Within-person variability and the minimal important difference of the EQ-VAS in adults with post-COVID-19 condition: a secondary analysis of data from a single-centre randomised controlled trial

Abstract

Objective

To describe the distribution and day-to-day variability of EuroQol Visual Analogue Scale (EQ-VAS) values and to estimate a minimal important difference (MID) in adults with post-COVID-19 condition (PCC).

Design

Secondary analysis of pooled EQ-VAS data from both arms of a 12-week randomised, placebo-controlled trial (PYCNOVID) of Pycnogenol (200 mg/day) versus placebo in adults with PCC.

Setting

Single centre study in Switzerland between June 2023 and July 2024.

Participants

Adults with PCC with at least one of the following symptoms: fatigue, post-exertional malaise, dyspnoea or brain fog.

Main outcome measure

EQ-VAS ratings over 7 consecutive days before baseline (randomisation) and at follow-up (after 12 weeks). Weekly mean EQ-VAS values and within-person SDs (wpSD) were calculated. The MID was estimated using distribution-based and anchor-based methods with prespecified correlation criteria, using the Functional Assessment of Chronic Illness Therapy-Fatigue (FACIT-Fatigue) instrument as the primary anchor, along with additional anchors.

Results

Among 153 participants, baseline EQ-VAS values were low, with a median of 48.7 on a 0–100 scale. Day-to-day variability was considerable, with a median wpSD of 6.4. Distribution-based MID estimates ranged from 5.9 to 8.9 points. FACIT-Fatigue was the only anchor meeting the correlation threshold (ρ≥0.30), yielding an anchor-based MID of 8.96 points. Triangulation of these estimates supported a final MID estimate of 8 EQ-VAS points.

Conclusions

Adults with PCC reported markedly impaired and highly variable health status. Weekly aggregation of daily EQ-VAS ratings likely provides a more robust estimate than single-day assessments. A PCC-specific EQ-VAS MID of 8 points, derived from a chronic, fatigue-predominant PCC population, facilitates interpretation of meaningful change and improves outcome definition and sample size planning in future studies. Applicability to earlier-stage or non-fatigue-predominant PCC populations warrants further investigation.

Trial registration number

NCT05890534.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This study used repeated daily EuroQol Visual Analogue Scale (EQ-VAS) assessments over 7 consecutive days to improve measurement validity.

• The heterogeneous sample, spanning mildly to severely affected participants, enhanced inclusiveness and representativeness of the cohort.

• People with lived experience were actively involved in the design of the study, including feedback supporting EQ-VAS as a relevant patient-reported outcome measure.

• The anchor-based analysis relied on a single anchor assessing fatigue, though this was supported by distribution-based estimates and so may not fully capture change related to other relevant post-COVID-19 sequelae.

• The minimal important difference was derived from a single-centre trial population with long average symptom duration and low baseline health status, which may limit its applicability to broader or less severely affected post-COVID-19 condition populations.

Introduction

Post-COVID-19 condition (PCC) continues to pose a substantial burden on individuals and health systems, with many affected people experiencing persistent symptoms that impair daily functioning, work ability and overall quality of life.^1–3 Due to the absence of reliable objective biomarkers or diagnostic tests, patient-reported outcome measures (PROMs) have become essential for characterising symptom severity and monitoring change over time.⁴ However, the interpretation of PROMs is often challenging due to limited evidence of what constitutes meaningful change. In PCC, health-related quality of life (HRQoL) is frequently assessed using the EuroQoL 5-Dimensional, 5-Level questionnaire (EQ-5D-5L) and Visual Analogue Scale (EQ-VAS).^{5 6} The EQ-VAS is an intuitive and low-burden measure that captures self-rated health on a 0–100 scale, where 0 represents the worst imaginable health and 100 the best imaginable health.⁷ The EQ-VAS is widely used and well validated in chronic respiratory diseases such as chronic obstructive pulmonary disease (COPD). Reported minimal important differences (MID) for chronic respiratory diseases range from about 6.5 to 8 points.⁸ COPD shares key symptoms with PCC, including breathlessness and fatigue but the two conditions differ in their disease trajectories, degree of symptom fluctuations and clinical predictability.^9–11 PCC is characterised by heterogeneous courses, pronounced day-to-day variability and prognostic uncertainty, which contrasts with COPD, where established clinical pathways, evidence-based treatment guidelines and well-defined management strategies support more predictable disease trajectories.^{10 12–14} Furthermore, individuals with PCC experience substantial fluctuations in symptom severity. Therefore, a single EQ-VAS assessment may not fully capture day-to-day variation.^{12 13} Capturing daily assessments of health status likely provides essential context for interpreting whether changes in EQ-VAS values exceed normal symptom fluctuation. The EQ-VAS has also been applied in observational studies, for example, to characterise symptom trajectories and recovery patterns over time, supporting its relevance for assessing overall health status in this population.¹⁵ However, it remains unclear whether the MID derived from chronic respiratory diseases can be transferred to PCC. Recently, MID estimates for EQ-VAS have been reported in a convenience sample of 42 individuals with PCC undergoing a home-based respiratory muscle training programme but its limited generalisability underscores the need for more robust evidence.¹⁶ Establishing a PCC-specific MID for EQ-VAS is essential for meaningful interpretation of changes in health status over time, and for informing trial design and sample size calculations.

