Study design and rationale

To identify the effect of a targeted modification of illness-related anxiety and dysfunctional symptom expectations on Long COVID symptoms, and to differentiate this effect from general modes of action, a randomised comparison between a specifically treated group, a group treated non-specifically in the same dose and a control group without additional treatment is required. A control group is necessary to test whether the experimental interventions have a positive effect compared with no intervention at all, and to investigate objectives 2 and 3. Thus, we will use the design of a three-arm randomised controlled trial, in which one third of participants will undergo targeted expectation management in addition to treatment as usual (TAU), one third will undergo unspecific supportive treatment in addition to TAU, and one third will receive TAU only. In the control group, we will additionally investigate the contribution of predefined risk factors to symptom persistence in Long COVID. The study will be fully observer-blinded with respect to all treatment conditions. It will be conducted monocentric with nationwide recruitment.

Setting

For recruitment, we will use our pulmonary out-patient clinic as well as our psychosomatic out-patient clinic and our established network of general practitioners and pulmonologists. We will also recruit via social media campaigns with support of cooperating self-help groups and patient organisations (e.g. ‘Landesverband Sachsen für Prävention und Rehabilitation von Herz-Kreislauf-Erkrankungen e.V. ‘, ‘Long COVID Vernetzungsstelle’, ‘Nationale Kontakt- und Informationsstelle zur Anregung und Unterstützung von Selbsthilfegruppen’). The experimental interventions will be carried out as personal online video consultations, which allows for a nationwide outreach of our study.

Inclusion criteria

Inclusion criteria are as follows: self-reported resolved SARS-CoV-2 infection confirmed by a positive polymerase chain reaction, serology or rapid antigen test; presence of Long COVID according to the NICE/AWMF S1-guidelines,⁷ at least moderately severe ongoing symptoms (Patient Health Questionnaire 15 (PHQ-15) score ≥10),^{Reference Kroenke, Spitzer and Williams25} aged ≥18 years and provision of informed consent.

Exclusion criteria

Exclusion criteria are as follows: acute SARS-CoV-2 infection, intensive care unit treatment for COVID-19, psychotherapeutic treatment in the past 3 months, necessity of acute emergency treatment, acute suicidality, a substance use disorder, acute psychosis, cognitive incapacity to comprehend the study materials, inability to complete outcome measures online or insufficient German language skills.

Experimental intervention 1 (COV.EXPECT ± TAU)

The experimental intervention consists of an expectation management intervention (COV.EXPECT) in addition to TAU. The manualised intervention primarily aims to reduce illness-related anxiety and optimise expectations about symptoms, treatment outcome and coping strategies.^{Reference Pan, Meister, Löwe, Kaptchuk, Buhling and Nestoriuc26} The design and dose of the intervention are based on the demonstrated effectiveness of the expectation management intervention from the PSY-HEART trial (Psychological Preoperative Interventions to Improve Outcome in Heart Surgery Patients),^{Reference Rief, Shedden-Mora, Laferton, Auer, Petrie and Salzmann27} on the SOMA. GUT study (Persistence of Gastrointestinal Symptoms in Irritable Bowel Syndrome and Ulcerative Colitis: From Risk Factors to Modification) within the SOMACROSS Research Unit^{Reference Löwe, Nestoriuc, Andresen, Vettorazzi, Zapf and Hübener28} and on other previous studies.^{Reference Pan, Meister, Löwe, Kaptchuk, Buhling and Nestoriuc26} The intervention consists of three individual online video consultation sessions (conducted fortnightly) and a booster session after 3 months, with each session lasting 45 min. In the first session, the patient's expectations regarding symptoms and treatment will be assessed through a semi-structured interview so that the intervention can be adapted to the individual patient expectations within the framework of the treatment manual. The intervention components include psychoeducation aimed at developing a biopsychosocial model of Long COVID and realistic, yet functional expectations regarding symptoms and treatment outcome, visualisation techniques to foster expectations of personal control and developing personal goals in managing symptoms to improve coping expectations. In a ‘toolbox’, illness-specific dysfunctional expectations (e.g. ‘I can't take the stairs because I'm too short of breath’) are assigned to specific therapeutic interventions (e.g. cognitive restructuring, fostering positive expectations via visualisation and addressing confidence in expectations, counteracting cognitive immunisation). To deepen the acquired skills, homework will be given after each session and discussed at the beginning of the following one. The intervention thus addresses the topics ‘dealing with anxiety’, ‘improving expectations’ and the need for information about their disease. By conducting focus groups, the intervention protocol and manual will be finalised in collaboration with patients from a Long COVID group currently offered in our psychosomatic out-patient clinic.

