Mail reaches the population; digital capture is not yet validated — and neither is disease-specific

Question: How do unsupervised digital and mail-based remote olfactory screening tools for neurodegenerative disease compare in diagnostic accuracy and adherence outside clinic settings?

Answer

Across these papers, the two modalities are not competing alternatives but successive stages of one workflow, and the comparison the question asks for has never been made. Mail-based dispatch of physical odorants is the only remote format validated at population scale: the PPMI remote hyposmia screen sent smell tests to 57,754 people and obtained 38,419 completed tests, identifying 6,649 hyposomic individuals [8], while a fully unsupervised at-home toolkit combining an 8-scent card test with an online questionnaire reached 0.98 sensitivity and 0.83 specificity for manifest Parkinson disease [6]. The digital layer in every retrieved study is response capture, instruction delivery and scoring wrapped around a physical consumable, and no study has validated such a test unsupervised outside a clinic: the self-administered digital SMELL-RS was tested in a tertiary care facility and is described as suitable for self-administration "in a clinical setting" [4], and the RFID/app-linked OS-card was evaluated in matched clinic samples [2]. Adherence evidence is correspondingly asymmetric — mail has documented multi-step attrition across tens of thousands of participants [8], whereas digital olfactory tools report no completion or adherence data at all, forcing reliance on adherence figures from remote digital cognitive batteries that range from 41% to 97.6% [12] [11]. Taken together, the papers support mail-plus-online-response hybrid designs as the only empirically supported remote option, and they locate the accuracy bottleneck in the odor-identification paradigm rather than in the delivery channel.

Mail-based odorant dispatch achieves population reach but inherits the low disease specificity of performance-based smell tests

The PPMI screen demonstrates that mailed smell testing can be deployed at a scale no clinic-based tool approaches, with 79,837 people consenting to remote hyposmia screening and 14.4% of those completing testing confirmed hyposomic [8]. Accuracy against a neurodegenerative outcome in an unselected population is, however, modest: in a population-based clinicopathological cohort, UPSIT alone predicted conversion to mild cognitive impairment with approximately 60% accuracy up to five years before conversion [5]. The NIH Toolbox Odor Identification Test differentiated amnestic MCI from healthy aging in older African American adults with an AUC of only .626, at an optimal cut-off yielding 76.3% specificity but 40.4% sensitivity [7]. High accuracy appears only when the tested sample is already enriched for synucleinopathy: among people with polysomnography-confirmed REM sleep behavior disorder, UPSIT predicted a positive CSF α-synuclein seed amplification assay with an AUC of 0.89 and hyposmia carried a 92% positive predictive value [9].

Digitizing the response does not change what the odorant can detect, and mild impairment remains the failure point

The two connected digital tests report convergent accuracy profiles in which severe loss is detected well and partial loss poorly: the OS-card correctly classified 100% of normosmic participants and 96% of anosmics but only 68% of dysosmics overall [2], and the self-administered digital SMELL-RS achieved an ICC of 0.71 and correlated with Sniffin' Sticks at r = 0.68 [4]. Digital delivery shortens administration substantially — under 10 minutes for OS-card [2] and roughly 5.5 to 5.9 minutes per SMELL-RS subtest, two to three times faster than the corresponding Sniffin' Sticks tests [4] — but no retrieved study reports that a screen-delivered instrument, without any physical odorant, discriminates neurodegenerative disease. The domain where olfactory screening is most needed, subjective cognitive decline, is also where effects are smallest: a meta-analysis of 11 studies found lower odor identification in SCD versus cognitively normal participants with a summary effect of SMD = -0.67 [3].

The strongest unsupervised accuracy figure is a composite index, so part of the reported accuracy is not olfactory

The simplified PD-RISK Index reached 0.98 sensitivity, 0.83 specificity, and likelihood ratios of 5.88 and 0.02 [6]; that figure belongs to an 11-item questionnaire plus an 8-scent test entered online, and the same study reports that the abbreviated 8-scent test and the mailed UPSIT booklet "performed equally well in detecting hyposmia" [6]. This matters because identification performance alone does not separate disease categories: a standard clinical olfactory test detected impairment in both Parkinson disease and non-Parkinson olfactory dysfunction but failed to distinguish them, whereas olfactory perceptual fingerprints derived from odor ratings and sniffing behavior separated the groups at 88% accuracy, rising to 94% with age and sex matching [1]. The discriminating signal in that study was sniff-duration dynamics — patients with Parkinson disease paradoxically increased sniff duration to unpleasant odors while controls and non-Parkinson dysosmics decreased it [1] — a measurement that currently requires laboratory instrumentation and has no validated unsupervised analogue.