Thus, the objectives of this study were to describe EQ-VAS values and their changes over time, to characterise day-to-day within-person variability of EQ-VAS values and to estimate an MID for EQ-VAS in a heterogeneous population of adults with PCC.

Methods

Study design and setting

For this study, we used data from a single-centre, placebo-controlled, quadruple-blind, parallel design randomised superiority trial (PYCNOVID) of adults with PCC, randomly allocated to receiving either Pycnogenol or placebo.¹⁷ Eligibility criteria were adults (≥18 years) with confirmed SARS-CoV-2 infection or physician-diagnosed PCC, reporting at least one key symptom at study entry (ie, fatigue, cognitive impairment, dyspnoea or postexertional malaise), with sufficient language skills, ability to attend study visits (study centre or home visit) and no expected medication changes. Individuals were excluded if they had severe comorbidities (eg, renal failure, advanced heart failure), acute infections, untreated or unstable psychiatric disease, recent COVID-19 vaccination (<4 weeks before baseline or during the study), intolerance to or regular use of Pycnogenol or participation in another interventional study. We used the WHO definition of PCC¹⁸ and assessed each participant’s clinical including the timing of SARS-CoV-2 infection(s), vaccination(s) and the onset and progression of symptoms over time.

The trial was performed at the University of Zurich, Switzerland, from June 2023 to November 2024. Participants completed four study centre visits (or home visits) consisting of screening, baseline (approximately 2 weeks later) and two follow-up visits, the first at 6 weeks after baseline and a second after 12 weeks which was the end of the intervention. For this study, only baseline and follow-up data (after 12-week intervention phase) were used. The study protocol including details about the design, inclusion and exclusion criteria and assessment procedures are described elsewhere.¹⁷ The study is reported in accordance with Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for observational studies, because the current work analyses the trial data as an observational cohort without focusing on treatment effects.¹⁹

Data sources and measurement

Participants completed online questionnaires using the Research Electronic Data Capture (REDCap, Vanderbilt University, USA) and underwent clinical assessments during study visits. Questionnaires assessed demographics and medical history, including chronic conditions.

Self-reported health status was assessed daily through online questionnaires using the EQ-VAS over 7 consecutive days before baseline and again for 7 consecutive days before follow-up (end of intervention). The EQ-VAS ranges from 0 to 100, where 0 indicates the worst and 100 the best imaginable health status.^{7 20} Mean EQ-VAS values were calculated for participants with ≥4 days of data. As the primary anchor for MID estimation, we used fatigue measured with the 13-item Functional Assessment of Chronic Illness Therapy-Fatigue (FACIT-Fatigue) instrument (MID=3).^21–23 The instrument demonstrated responsiveness to change in individuals with PCC following exercise-based rehabilitation.²⁴ Additional anchors included dyspnoea assessed with Chronic Respiratory Questionnaire (CRQ) dyspnoea domain (MID=0.5),^25–27 depression and anxiety measured using the Hospital Anxiety and Depression Scale (HADS) (MID=1.5),²⁸ and HRQoL assessed with the EQ-5D-5L (MID=0.03).²⁹ Since MIDs for these instruments have likewise been established in other clinical populations rather than in PCC, we used their published values to support anchor-based analyses in the absence of PCC-specific thresholds.