Experimental intervention 2 (COV.SUPPORT ± TAU)

The experimental intervention consists of a non-specific supportive intervention (COV.SUPPORT) in addition to TAU. COV.SUPPORT is identical to COV.EXPECT in terms of common and non-specific treatment elements (i.e. time, personal attention and emotional support), but does not use specific interventions to modify illness-related anxiety and expectations. In contrast to COV.EXPECT, which focuses primarily on changing dysfunctional symptom expectations for the future, COV.SUPPORT focuses exclusively on coping with stressful situations in the present. COV.SUPPORT is manualised and adapted from the supportive therapy we use in the PSY-HEART-II trial.^{Reference Salzmann, Laferton, Shedden-Mora, Horn, Gärtner and Schröder29} This unspecific intervention is needed to differentiate our proposed mechanism of action from general modes of action.

Control intervention (TAU)

The control intervention consists of TAU only. TAU in all study groups implies that the patient receives their usual treatment without any interference by the study. This group is also needed to identify risk factors for the persistence of Long COVID symptoms (objective 2) and compare risk factors for symptom persistence across conditions with results from the SOMACROSS Research Unit (objective 3).

Measurement points

Assessments are carried out at baseline and after 6 weeks (intermediate), 3 months (post-interventional) and 6 months (follow-up). The intermediate assessment after 6 weeks is necessary for mediator analyses that investigate whether a change in symptom severity is mediated via changes in illness-related anxiety and dysfunctional symptom expectations. Except for an additional 12-month follow-up in the SOMACROSS Research Unit, assessment times are identical, allowing for comparison of results. Primary and secondary outcomes and mediator variables will be collected through electronic data entry by patients at home; only the diagnostic interview will be collected via telephone interview. The primary and secondary outcome variables and mediator variables will be measured at each assessment point, and are displayed in Table 1.

Table 1 Outcome instruments of the SOMA.COV study

Primary outcome

To test the effect of the expectation management intervention on Long COVID symptoms, the primary outcome for this study is the baseline to post-interventional change in overall somatic symptom severity (3-month follow-up). As harmonised core outcome sets for Long COVID and post-COVID-19 conditions are currently still being developed, overall symptom severity will be assessed with the established PHQ-15,^{Reference Kroenke, Spitzer and Williams25} which is validated in English and German in various diseases and has proven its sensitivity to change as an outcome instrument in many studies.^{Reference Kroenke, Spitzer, Williams and Löwe30} On a scale of 0 (‘not bothered at all’) to 2 (‘bothered a lot’), the PHQ-15 uses 15 items to measure somatic symptom severity during the past 4 weeks. It includes most of the common Long COVID symptoms and enables us to compare our results to the SOMACROSS Research Unit, which also uses the PHQ-15 as a primary outcome instrument.^{Reference Löwe, Andresen, Van den Bergh, Huber, von dem Knesebeck and Lohse9}

Secondary outcomes

Secondary outcomes include symptoms that have been found to be of special relevance in Long COVID, such as fatigue,^{Reference Chalder, Berelowitz, Pawlikowska, Watts, Wessely and Wright31} post-exertional malaise^{Reference Cotler, Holtzman, Dudun and Jason32} and pain.^{Reference Tait, Pollard, Margolis, Duckro and Krause33} We will also include a self-developed screening questionnaire on Long COVID as well as other post-infectious symptoms. This self-report instrument, called the PHQ-15 Post-Acute Infection Syndromes (PHQ-15 PAIS), is based on the PHQ-15^{Reference Kroenke, Spitzer and Williams25} and extended for the most common symptoms in post-acute infection syndromes.^{Reference Choutka, Jansari, Hornig and Iwasaki34} Illness-related anxiety (Somatic Symptom Disorder – B Criteria Scale (SSD-12)),^{Reference Toussaint, Murray, Voigt, Herzog, Gierk and Kroenke35} expectations of symptom severity, burden, treatment outcome and coping with symptoms (Treatment Expectation Questionnaire (TEX-Q); European Research Network to Improve Diagnosis, Treatment and Health Care for Patients with Persistent Somatic Symptoms (EURONET-SOMA) numeric rating scales)^{Reference Rief, Burton, Frostholm, Henningsen, Kleinstäuber and Kop36,Reference Alberts, Löwe, Glahn, Petrie, Laferton and Nestoriuc37} will be investigated as pre-specified mediator variables. The treatments received will be documented. In addition to Long COVID-specific outcome variables, we will apply the joint core measures of the SOMACROSS Research Unit to identify further risk factors and mechanisms for symptom persistence in Long COVID, and to make the results of the studies comparable. The joint core set of instruments is described in detail in the overarching protocol of the SOMACROSS Research Unit.^{Reference Seeßle, Waterboer, Hippchen, Simon, Kirchner and Lim12} Supplements from the core set include, for example, adverse childhood experiences, negative affectivity, stigmatisation, health-related quality of life and healthcare utilisation. As in the research unit, we will also conduct the Structured Clinical Interview for the DSM-5 (SCID-5) for the diagnosis of somatic symptom disorder according to the DSM-5, to investigate the diagnosis as a potential predictor of treatment response.