Adherence evidence favors mail for reach and digital capture for tolerance, but no olfactory study measures both

Mailed testing shows a steep but quantifiable per-step funnel: of 79,837 consents, 57,754 were sent a test and 38,419 completed it (66.5% of those sent), after which only 1,720 completed dopamine transporter imaging [8]. The attrition therefore concentrates at the transition from consent to returning a physical kit, and again at the transition to any clinic-based confirmation step. The at-home toolkit study reports diagnostic performance but no completion or return rate, so unsupervised olfactory adherence cannot be estimated from it [6]. Adherence benchmarks for unsupervised digital testing outside clinic come only from non-olfactory batteries: 82.2% adherence and favorable usability for a smartphone cognitive battery in a registry cohort [10], 97.6% completers and 96% adherence over 28 days for a remote Alzheimer assessment suite [11], 78.5% completion for at-home online cognitive screening before primary care visits [14], and only 41% average adherence when participants were prompted four times daily for seven days [12]. The pattern suggests adherence is governed by protocol burden more than by delivery channel, a point the scoping literature on remote unsupervised assessment in preclinical disease also flags as an open feasibility question in older adults [13].

Conflicts and gaps

No retrieved paper compares a digital and a mail-based olfactory tool head to head, in the same participants, against the same reference standard, so the question as posed is answered only by triangulation. Reported accuracies are not commensurable: PD-RISK's specificity of 0.83 is against healthy controls and mixed neurological conditions [6], OS-card's 68% dysosmic classification is against ETOC-derived status categories [2], and the NIHTB-OIT figure of 40.4% sensitivity is against a clinical aMCI diagnosis [7]. Two of the three digital instruments were validated in clinic samples, so their reported reliability cannot be assumed to transfer to unsupervised use [4] [2]. Adherence is defined inconsistently across the field — consent-to-completion in PPMI [8], session-level percentage of expected prompts in digital cognitive studies [12] — and no olfactory study reports repeat-test adherence, which is the metric that matters for longitudinal prodromal surveillance. Whether the digital component adds diagnostic information or merely convenience is untested.

What would change the answer

A within-participant equivalence trial in a prodromal-enriched cohort, with biomarker-confirmed status (CSF α-synuclein seed amplification assay or amyloid PET) as the reference standard, in which the same people complete a mailed UPSIT, a mailed 8-scent card, and an app-linked card test unsupervised at home, would directly settle the accuracy comparison; the PPMI and PD-RISK designs already supply the recruitment and diagnostic scaffolding for this [8] [6]. Adding passive capture — response latency, sniff or sampling behavior, video verification of card use — to the digital arm would test whether the specificity currently requiring laboratory sniff monitoring can be recovered remotely [1]. The adherence question requires publishing full per-step funnels with return rates and repeat-testing rates for each modality, and a randomized burden comparison (single vs repeated administration) modelled on the high-frequency remote designs that produced 41% adherence [12]. Finally, external validation of the connected tests in unassisted community samples, rather than in tertiary clinic samples [4] [2], is the minimum condition for calling any of these tools remote.

References

[1] Andelman-Gur, M. M., Shushan, S., Snitz, K., Pinchasof, G., Honigstein, D., Gorodisky, L., Ravia, A., Ezra, A., Hezi, N., Gurevich, T., Sobel, N.. "The World Smells Different in Parkinsons Disease". medRxiv. 2026-02-24T00:00:00.000Z. URL: https://doi.org/10.64898/2026.02.22.26346819

[2] Moussy, E., Fournel, A., Daude, C., Fieux, M., Ferdenzi, C., Bensafi, M., Richard, M.. "A rapid and connected test for olfactory loss screening". medRxiv. 2026-07-27T00:00:00.000Z. URL: https://doi.org/10.64898/2026.07.23.26358424

[3] Jobin, B., Zigrand, C., Frasnelli, J., Boller, B., Albers, M. W.. "Lower Odor Identification in Subjective Cognitive Decline: A Meta-analysis". medRxiv. 2025-05-04T00:00:00.000Z. URL: https://doi.org/10.1101/2025.04.15.25325887

[4] Hsieh, J. W., Dougherty, M., Poulopoulou, A., Blidariu, D., Senn, P., Hopper, R., Patel, D., Maggioni, E., Obrist, M., Vosshall, L. B., Keller, A., Landis, B.. "SMELL-RS: A Self-administered, Digital Test for Olfactory Dysfunction that is Rapid, Reliable, and Accurate". medRxiv. 2026-03-31T00:00:00.000Z. URL: https://doi.org/10.64898/2026.03.28.26349316