Outcomes

Our outcomes were: (1) EQ-VAS values at baseline and follow-up and distribution of EQ-VAS values at baseline, (2) day-to-day within-person variability in EQ-VAS during the 7-day baseline assessment period and (3) change from baseline to follow-up for the estimation of the MID for EQ-VAS.

Statistical analysis

We summarised continuous variables using medians with IQRs and categorical variables were summarised using counts and percentages. Where appropriate, we additionally reported median together with 2.5th and 97.5th percentiles to describe the central 95% of observed values. Weekly EQ-VAS values at baseline and follow-up were calculated as the mean of up to 7 daily ratings, and changes were defined as follow-up minus baseline. Day-to-day variability in EQ-VAS during the baseline week was assessed using within-person SD (wpSD), calculated for each participant as the SD of their available daily EQ-VAS ratings across the 7-day baseline period, and required at least four valid daily measurements.

To determine the MID for the EQ-VAS, we used a combination of distribution-based and anchor-based approaches using triangulation.³⁰ Change (Δ) in scores of EQ-VAS and candidate anchors was defined as follow-up minus baseline. We estimated MIDs using four distribution-based methods^{30 31}: (1) SD-based approach³⁰ (0.5*SD baseline) using weekly mean EQ-VAS at baseline; (2) SE of measurement-based approach (SD baseline*square root (1-ICC)) where intraclass correlation coefficient (ICC) refers to the ICC. The ICC was estimated from a random-intercept mixed model of daily EQ-VAS ratings during the baseline week and converted to the reliability of the weekly mean (7 days) using the residual variance divided by 7. SEM was then computed as SD of the weekly mean EQ-VAS at baseline multiplied by √(1−ICC); (3) Cohen’s effect size (0.5*SDΔ) of Δ EQ-VAS and (4) empirical rule effect size (0.08*6*SDΔ) of Δ EQ-VAS.^{30 32} Anchor-based analyses used candidate anchors expected to improve from baseline to follow-up.^{30 31} Anchors included FACIT-Fatigue score,²¹ CRQ dyspnoea score,³³ HADS total score, HADS anxiety score, HADS depression score³⁴ and EQ-5D-5L utility Dutch index score.⁷ Anchors were evaluated using correlation and receiver operating characteristic (ROC)-based analyses. Anchor-based analysis in this study therefore served two distinct purposes: (1) estimating a group-level MID via the mean-change methods and (2) deriving an individual-level classification threshold (MIC-type estimate) via ROC analysis.^{30 31} These approaches address related but conceptually distinct questions. The former estimates the average change observed among participants reporting minimal but meaningful improvement, whereas the latter identifies the change threshold that best discriminates improved from non-improved individuals.^{30 31} Accordingly, both estimates are reported separately because they address different methodological objectives.

Anchors were considered suitable if the Spearman correlation coefficient between ΔEQ-VAS and the corresponding anchor change met a prespecified threshold of at least 0.30.³⁰ For retained anchors, the anchor-based mean-change MID was estimated as the mean ΔEQ-VAS among participants classified as improved according to the anchor’s established MID (with 95% CIs derived from a one-sample t-test). In ROC-based analyses, improvement was defined using each anchor’s established MID. We quantified discrimination using the area under the ROC curve (AUC) and derived EQ-VAS cut-offs, based on the Youden index, only for anchors with AUC values of at least 0.70. This ROC-derived cut-off represents the EQ-VAS change value that best discriminates participants meeting the anchor-defined improvement criterion from those who did not and was interpreted as a responder/classification threshold (MIC-type estimate) rather than group-based MID.³⁵ This cut-off value was reported descriptively and was not incorporated into the triangulated MID.³⁵

We triangulated MID evidence by prioritising the anchor-based mean-change estimate and comparing these against distribution-based benchmarks. The final proposed MID was chosen as the value that best aligned with both the anchor estimate and the distribution-based range and was further supported by external plausibility checks from published EQ-VAS MIDs in other populations.^{16 28}

Although a small amount of data was missing across outcomes, completeness was high, and we therefore conducted available analyses without imputation.