Safety outcomes

To the best of our knowledge, there is no risk for serious adverse events caused by the application of expectation management interventions.^{Reference Rief, Shedden-Mora, Laferton, Auer, Petrie and Salzmann27} However, patients may develop severe somatic complications of Long COVID or other medical conditions. In these cases, patients will be informed and advised to initiate appropriate treatment. In case of an emergency, medical treatment will be initiated at the University Medical Center Hamburg-Eppendorf. Any adverse events will be systematically recorded at each follow-up assessment.

Sample size

This trial is powered with regard to the difference between intervention 1 (COV.EXPECT + TAU) and the control condition (TAU), as this is the main interest. Because of the closed testing principle, the type one error (alpha) does not need to be adjusted for multiplicity. As there are no comparable trials on Long COVID available, we based our estimation both on our DEPSCREEN-INFO trial (Increasing the Efficiency of Depression-Screening using Patient-Targeted Feedback),^{Reference Löwe, Blankenberg, Wegscheider, König, Walter and Murray38} which yielded a between-group effect size of d = 0.33 in PHQ-15 scores with an intervention of much lower intensity (written feedback after depression screening), and on the PSY-HEART trial,^{Reference Rief, Shedden-Mora, Laferton, Auer, Petrie and Salzmann27} which yielded a medium between-group effect size of d = 0.71 with an expectation management intervention, but using a different outcome instrument (modified version of the Pain Disability Index). Based on these two studies, we expect a medium effect size (d = 0.50) that can be detected with a power of 80%, using a two-sided alpha of 5%, by including 64 patients in each group, resulting in a total sample size of N = 192. Based on the results of our previous studies, we assume a loss to follow-up between baseline and the primary outcome measurement (i.e. 3-month follow-up) of 25%, resulting in a total of 258 randomised patients. Assuming that 50% of patients will meet the inclusion criteria, a total of 516 patients will be assessed for eligibility. Figure 2 shows the anticipated flow of participants throughout the trial.

Fig. 2 Anticipated flow of participants through the course of the study. Outcomes after 6 months are secondary and were not included in the sample size estimation. COV.EXPECT, expectation management intervention; COV.SUPPORT, supportive intervention.

Statistical methods

The primary analysis and all prespecified secondary analyses will be conducted in the intention-to-treat sample consisting of all randomised patients with the full analysis set, which is as close as possible to the intention-to-treat population. In secondary analyses, the data of the per-protocol population (subgroup of patients without major protocol violations) will be used and multiple imputation will be applied.

For objective 1, a linear mixed-effects model with the change from baseline for all measured time points will be used to investigate the group differences in the PHQ-15, with treatment group, time and gender as fixed effects, patient as the random effect and baseline PHQ-15 score as the covariate. The interaction between time and treatment group will be tested and excluded from the model if P ≥ 0.05. For primary hypothesis, the global treatment effect after 3 months will be considered first. If the two-sided P-value is <0.05, the contrasts in change after 3 months for COV.EXPECT versus TAU, COV.SUPPORT versus TAU and COV.EXPECT versus COV.SUPPORT at the 5% level (two-sided hypothesis) will be considered. Because of the closed testing principle, an adjustment of the contrast P-values for multiplicity is not necessary.

Secondary end points will be analysed analogously to the primary end point, with the regression model corresponding to the measurement scale (mixed linear model for metric outcomes, mixed logistic model for binary outcomes, cumulative logit model for ordinal outcomes).

In subsidiary analyses, a potential effect of recruitment settings will be investigated by including the setting as a further fixed effect in the analysis. To analyse whether treatment effects on Long COVID resulted through changes in illness-related anxiety or symptom expectations, we will conduct mediation analyses.