[5] Tremblay, C., Choudhury, P., Driver-Dunckley, E., Alasmar, Z., Serrano, G. E., Shill, H. A., Mehta, S., Fereshtehnejad, S.-M., Ho, A., Shprecher, D., Lorenzini, I., Belden, C. M., Atri, A., Adler, C. H., Beach, T. G., Dadar, M., Zeighami, Y.. "Olfactory decline in aging: longitudinal trajectories and associations with cognitive decline and postmortem neuropathology". medRxiv. 2026-04-28T00:00:00.000Z. URL: https://doi.org/10.64898/2026.04.27.26351817

[6] Li, J., Grimes, K., Saade, J., Lewis, D., Tomlinson, J. J., Frank, A., Ramsay, T., Salmaso, N., Manuel, D., Schlossmacher, M. G.. "PD-RISK Index: Piloting an unsupervised toolkit to identify adults with Parkinson disease". medRxiv. 2026-10-19T00:00:00.000Z. URL: https://doi.org/10.64898/2026.04.30.26352172

[7] Berivan Ece, Bruno Giordani, Felicia Goldstein, Sandra Weintraub, Jennifer Manly, Pamela Dalton, Valentina Parma, Richard Gershon, Emily Ho. "NIHTB odor identification in African American older adults with and without aMCI". psyarxiv. 2026-02-16T13:53:12.732Z. URL: https://osf.io/983gh_v1

[8] McMahon, B., Dobkin, R., Kuhl, M., Heathers, L., Lemon, L., Dimos, J., Brown, E., Pierz, K., Marek, K.. "Overcoming challenges to successful recruitment of individuals at the earliest stages of disease: An example from Parkinson's disease research". medRxiv. 2025-06-18T00:00:00.000Z. URL: https://doi.org/10.1101/2025.06.17.25329789

[9] Brown, E., Chahine, L. M., Videnovic, A., Marshall, M. J., Kurth, R., Caspell-Garcia, C., Brumm, M. C., Coffey, C., Siderowf, A., Simuni, T., Marek, K., Tanner, C. M., on behalf of the Parkinsons Progression Markers Initiative,. "Smell testing to identify early alpha-synucleinopathy among people with dream enactment behavior". medRxiv. 2024-12-12T00:00:00.000Z. URL: https://doi.org/10.1101/2024.12.11.24318857

[10] Dhanam, S., Sanderson-Cimino, M., Taylor, J. C., Paolillo, E. W., Fregly, R., Kwang, W., Maruff, P., Wise, A., Heuer, H. W., Forsberg, L. K., Kramer, J. H., Boeve, B. F., Rosen, H. J., Mackin, R. S., Weiner, M. W., Nosheny, R. L., Boxer, A. L., Staffaroni, A. M., Brain Health Registry,. "Remote, self-administered, smartphone cognitive testing in a registry-based cohort: Feasibility, reliability, and validity findings". medRxiv. 2025-10-30T00:00:00.000Z. URL: https://doi.org/10.1101/2025.10.28.25338686

[11] Taylor, K. I., Wolfer, A. M., Kurniawan, I. T., Veloso, M., Keita, G., Hagenbuch, N., Shi, B., Orfaniotou, F., Aponte, E. A., Colell, M. G. V., Chatham, C. H., Holiga, S., Ullmann, R., Abouelkheir, W., Rey-Riek, S., Poon, E., Watson, D., Boada, M., Perumal, T. M.. "From feasibility to neuroanatomic validity of remote cognitive smartphone assessments in early Alzheimers disease". medRxiv. 2026-05-21T00:00:00.000Z. URL: https://doi.org/10.64898/2026.05.19.26353554

[12] Andrew Aschenbrenner, Hannah Wilks, Matthew Welhaf, Samhita Govindan Katteri, Sarah Stout, Martin Sliwinski, Tammie Benzinger, Brian Gordon, Yan Li, Richard Perrin. "Remote Digital Cognitive Assessment in a Trial-Ready Alzheimer’s Cohort: A Scalable Approach for Early Intervention Studies". psyarxiv. 2025-10-03T15:32:13.465Z. URL: https://doi.org/10.1002/trc2.70311

[13] Polk, S. E., Öhman, F., Hassenstab, J., Papp, K. V., König, A., Schöll, M., Berron, D.. "A scoping review of remote and unsupervised digital cognitive assessments in preclinical Alzheimer's disease". medRxiv. 2025-03-25T00:00:00.000Z. URL: https://doi.org/10.1101/2024.09.25.24314349

[14] Thompson, L., Eaton, C., Prieto, S., Lawrence, M., Czech, S., Rosenbaum, J., Anthony, D., Chandran, R., Kelly, L., Ngo, I., Goldberg, A., Gillooly, C., Richardson, C., Elwy, R., Jones, R., Rentz, D.. "Feasibility of at-home online cognitive screening prior to primary care wellness visits for older adults.". medRxiv. 2026-03-05T00:00:00.000Z. URL: https://doi.org/10.1101/2025.08.19.25333983