Patient and public involvement

People with lived experience were involved in the design of the original trial, including providing feedback on the selection of EQ-VAS as a primary outcome measure; they agreed that this instrument captures their health status and global symptom burden relevant to their experience. This input directly informed the choice of outcome measure examined in the current secondary analysis. No additional patient and public involvement activities were conducted specifically for this secondary analysis, as it is based on outcome data already collected within the original trial. We plan to share results with the study participants and other stakeholders (eg, Altea, Long COVID Schweiz).

Results

Between 14 June 2023 and 5 July 2024, 170 individuals were screened for eligibility. Among them, 153 met the eligibility criteria and consented to participate. Of these, 150 participants had valid EQ-VAS data at both baseline and follow-up and were included in the MID analyses (table 1). Table 1 summarises participants’ characteristics. Most participants were female (75.8%), and the mean age was 44.6 years (range 18–80 years). A high proportion (82.9%) of participants were suffering from fatigue (FACIT-Fatigue score <34). The mean EQ-VAS at baseline was very low with a median of 48.7 (37.3–64.1), indicating poor self-reported health status (table 1). Out of 153 participants, 122 (79.7%) completed the EQ-VAS on all 7 days. 24 (15.7%) completed 6 days and 7 (4.6%) completed 4 or 5 days.

Table 1. Participants characteristics at baseline.

Characteristics	All (n=153)
Female, n (%)	116 (75.8)
Age, years	45.0 (34.0–55.0)
Vaccinated against SARS-CoV-2	143 (93.5)
Time since onset of symptoms, weeks	101.3 (74.4–135.4)
Hospitalised for SARS-CoV-2 infection, n (%)	6 (4.0)
Comorbidities, n (%)
Hypertension	12 (7.8)
Diabetes	3 (2.0)
Cardiovascular disease	7 (4.6)
Chronic kidney disease	3 (2.0)
Chronic respiratory disease	22 (14.4)
Psychiatric disease	20 (13.1)
Autoimmune disease	8 (5.2)
Obesity	30 (19.6)
Patient-reported outcomes
FACIT-Fatigue	23.0 (16.0–30.0)
CRQ dyspnoea	6.0 (5.0–6.6)
HADS total score	11.0 (7.0–16.0)
HADS anxiety score	5.0 (3.0–8.0)
HADS depression score	6.0 (4.0–9.0)
EQ-5D-5L index score	0.60 (0.5–0.8)
EQ-VAS	48.7(37.3–64.1)

Data are presented as number (percentages) or median (IQR). For each participant, EQ-VAS values were averaged across the 7-day assessment window (24 valid days required) to obtain aparticipant-level weekly mean. In table 1, EQ-VAS values are reported as the median (IQR) of these participant-level weekly means.

CRQ, Chronic Respiratory Questionnaire; EQ-5D-5L, EuroQol 5-Dimension 5-Level; EQ-VAS, EuroQol Visual Analogue Scale (0–100); FACIT-Fatigue, Functional Assessment of Chronic Illness Therapy-Fatigue; HADS, Hospital, Anxiety and Depression Scale.

Changes in daily and weekly median EQ-VAS values from baseline to follow-up (ie, after 12 weeks) in the Pycnogenol and placebo groups are shown in online supplemental figures 1 and 2. There were no between-group differences in baseline-adjusted EQ-VAS (β=0.54, 95% CI −3.45 to 4.54, p=0.79).

Day-to-day within-person variability

Over the baseline week, the distribution of mean EQ-VAS values across participants centred around a median of 48.7 with the 2.5th and 97.5th percentile range spanning 18.0 to 81.5. The day-to-day within-person variability in EQ-VAS within participants showed a median wpSD of 6.4, with the 2.5th and 97.5th percentile range spanning 1.8 to 16.8 (figure 1).

Figure 1. Distribution of participants’ mean EQ-VAS values (A) and within-person SD (B) based on daily assessments over 7 consecutive days at baseline. EQ-VAS, EuroQol Visual Analogue Scale.

Figure 2 shows daily EQ-VAS trajectories across the 7-day baseline week. Participants were grouped into six bands defined by the wpSD of their daily EQ-VAS ratings during the baseline week. Thin grey lines represent individual participants’ daily EQ-VAS values within each wpSD band. The coloured line with points shows the day-specific median EQ-VAS across participants within each band. In the lowest-variability sextile (≤17th; n=26), most participants showed relatively stable EQ-VAS values across the baseline week, and the daily median remained around ~40 throughout the week. In the higher-variability band, individuals displayed larger day-to-day variability, with some trajectories showing marked short-term fluctuations across a wide range of EQ-VAS values. Furthermore, the maximum observed within-person EQ-VAS range during the 7-day baseline period was 10 points in panel A, 15 points in panel B, 19 points in panel C, 26 points in panel D, 31 points in panel E, and 65 points in panel F (figure 2).

