A randomised controlled trial testing the efficacy of Fit after COVID, a cognitive behavioural therapy targeting severe post-infectious fatigue following COVID-19 (ReCOVer): study protocol

Several trial registers were searched on 8th and 9th of July 2021. While various trials testing the efficacy of exercise therapy and rehabilitation on fatigue (as primary or secondary outcome) were identified, no trial has been registered testing the effect of CBT or another psychological intervention on fatigue following COVID-19.

Patients are recruited by medical professionals of participating hospitals in the Netherlands: Amsterdam University Medical Centers (Amsterdam UMC), location Academic Medical Center (AMC) and VU medical center (VUmc) in Amsterdam, Radboud university medical center (Radboudumc) in Nijmegen, Jeroen Bosch Hospital (JBZ) in Den Bosch, Catharina hospital in Eindhoven and Bernhoven in Uden. In addition, general practitioners (GP’s), public health services, patient organisations and the general public are informed about the study by flyers, leading to referrals via GP’s and self-referrals. For recruitment via self-referral, the study is advertised via social media, i.e. LinkedIn and Facebook, and adverts in local newspapers. Additionally, a study-website (www.​moenacovid.​nl) has been developed, where interested individuals can find information about the study, fill out a brief test to check whether they fulfil inclusion criteria, and find information on how to contact the research assistant.

After (self)referral, interested patients are contacted by telephone by the research assistant to inform them about the study. All patients are screened for eligibility. The inclusion and exclusion criteria are shown in Table 1. For patients referred via a physician, the medical inclusion criteria (a and b) and exclusion criterion a (a somatic condiotion that can explain the fatigue) are checked by a physician or research nurse to ensure the patient has been diagnosed with COVID-19 and has no other psychiatric or somatic condition that can explain the fatigue. For self-referrals, these eligibility criteria are checked by contacting the patient’s GP with the permission of the patient. Self-referrers are also asked to send a copy or screenshot of their positive SARS-CoV-2 test result. In case the information in the medical record of the GP is insufficient to determine eligibility, self-referrals are sent to the outpatient clinic of the Amsterdam UMC (location AMC) or Radboudumc for screening. After that, they follow the procedure for inclusion of patients directly referred by a physician.

After obtaining informed consent, online screening questionnaires are administered by the researcher to verify the remaining eligibility criteria (see Table 1). The subscale fatigue severity of the CIS-questionnaire [38] is used for inclusion criterion c, to screen for the presence of severe fatigue. For inclusion criterion d, the presence of limitations in physical or social functioning, the subscale physical functioning of the Short Form Health Survey (SF-36) [40] and the Work and Social Adjustment Scale (WSAS) [41] are used. Participants are screened for the presence of post-traumatic stress disorder (PTSD) with the PTSD Checklist for DSM-5 (PCL-5) [42], and for the presence of a depressive disorder with the Beck Depression Inventory for Primary Care (BDI-PC) [43, 44] (exclusion criterion a, i.e. the presence of a psychiatric condition that can explain the fatigue). If the score on the BDI-PC is ≥ 4 or the score on the PCL-5 is ≥ 33, the Mini-International Neuropsychiatric Interview (M.I.N.I.) [45] is conducted by a trained research assistant via phone to determine if patients meet the criteria of a depressive disorder or PTSD. In that case, patients will not be included. Instead, the GP is contacted for referral to appropriate treatment. If two persons from the same household are screened and found to be eligible, only one of them can participate (based on their choice). This is to prevent contamination. The other person is allocated to the same condition outside the study context. If a patient has been infected twice, the first diagnosis with COVID-19 or hospital discharge must be from 3 months up to and including 12 months ago (inclusion criterion a). Only in those cases where fatigue commenced or increased substantially directly after the second infection, the second infection is used as the reference point to determine eligibility.

At the beginning of the study, the upper limit for including patients was extended from initially 6 months to 12 months after diagnosis or hospital discharge (inclusion criterion b). This enhanced the feasibility of the project as patients from the first COVID-19 wave in the Netherlands were then eligible when initiating recruitment in October 2020.

Inclusion criterion a, symptomatic and confirmed COVID-19, was adapted following the introduction of new SARS-CoV-2 tests. Part of original criterion a, typical symptoms and being part of a household in which another person tested positive for SARS-CoV-2 by PCR 2 weeks before or after the first day of illness, was dropped because of increased test capacity. No patients were included based on this criterion.

