The corticosteroids for COVID-19 induced loss of Smell (COCOS) trial was a single-centered, randomized, double-blinded, placebo-controlled study in the Netherlands to determine the efficacy of a short prednisolone treatment on olfactory disorders after COVID-19. The trial consisted of a baseline visit at the outpatient Ear, Nose, and Throat (ENT) clinic and a second visit (follow-up) after 12 weeks (Fig. 1). The Institutional Review Board of the participating hospital approved the research protocol (protocol number: 21–635/G-D, October 2021). This study was conducted according to the principles of the Declaration of Helsinki (2013, Fortaleza). The CONSORT 2010 guideline was performed. The recruitment phase started November 2021 and ended February 2022. The trial ended May 2022.

Study design. TDI score; Threshold-Discrimination-Identification score; SST; Sniffin’ Sticks Test; TST; Taste Strip Test

Participants were identified by the Dutch Patients Association for smell and taste disorders and via the Dutch media that had been approached by the participating hospital. Interested patients could contact the research team via our website during the screening period. Investigators contacted interested patients by telephone to check inclusion and exclusion criteria and medical status. Medical status consisted of medication use, medical history, date of confirmed positive COVID-19 test, and date of onset of the olfactory disorder. Patients were included if they were > 18 years old, if they had persistent (> 4 weeks) olfactory disorders within 12 weeks after COVID-19 diagnosis based on a positive test (PCR or antigen), and if they understood the Dutch language. Patients were excluded if they used oral anticoagulants without stomach protection or if they suffered from pre-existing olfactory disorders including chronic rhinitis or rhinosinusitis or diseases which contra-indicate the use of steroids (diabetes mellitus for which drugs are used, stomach ulcers/bleeding, psychoses or ongoing oncological disease). Women who were pregnant, or who intended to become pregnant, were excluded. Eligible patients were invited for a baseline visit at the outpatient ENT clinic at the participating hospital. At the baseline visit, patients could still be excluded if they had no objective hyposmia (reduced loss of smell) or anosmia (total loss of smell) confirmed with a Threshold-Discrimination-Identification (TDI) score > 30.5 on Sniffin’ Sticks Test (SST) or if they had other causes for olfactory disorders objectified by nasendoscopy.

We collected further baseline characteristics such as vaccination status and COVID-19 symptoms at first visit. Furthermore, we performed a nasendoscopy in order to eliminate other causes for olfactory disorders. Patients underwent objective smell and taste tests, and filled in three additional questionnaires. At the baseline visit, patients received their randomly allocated blinded study medication (40 mg prednisolone once daily for 10 days or matching placebo) and were instructed to start their 10 days of study medication the next morning. Researchers contacted patients by telephone ten days after the baseline visit to assess possible side-effects and treatment compliance. Patients in both groups performed 12 weeks olfactory training twice a day, coming to a total of 168 sessions. Patients crossed off a daily schedule allowing researchers to monitor olfactory training compliance. The follow-up visit was scheduled 12 weeks after the start of treatment. Outcome measurements were collected at the first visit (baseline) and second visit (follow-up) to compare outcomes (Fig. 1). All outcomes were registered in an electronic case report form (eCRF), the endorsed system Castor EDC.

Patients were randomly allocated to receive prednisolone or placebo. Half of the group was treated with capsules of 40 mg of prednisolone, once daily for 10 days. The other half received capsules of placebo, once daily for 10 days. The pharmacy that prepared prednisolone and placebo capsules made a block randomization sequence list, on which the patient subject number was linked to the study medication number. This pharmacy is a Dutch state-of-the-art good manufacturing practice compounding pharmacy independent from our department. To minimize seasonal effects between groups, randomization occurred in block sizes of four patients. The blinding of researchers, physicians, outcome assessors, and patients to the treatment allocation broke after all the analyses were finished.

The primary outcome was the objective difference between the two groups on the TDI score post-treatment at 12 weeks, measured with the Sniffin’ Sticks Test (SST, Burghart). The SST consists of three parts: a threshold test (score ranging 1–16), discrimination test (score ranging 0–16), and identification test (score ranging 0–16). The TDI score is the sum of these three tests and ranges from 1 to 48. The higher the score, the better the olfactory function. A score of ≤ 16.5 is considered as anosmia, a score of > 30.5 as normosmia, and scores between these values are considered as hyposmia. A difference of 5.5 on TDI score was determined as a clinically relevant difference for the primary outcome [23]. Secondary objective outcome was gustatory function measured by Taste Strip Test (TST, Burghart), assessing recognition thresholds and identification of the four basis tastes [24]. Total taste score range from 0 to 16 since scores for each taste range from 0 to 4. High scores indicate a better taste function. Clinical improvement was set at > 2 points [25].

Secondary subjective outcomes were olfactory, gustatory and trigeminal function, impact of smell/taste changes on quality of life, and nasal symptoms measured by subjective questionnaire outcomes. These contained the validated Sino-Nasal Outcome Test-22 questionnaire (SNOT-22) [26], self-reported smell, taste, trigeminal sensations questionnaire by medians of visual analog scale (VAS), ranging 0–10 [27], and the translated Olfactory Disorders Questionnaire (ODQ) [28, 29]. All outcomes where assessed during both first and second visit. Details of all outcomes, examinations, and questionnaires are reported in the protocol, Sect. 7.5 and 8.1.2 (appendix), http://dx.doi.org/10.1136/bmjopen-2021-060416.

All statistical analyses were performed using the IBM SPSS Statistics 26.0.0.1 software and R statistical computing. We performed analysis on an intention-to-treat basis. Sample size was calculated based on means and standard deviations of an earlier pilot study [19]. With a power of 0.90, an alpha of 0.05, and a mean difference of 5.5 (SD 8.0) on SST-scores, the total sample size was 92. To correct for possible non-parametric testing, the sample size was increased with 15%. As the study is limited in time and effort for the patient, a maximum of 10% dropout was expected. This gives a total sample size of initial 116 patients, with 58 in every group. A test for normality was used to assess whether variables were normally distributed. Since all our outcomes were not-normally distributed, a Mann–Whitney U test was performed to determine statistical significant differences between the prednisolone and placebo group. The differences in continuous variables between the groups was calculated using Hodges-Lehmann estimation. Confidence intervals for differences between groups were reported.

The national patients association was involved in the conduct of the study, applying for the funding, and in recruiting patients. No patients were involved in the research questions or outcome measurements. We acknowledged and thanks the participants of our trial for their contribution. Patients who participate in this trial, who prefer, will be informed of the results.