Figure 2. Daily EQ-VAS trajectories during the baseline week, stratified into sextiles of wpSD (A-F). Grey lines represent individual participants’ EQ-VAS values recorded on 7 consecutive baseline days; the blue line shows the daily median within each stratum. Participants were assigned to one of six equally sized groups based on the wpSD of their baseline EQ-VAS values, with panel A (≤17th) indicating the lowest variability and panel F (>83rd–≤100th) indicating the highest variability. EQ-VAS, EuroQol Visual Analogue Scale.

Minimal important difference

For the anchor-based analyses, the number of participants with valid data varied by instrument: 146 participants had complete data for FACIT-Fatigue, HADS and EQ-5D-5L, and 143 participants had complete CRQ data (dyspnoea domain). Table 2 displays change scores from baseline to follow-up, which served as the basis for the anchor-based MID estimation and for distribution-based methods that rely on change-score variability. Distribution-based MID estimates for EQ-VAS ranged from 5.9 to 8.9 points. FACIT-Fatigue was the only anchor meeting the prespecified correlation criterion (Spearman ρ=0.43) and the mean ΔEQ-VAS among FACIT-defined improvers was 8.96 points (95% CI 6.49 to 11.43). Triangulation consisted of comparing the anchor-based estimates with the distribution-based benchmarks. The FACIT-Fatigue anchor-based estimate (8.96 points) closely matched the primary distribution-based benchmark (0.5 SD at baseline: 8.9 points); therefore, we propose an MID of 8 points (table 3). ROC analysis for FACIT-Fatigue showed acceptable discrimination (AUC=0.72) with an optimal EQ-VAS change cut-off of 3.31 points (sensitivity 0.69, specificity 0.67). Other candidate anchors (CRQ dyspnoea, HADS measures, EQ-5D-5L) did not meet the correlation criterion (ρ<0.30).

Table 2. Patient-reported outcomes at follow-up 2 and change from baseline.

Outcome	Follow-up 2	Δ change
FACIT-Fatigue (n=146)	27.0 (22.0 to 33.0)	4 (−1 to 8)
CRQ dyspnoea (n=143)	6.2 (5.0 to 6.8)	0 (−0.2 to 0.4)
HADS total score (n=146)	10.0 (6.0 to 15.0)	−1 (−4 to 2)
HADS anxiety score (n=146)	5.0 (3.0 to 8.0)	−1 (−2.8 to 1)
HADS depression score (n=146)	5.0 (3.0 to 7.0)	−1 (−2 to 1)
EQ-5D-5L index score (n=146)	0.7 (0.6 to 0.8)	0.02 (−0.05 to 0.13)
EQ-VAS (n=150)	56.1 (36.8 to 68.7)	3.3 (−1.9 to 10.8)

Data are presented as median (IQR). For each participant, EQ-VAS values were averaged across the 7-day assessment window (≥4 valid days required) to obtain a participant-level weekly mean. In table 1, EQ-VAS is reported as the median (IQR) of these participant-level weekly means. Higher FACIT-Fatigue values indicate less fatigue; higher CRQ dyspnoea scores indicate less dyspnoea; lower HADS scores indicate less severity; higher EQ-5D-5L and EQ-VAS values indicate higher HRQoL.

CRQ, Chronic Respiratory Questionnaire; EQ-5D-5L, EuroQol 5-Dimension 5-Level; EQ-VAS, EuroQol Visual Analogue Scale (0–100); FACIT-Fatigue, Functional Assessment of Chronic Illness Therapy-Fatigue; HADS, Hospital, Anxiety and Depression Scale.

Table 3. MID estimates for EQ-VAS from anchor-based (A) and distribution-based approaches (B).