Eligible and consenting participants start with the baseline assessment (T0). Following T0, participants are randomised to the CBT group or care as usual. In case of randomisation to CBT, the first treatment session is planned within 2 weeks after randomisation. Nineteen weeks after randomisation all participants are assessed again (T1). For participants assigned to CBT this is the post-intervention assessment. At follow-up (T2), 6 months after T1, all participants are assessed again. Participants randomised to the CBT group are also assessed at long-term follow-up (T3), 12 months after treatment. Due to ethical considerations, participants from the care as usual arm are offered referral for existing CBT for chronic fatigue after the follow-up assessment (T2) and hence, cannot serve as a control for the long-term follow-up assessment. A flow-chart of the study design is shown in Fig. 1.

CBT for severe fatigue has been found to be effective in several RCT’s [27, 29, 30]. The efficacy of the internet version of this treatment protocol is also confirmed [29]. The adapted version targeting fatigue after COVID-19 is named Fit after COVID (in Dutch: Fit na COVID). The main part of the intervention is based on an existing CBT manual for Chronic Fatigue Syndrome (CFS) by our research group [46] and was adapted by experienced cognitive behavioural therapists (HK, TK). As the original CBT manual for CFS, Fit after COVID is based on a cognitive behavioural model of fatigue [26]. According to this model, a disease or stressor (here: COVID-19) initially triggers fatigue while cognitive behavioural variables perpetuate fatigue. The seven perpetuating factors addressed in Fit after COVID are (1) disrupted sleep-wake pattern, (2) dysfunctional beliefs about fatigue, (3) low or unevenly distributed level of activity, (4) perceived low social support, (5) problems with processing the acute phase of COVID-19, (6) fears and worries regarding COVID-19, and (7) poor coping with pain.

In the adapted version, the psychoeducation and the explanation about perpetuating factors now consistently refer to COVID-19 as the trigger of the fatigue. The modules on processing the acute phase of COVID-19 and fears and worries regarding COVID-19 were adapted from a treatment manual for fatigue in cancer survivors (i.e. coping with cancer and cancer treatment, and fear of cancer recurrence) [47].

The intervention Fit after COVID is personalised in two ways. First, the content of the core modules is adapted to the participant based on the baseline assessment. The module on disrupted sleep-wake pattern is adapted based on the sleep diary and the Insomnia Severity Index (ISI) [48]. The module on dysfunctional beliefs about fatigue is adapted based on questionnaires assessing specific dysfunctional beliefs, i.e. catastrophizing assessed by the Jacobson-Fatigue Catastrophizing Scale (J-FCS) [49], focussing on fatigue assessed by the Illness Management Questionnaire (IMQ) [50] and low self-efficacy assessed by the Self-Efficacy Scale (SES) [51, 52]. The module on graded activity contains a version for relatively active participants and a version for low active participants, based on the activity pattern assessed with an actigraph [53]. Participants with low activity pattern immediately start with a gradual increase in their daily physical activity, while participants with a relative active activity pattern learn first to evenly distribute their activities during the day and then subsequently gradually increase their daily activity. As the second way to personalize the treatment, participants only follow the optional modules aimed at fatigue-perpetuating factors that apply to them. The modules are selected based on their scores on baseline questionnaires and on information collected by the therapist during the intake session. Questionnaires indicating whether optional modules are relevant are the Van Sonderen Social Support Inventory Discrepancy scale (SSL-D) and Interactions scale (SSL-I) [54, 55] for the module on perceived low social support, the Impact of Event Scale (IES) [56, 57] for the module on problems with processing the acute phase of COVID-19 and the subscale pain of the SF-36 [40] for the module on poor coping with pain. For the selection of the module on fears and worries regarding COVID-19, the Cancer Worry Scale (CWS) [58] was adapted to the COVID-19 Worry Scale (COWS).

See Table 2 for an overview and description of each module, the assessment tools and accompanying cut-off points.

Six therapists working at the participating hospitals completed a 4-day training programme provided by a senior clinical psychologist (HK). The training consists of discussing each treatment module and practising the interventions in role-playing with simulation patients, being professional actors. An introduction to the online platform, exercises in writing e-mails and giving online feedback are part of the training. Five therapists working at the Expert Centre for Chronic Fatigue (ECCF) of the Amsterdam UMC were already trained in the protocol for CBT for CFS, but followed an additional 2-hour training programme on COVID-19-specific issues.

All therapists are supervised bi-weekly by an experienced clinical psychologist (HK, TK) to ensure treatment integrity. In this supervision, all cases are discussed. Therapists register each face-to-face, video, phone and e-mail contact with the participant. The duration of the contact and treatment modules delivered are registered. All activities of the participant and therapist on the platform are registered via log data and are available for analysis.

