for a placebo-controlled trial that uses the logging kit to measure adherence — the semester deliverable is the trial design, not the trial. Analysis door: compare the ten trials in the 2026 meta-analysis by control type (four placebo, six no training) and training duration, using only their published tables. Access note: testing with people who have smell loss needs an ENT or smell clinic partner and ethics review; the facility-free door uses volunteers with normal smell for the usability and logging tests. Relevant skills: product and interaction design, sensors and embedded systems, trial methodology, sensory science.

Genes

Related problems

Stuck the same way, in other fields.

field: food & the living worldthe tractor can't talk to the sensorPrecision agriculture data interoperability failurebuilt without the people who'd live with itsource: primaryfield: food & the living worldthe warning system nobody maintainsFlood early warning systems often lack the ongoing funding that keeps them runningbuilt without the people who'd live with itsource: primaryfield: food & the living worldthe mold cure nobody usesAflasafe cuts aflatoxin contamination by 80–100% but remains poorly adopted while people get liver cancerbuilt without the people who'd live with itsource: primaryfield: the built worldthe stove is not a shelfIn Finland about half of stove fires start from things left on the stove, not food; stove guards, now required only for new stoves in senior and special-needs housing, are certified against cooking-oil fires and in tests alarmed late when a pizza box or pot holder hid the hot platebuilt without the people who'd live with itsource: primary

Sources

  • Mogensen DG, Aanaes K, Andersen IB, Jarden M, Backer V. "Effect of Olfactory Training in COVID‐19 Related Olfactory Dysfunction—A Placebo‐Controlled Trial." The Laryngoscope 2025;135(12):4849–4857, doi:10.1002/lary.32275, accessed 2026-10-22. Link ↗
  • Albazee E, Aldawas G, Alzeeby M, Alebrahim AH, Alwael A. "Olfactory Training for COVID-19-Related Olfactory Dysfunction: A Systematic Review and Meta-Analysis of Randomized Controlled Trials." Turkish Archives of Otorhinolaryngology 2026, doi:10.4274/tao.2026.2026-3-27, accessed 2026-10-22. Link ↗
  • Serrano TLI, Antonio MA, Giacomin LT, Morcillo AM, Dirceu Ribeiro J, Sakano E. "Olfactory training for the treatment of COVID-19 related smell loss: a randomised double-blind controlled trial." Rhinology 2025, doi:10.4193/rhin24.081 (abstract only), accessed 2026-10-22. Link ↗
  • Haas M, Raninger J, Kaiser J, Mueller CA, Liu DT. "Treatment adherence to olfactory training: a real-world observational study." Rhinology 2024, doi:10.4193/rhin23.207 (the DOI does not resolve at doi.org; full text free on the journal site), accessed 2026-11-12. Link ↗
  • Saatci O, Altundag A, Duz OA, Hummel T. "Olfactory training ball improves adherence and olfactory outcomes in post-infectious olfactory dysfunction." European archives of oto-rhino-laryngology 2020, doi:10.1007/s00405-020-05939-3 (abstract only) Link ↗
  • Winter AL, Henecke S, Thunell E, Swartz M, Martinsen J, Sahlstrand Johnson P, Lundstrom JN. "Olfactory training using nasal inserts is more effective due to increased adherence." Rhinology 2025, doi:10.4193/rhin24.369 (full text free on the journal site), accessed 2026-11-12. Link ↗
  • Heim B, De Cleene N, Jagusch F, et al. "Olfactory Training for COVID-19–Related Smell Loss: The SMELL Trial." Research Square preprint 2026, doi:10.21203/rs.3.rs-9750365/v1 (abstract only) Link ↗
  • Bérubé S, Demers C, Bussière N, Cloutier F, Pek V, Chen A, Bolduc-Bégin J, Frasnelli J. "Olfactory Training Impacts Olfactory Dysfunction Induced by COVID-19: A Pilot Study." ORL; journal for oto-rhino-laryngology and its related specialties 2023;85(2):57-66, doi:10.1159/000528188, accessed 2026-10-22. Link ↗
  • Filiz G, Bérubé S, Demers C, Cloutier F, Chen A, Pek V, Hudon É, Bolduc-Bégin J, Frasnelli J. "Can Multisensory Olfactory Training Improve Olfactory Dysfunction Caused by COVID-19?" Multisensory research 2024;37(4-5):299-316, doi:10.1163/22134808-bja10127 (abstract only), accessed 2026-10-22. Link ↗
  • Beşevli C, Marques A, Brianza G, Dawes C, Obrist M. "Flipping Perspectives: Visualising Digital Smell Training." DIS '25 (ACM Designing Interactive Systems), 2025, doi:10.1145/3715336.3735425, accessed 2026-10-22. Link ↗
  • Besevli C, Brianza G, Dawes C, Mathur S, Beganovic S, Boak D, Fatah Gen Schieck A, Lechner M, Maggioni E, Philpott C, Obrist M. "Digital intervention for smell training: a real-world study on engagement, adherence, and behavioural dynamics." Rhinology 64(3):367-376, 2026, doi:10.4193/rhin25.290 (record via Europe PMC, PMID 41574552; full text read from the journal's free PDF), accessed 2026-11-12. Link ↗