To identify further risk factors involved in the persistence of Long COVID (objective 2), we will use longitudinal data from the control group and conduct multivariate regression analyses with symptom persistence as outcome, taking into account the number of predictors and sample size. Patients from the intervention groups will not be included in these analyses because the interventions could bias the results.

To compare risk factors and mechanisms of somatic symptom persistence in Long COVID exploratively with those identified in the other conditions investigated in the SOMACROSS Research Unit (objective 3), we will compare the results of the disease-specific regression analyses from the SOMACROSS Research Unit with the derived model from objective 2, and conduct further exploratory analyses.

Subgroup analyses will be conducted on differences in the primary outcome with respect to the following variables of interest: age, gender, education level, social support, duration and severity of Long COVID, any chronic physical disease, any mental health comorbidity, depression and anxiety. For safety analyses, adverse events will be compared between groups and analysed descriptively. Further details of statistical analyses will be described in detail in a statistical analysis plan, which will be finalised before breaking the treatment blind.

Methods against bias

A fixed randomisation list with variable block size and length, stratified by gender, will be generated by an independent member of the biostatistics unit and uploaded into REDCap version 13.7.7 for Windows (REDCap Consortium; see www.projectredcap.org), a software for building and managing online surveys and databases. Assignment will be performed automatically after inclusion in the database. Patient drop-out will be minimised by contacting patients according to a schedule of repeated contact attempts, and by allowing solely written or telephone data collection if electronic collection is not feasible. Because of the electronic collection of outcome variables and having the telephone interviews conducted by trained interviewers who are not involved in the treatment, the study is fully observer-blinded with respect to all treatment conditions. Treating physicians will not be informed about the group allocation or the type of intervention. As in most psychological intervention studies, full patient and therapist blinding is not feasible as their active involvement in the intervention is necessary. Still, patients in the active intervention groups will be blinded. Both interventions will be manualised and therapists and interviewers will be trained and supervised. As a manipulation check regarding potentially overlapping content, contamination and carry-over effects between the two interventions, patients will complete a rating scale on treatment content and on subjective treatment mechanisms after the post-intervention outcome assessment. Potential sampling bias will be avoided by recruiting patients both within and outside of psychotherapeutic care, and by aiming to include both patients with shorter and longer symptom duration in the study. Any questions regarding patient exclusions, serious adverse events and potential study termination will be reviewed by the study's Data Safety and Monitoring Board. Finally, report of the trial and results will follow the Consolidated Standards of Reporting Trials (CONSORT) 2010 recommendations.^{Reference Moher, Hopewell, Schulz, Montori, Gøtzsche and Devereaux39}

Feasibility of recruitment

In previous studies, we were able to successfully recruit patients within our network of cooperating general practitioners, e.g. in the three-arm GET.FEEDBACK.GP-RCT (depression screening using patient-targeted feedback in general practices).^{Reference Kohlmann, Lehmann, Eisele, Braunschneider, Marx and Zapf40} In addition, resident pulmonologists, our pulmonary out-patient clinic and our psychosomatic out-patient clinic, three self-help groups and patient associations, and social media will support recruitment. The format as an online video consultation and the brevity of the intervention will also facilitate patient enrolment. We expect a high willingness to participate in a psychosocial intervention study for Long COVID symptoms, as patients express high needs in this direction and this area is considered to be massively undersupplied so far.

Ethical approval

The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. All procedures involving human patients were approved by the Local Psychological Ethics Committee at the Center for Psychosocial Medicine of the University Medical Center Hamburg-Eppendorf on 14 February 2022 (approval number LPEK-0446). The trial will further be conducted in accordance with guidelines for Good Clinical Practice, and national and local laws. Before inclusion, eligible participants will provide written informed consent. Data will be stored in the REDCap database in pseudonymised form. A corresponding data management manual and a data review plan will be finalised before recruitment of the first patient. Any changes to the study protocol will be listed in the study registry and publications. The study is registered with the ISRCTN Registry (identifier ISRCTN15068418).

Data sharing

In accordance with the ethics committee approval and the DFG guidelines for the handling of research data adopted in 2015, de-identified individual patient data will be made publicly available. Data sharing will be in accordance with the FAIR data principles (Findable, Accessible, Interoperable, and Reusable) and international naming conventions (e.g. Systematised Nomenclature of Medicine) to maximise transparency and scientific reproducibility. According to the WHO Statement on Public Disclosure of Clinical Trials (https://www.who.int/ictrp/reporting-on-findings), the main findings will be submitted for publication in a peer-reviewed journal within 12 months of study completion.