Allocation of participants to either CBT or care as usual is stratified based on (1) illness severity during the acute stage of COVID-19 (no admission to hospital; admitted to hospital, not on the intensive care unit (ICU); ICU during hospitalisation) and (2) dyspnoea based on the Medical Research Council (MRC) [59] score (< 3 vs. ≥ 3). This is because dyspnoea might influence the response to CBT by interfering with the graded activity module. A web-based randomisation programme in Castor EDC [60] is used to generate the randomisation sequence. The allocation ratio is 1:1. Block-randomisation with randomly selected block sizes (i.e. 2, 4 or 6) is used. The randomisation is performed by a research assistant in the presence of the participant who is on the phone. The research assistant, the researcher and the participants are blinded to the allocation sequence. Due to the nature of the intervention, blinding to allocation is not feasible. Data will be analysed by an independent statistician using a data file which is blinded for treatment allocation.

Outcome measures consist of self-reported questionnaires, which are assessed at T0, T1, T2 and T3. See Table 3 for the measurements at all time points. The questionnaires are administered online using Castor EDC [60]. Participants receive an e-mail with a personal link to fill out the questionnaires. For optimal retention, participants are contacted by the research assistant if they have not completed the questionnaires within one week. When participants are not willing to complete all measurements at T1, T2 and/or T3, they are asked to at least fill out the CIS-fatigue [38], which is the primary outcome measure. At T0 and T1, data on physical activity level and sleep are also gathered by using an actigraph and completing a sleep diary.

Fatigue severity is assessed by the subscale fatigue severity of the CIS [38]. The CIS consists of 20 items scored on a 7-point Likert scale and measures four dimensions of fatigue, i.e. fatigue severity, concentration problems, motivation and activity. The CIS-fatigue, the primary outcome, consists of eight items, each item is scored on a 7-point Likert scale. The range of scores is 8 to 56, with a higher score indicating more severe fatigue. The cut-off score for severe fatigue is ≥ 35 [38] on the CIS-fatigue. Previous research supports the reliability and validity of the CIS [38, 61, 62].

Limitations in physical functioning are assessed by the subscale physical functioning of the SF-36 [40] which consists of ten items scored on a 3-point scale. The weighted subscale score ranges from 0 to 100, with a higher score indicating less limitations. A score of ≤ 65 is indicative of substantial limitations in physical functioning [63]. The SF-36 is a reliable and valid instrument [64].

Social disability is assessed by the WSAS [41, 65] which consists of five items assessing different domains of functioning, e.g. work and social leisure activities, scored on an 8-point Likert scale. A score of ≥ 10 is indicative of significant functional impairment [41, 65]. The WSAS is a reliable and valid instrument [41, 65.

Somatic symptom severity is assessed by the somatic symptoms scale of the Patient Health Questionnaire (PHQ) [66]. The PHQ-15 consists of 15 items referring to somatic symptoms. Each symptom is scored on a 3-point scale with scores of 5, 10 and 15 indicate low, medium, and high symptom severity. Research supports the reliability and validity of the PHQ-15 [66].

The demographic characteristics include age, sex, relationship/cohabitation status, level of education and employment status. Patient characteristics include lenght, weight, morbidities, hospital/ ICU admission, start date of COVID-19 symptoms, date of COVID-19 test, type of COVID-19 test, physician visit for COVID-19 and vaccination status. All of the characteristics are assessed by self-report during screening. Vaccination status is also assessed at T1.

Activity: An actigraph is used to assess the participant’s level of physical activity. The actigraph is worn around the wrist for 14 consecutive days and nights for a reliable estimate of daily activity for 12 full days. The actigraph has been shown to be a reliable and valid instrument for the assessment of physical activity [53, 67].

Sleep parameters: During the 14 days participants wear the actigraph, a sleep diary [68] is also completed daily, if possible within one hour of getting out of bed in the morning. Participants have to fill in (1) the time they go to bed (in hours and minutes), (2) the time they try to fall asleep, (3) the sleep onset latency, (4) time awake at night, (5) time out of bed, (6) time of waking-up, (7) time of getting out of bed and (8) a rating of the sleep quality (scale 0–10) [69]. Sleep problems are also assessed by the ISI [48] and the subscale sleep-rest of the Sickness Impact Profile (SIP) [70, 71].