Also cited in the brief

  • Haas et al. 2024 and Winter et al. 2025. Link ↗
Revision notes

This brief gets checked against its sources and contains the following revisions. Future revisions get added here.

Worth knowing

  • The Austrian dropout figures (31% stopped because they felt better, 51% because they did not) come from clinic patients with smell loss from several causes over ten years, not from COVID-19 patients.

Full check

Sources read

  • Haas et al., why patients stop smell training (Rhinology, 2024): full text
  • Winter et al., nasal-insert smell training trial (Rhinology, 2025): full text
  • Mogensen et al., placebo-controlled smell training trial (Laryngoscope, 2025): full text
  • Albazee et al., meta-analysis of smell training trials (2026): full text
  • Bérubé et al., placebo-controlled pilot trial (ORL, 2023): full text
  • Beşevli et al., digital smell-training device study (DIS, 2025): full text
  • ClinicalTrials.gov, ScentCare trial record: full text
  • Serrano et al., placebo-controlled smell training trial (Rhinology, 2025): abstract only
  • Saatci et al., training-ball trial (2020): abstract only
  • Filiz et al., trial against an odourless control (2024): abstract only
  • SMELL trial preprint: abstract only
  • Besevli et al., six-month home study of a smell-training device (Rhinology, 2026): full text

What changed

  • Not found in the sources, replaced with what they say: that the Danish trial used a paper diary (now: a training diary, the paper's words).
  • Stated more strongly than the source: the Danish authors now say the quality-of-life gain could be due to taking part in an unblinded treatment, as they put it, not that it is.
  • Numbers: the meta-analysis result is now said to pool seven of its ten trials, three against placebo and four against no training, not all ten; the Brazilian trial's 84.5% now reads as a share overall, since the summary does not say of which group.
  • Credit: the monthly adherence percentages in the digital device study were printed by the researchers, not shown by the app.
  • Links: the clinic dropout study and the nasal-insert trial now point to the full texts read.
  • Other: the page now says the nasal-insert trial was read in full and the training-ball trial only as an abstract.

Left open

  • the Brazilian placebo trial's details, including what the 84.5% is a share of (abstract only)
  • the size of the gain over an odourless control in the Filiz trial (abstract only)
  • whether the training ball records use objectively (abstract only)
  • the SMELL trial's results (preprint, abstract only)

Separate check of specific claims: A six-month home study of a smell-training device and app, with its logged session figures, was added as an earlier attempt, and the notes on which sources were read in full were brought up to date.

Earlier full check: Four points were brought in line with the full texts: what the Canadian pilot found, how hard blinding is, which kinds of control the meta-analysis pools, and where one paper was read.

Separate check of specific claims: The long title now says placebo-controlled trials disagree (display title unchanged), a trial's country was corrected, positive findings and the authors' view of dropout bias were added, and a smell-training app and a recruiting trial were added as earlier attempts.