Fatigue-related behaviours, cognitions and emotions: Based on the cognitive behavioural model for fatigue in CFS and fatigue in other long-term conditions, several fatigue-related behaviours, cognitions and emotions are assessed. These are dysfunctional beliefs assessed by the J-FCS [49], IMQ [50] and the SES [51, 52, social support assessed by the SSL-D and SSL-I [54, 55], problems with processing assessed by the IES [56, 57], and pain assessed by the subscale pain of the SF-36 [40]. To assess COVID-19 related worries, we adapted the Cancer Worry Scale [58]. The resulting COVID-19 Worry Scale (COWS) consistently refers to COVID-19 instead of cancer. For participants randomised to CBT, the T0 actigraph data, sleep data and questionnaires about fatigue-related behaviours, cognitions and emotions are used to tailor Fit after COVID to the individual (see Intervention paragraph).

Other relevant parameters: The criteria for CFS are assessed by the US Centers for Disease Control and Prevention (CDC) criteria revised in 2003 [72]. Mental health is assessed by the patient health questionnaire, somatic, anxiety and depressive symptoms (PHQ-SADS) [73]. Social functioning is assessed by the subscale social functioning of the SF-36 [40] and the SIP [71]. Illness perceptions are assessed by the Brief Illness Perception Questionnaire (B-IPQ) [74, 75]. Illness attributions are assessed by the Causal Attribution List (CAL) [76]. Responses to symptoms are assessed by the Cognitive Behavioural Responses to Symptoms Questionnaire (CBRSQ) [77, 78]. Fear of movement is assessed by the Tampa Scale of kinesiophobia (TSK) [79]. Invalidation experiences are assessed by the Illness Invalidation Inventory (3*I) [80]. Resilience is assessed by the Brief Resilience Scale (BRS-6) [81, 82]. Expectations about treatment are assessed by the Treatment Outcome Expectations Questionnaire (TOEQ) [83]. All study parameters are assessed as shown in Table 3.

The primary objective is to investigate whether CBT will lead to a significantly lower mean fatigue severity score measured with the CIS-fatigue across follow-up visits T1 and T2 as compared to care as usual.

Secondary objectives are to investigate whether CBT as compared to care as usual will result in:

At T1, all participants are asked to report if they experienced new symptoms or an increase in existing symptoms during CBT or care as usual. All AE’s reported by the participant or observed by the therapists or research staff are recorded and reported.

Personal data is handled confidentially and is coded in compliance with the EU General Data Protection Regulation Act (in Dutch: Uitvoeringswet AVG, UAVG). Identifying patient information is coded for all study procedures. Only the PI (HK) and one researcher (TK) are allowed to access the keyfile with both participant codes and identifying participant data. The data collected with the actigraph is stored and processed locally on the secure servers of the Amsterdam UMC. The sleep diary is completed by the participants on a paper-and-pencil sheet. This sheet does not include personal information but only the patient’s identification code. The data cannot be traced back to specific participants in reports and publications on the study. Data will be stored at the Department of Medical Psychology of the Amsterdam UMC, location AMC, for 15 years following completion of the project. Data of Fit after COVID on the online platform will also be stored for 15 years on a secure server.

This study is subject to on-site monitoring in accordance with the quality assurance advice of the Dutch Federation of University Medical Centres regarding research involving human subjects [85]. On-site monitoring is based on the risk classification negligible. The investigator will submit a summary of the progress of the trial to the ethics committee once a year. Information will be provided on the date of inclusion of the first subject, numbers of subjects included, numbers of subjects that have completed the trial, SAEs and amendments.

This study was reviewed and approved by the medical ethics committee of the Amsterdam UMC, location AMC (registration nr. 2020_182, NL74828.018.20) and by the local ethics committees of the participating hospitals. The study is registered in the Netherlands Trial Register (NTR) (NL8947). The NTR is updated in case of changes to the study protocol. Written informed consent will be obtained from all participants by the researcher (TK) and the research assistant (DP). All changes in the protocol are sent for approval to the medical ethics committee.

Participants can leave the study at any time for any reason without any consequences. The reasons are documented. The researcher can withdraw a participant from the study in case of incorrect enrolment. Participants are informed of their possibility to withdraw from the study during the intake and in the information letter. Despite leaving or being withdrawn from the study, treatment can be continued. The investigator, the treating physician, or the treating psychologist can decide to withdraw a participant from the study for urgent (medical) reasons. These reasons are documented. Randomised participants who drop out of the study are not replaced.

The sample size calculation is based on testing the primary hypothesis that CBT leads to a significantly lower mean fatigue severity score (CIS-fatigue) across the first two follow-up assessments (T1 and T2) as compared to care as usual. A difference of 6 points on the CIS-fatigue is considered clinically relevant [38]. Based on previous research we assume a common standard deviation of 12 and correlation coefficients of 0.4 among CIS-fatigue scores assessed at T1 and T2 [27, 31].

With a sample size of 45 in group 1 (CBT) and 45 in group 2 (care as usual), a two-sided test for the time averaged difference between two means in a repeated measures design with a 0.05 significance level has 80% power to detect a difference in means of 6 in a design with 2 repeated measurements when the standard deviation is 12 and the between-level correlation is 0.4. Assuming a conservative dropout of 20%, we will randomise 114 participants, 57 to each condition.

Descriptive statistics (i.e. participant characteristics) will be presented in tables, with separate columns for participants randomised to CBT vs. participants randomised to care as usual. Categorical variables (e.g. sex, relationship status) will be presented in participant numbers (n) and percentages (%). Continuous variables (e.g. age, CIS-fatigue baseline score) will be presented in means and standard deviations (SD) or, if appropriate due to skewness of data, as median with its interquartile range. Participant numbers, dropout and reasons for dropout will be presented in the study flow-chart. In all analyses, two-sided p values of < 0.05 are considered to indicate statistical significance. All analyses will be conducted in SPSS.

Change in fatigue scores between T0 and T1 for each individual participant will be shown in a figure using the Leeds RCI calculator [86]. This figure will show whether the change was clinically significant and reliable, or reliable only, and if there was no change or a deterioration in fatigue.

The primary study parameter will be analysed according to intention-to-treat, i.e. all randomised participants will be included in the analysis and they are analysed in the group to which they were randomly allocated. The primary hypothesis will be tested using a mixed linear model. The model will include T1 and T2 fatigue severity score as a dependent variable, condition (CBT vs. care as usual), time (T1, T2), condition by time interaction as fixed effects, a random intercept and the fatigue severity score at baseline as covariate. It will be examined if the main effect of condition from the mixed linear model, i.e. the mean difference in CIS-fatigue scores across T1 and T2 while controlling for baseline fatigue, is statistically significant. If the main effect of condition is found to be statistically significant, the statistical significance of the between group differences at the separate time points (T1 and T2) will be interpreted.

Cohen’s d effect sizes will be calculated by dividing the parameter estimate for the mean difference in CIS-fatigue scores between both conditions from the mixed linear model by the pooled standard deviation of both conditions combined. Effect size magnitudes will be interpreted as small (0.2 to 0.5), medium (0.5 to 0.8) and large (greater than 0.8) [87].

To explore the robustness of our findings from the primary analysis, four sensitivity analyses will be conducted. First, per protocol analysis will be conducted. Treatment completion will be operationalized in two ways, i.e. by including (a) participants who have filled out the treatment goals and opened the standard modules of the intervention and (b) participants who have completed the intervention according their therapist.

Second, as the primary outcome consists of the mean CIS-fatigue score across T1 and T2, the extent to which our result would change if based on a single time point will be explored. Therefore, the analysis with the CIS-fatigue score at T1 as a dependent variable and with the CIS-fatigue score at T2 as a dependent variable adjusted for CIS-fatigue score at baseline using analysis of covariance will be repeated. For each of the time points Cohen’s d effect sizes will be calculated.

Third, the extent to which dyspnoea at T0, disease severity, time since diagnosis of COVID-19, age and sex have an impact on the primary outcome will be explored, and whether this impact differs between CBT and care as usual. These analyses will be conducted by including these variables and their interaction-terms with condition as covariates in the mixed linear models.

Fourth, missing values will be replaced with multiple imputation using chained equations and the data will be reanalysed. The imputation model will include all baseline sociodemographic and clinical characteristics assessed at baseline and follow-up measurements. A total of five data sets will be imputed and pooled according to Rubin’s rule.

Secondary outcomes will also be analysed according to intention-to-treat. For T1 and T2, it will be determined if there is a clinical meaningful improvement in fatigue, defined as a RCI [84] of more than 1.96 and a decrease of the fatigue level to a normal range (i.e. a score of < 35 on the CIS-fatigue). For secondary outcomes that are dichotomous, separate logistic regression analysis will be done for T1 and T2. For secondary outcomes that are continuous (i.e. physical functioning, work and social adjustment, somatic symptoms, cognitive symptoms), differences between CBT and care as usual will be analysed using mixed linear models as described above.

For the long-term follow-up (T3) examining the course of fatigue over time within the CBT group, linear mixed model analyses will be applied, where the depended measure CIS-fatigue is assessed at T0, T1, T2 and T3.