A. Distribution-based approaches
Distribution-based approach		Calculation		MID estimate
SD		0.5* SD baseline		8.9
SE of measurement		SD baseline*sqrt (1-ICC)		7.7
Cohen’s effect size		0.5* SDΔ		6.2
Empirical rule effect size		0.08*6*SDΔ		5.9
B. Anchor-based approaches
Anchor-based approach	Correlation coefficient	Used as anchor	Group-based MID mean EQ-VAS (95% CI)	AUC	ROC cut point (responder threshold)*
FACIT-Fatigue	0.43	Yes	8.96 (6.49 to 11.43)	0.72	3.31 (Sens 0.69, Spec 0.67)
CRQ Dyspnoea	0.26	No	NA	0.63	NA
HADS, Total	−0.10	No	NA	0.53	NA
HADS-Anxiety	−0.03	No	NA	0.50	NA
HADS-Depression	−0.14	No	NA	0.52	NA
EQ-5D-5L	0.23	No	NA	0.6	NA
Proposed MID: 8 units

^*ROC derived cut-points are reported descriptively and were not incorporated into the MID, see Methods section.

AUC, area under the curve; CRQ, Chronic Respiratory Questionnaire; EQ-5D-5L, EuroQol 5-Dimension 5-Level; FACIT-Fatigue, Functional Assessment of Chronic Illness Therapy-Fatigue; HADS, Hospital Anxiety and Depression Scale; ICC, intraclass correlation coefficient; MID, minimal important difference; NA, not available; ROC, receiver operating characteristic; Sens, sensitivity; Spec, specificity; VAS, Visual Analogue Scale.

Discussion

Our study provides a detailed description of EQ-VAS levels, their day-to-day variability, and estimation of a MID in a heterogeneous group of adults with PCC. At baseline, participants reported EQ-VAS values around a median of 48.7 with 95% of the values spreading between 18.0 and 81.5, highlighting marked heterogeneity in perceived health status within the cohort. Within participants, day-to-day EQ-VAS variability was also considerable (median wpSD=6.4, 95% range=1.8 to 16.8), indicating pronounced individual fluctuations. Using both distribution-based and anchor-based methods, we generated a triangulated EQ-VAS MID of 8 points, representing the amount of change likely to reflect meaningful improvement beyond normal within-person variation.

The EQ-VAS is a widely used and validated PROM across clinical studies and populations, including PCC.^{5 6 15} We found a low self-reported health status across our heterogeneous PCC population compared with the general population in different European countries, where EQ-VAS values typically range between 71.4 and 82.0, with values in Germany even above 90.^{36 37} This aligns with other analyses showing substantial reductions in HRQoL among individuals with PCC.^38–40 In addition, the pronounced individual day-to-day fluctuations we observed are coherent with existing literature, as considerable variation has been consistently demonstrated across multiple PCC-related symptoms, including fatigue, dyspnoea and cognitive issues.⁴¹ Such fluctuations are further reinforced by evidence that physical, cognitive and social activities trigger short-term symptom worsening, contributing to unpredictable day-to-day variation in perceived health status.¹² These findings suggest that a single-day assessment may be sensitive to ‘good-day’ or ‘bad-day’ effects and therefore do not sufficiently reflect the typical health status among people with PCC. The subgroup analyses by baseline variability also support the use of repeated assessments. The maximum within-person EQ-VAS range observed during the baseline week was 10 points in the lowest variability group and 65 points in the highest variability group, indicating that reliance on a single assessment day may misrepresent typical health status, particularly in more variable individuals. Thus, weekly aggregation of repeated daily ratings can provide more stable estimates of health status and change over time and may improve sensitivity in detecting intervention effects in clinical and research settings.^{41 42}

The MID estimated in this study is consistent with values reported in chronic respiratory diseases such as COPD and aligns with a previously published estimate from a smaller, intervention-specific sample of individuals with PCC.^{16 43} Compared with that earlier work, our study provides a more generalisable benchmark by using a larger sample and scheduled daily EQ-VAS measurements at baseline and follow-up to derive weekly means and characterise within-person variability. The anchor-based mean-change MID was 8.96 points, which fell slightly above the range of our four distribution-based estimates (5.9–8.9 points). We chose an MID value of 8 points as it lies at the upper boundary of the distribution-based range while remaining close to the anchor-based estimate, reflecting the closest alignment between the two approaches; this value was further supported by external plausibility checks against published EQ-VAS MIDs in other populations. This benchmark supports interpretation of change in a condition with high short-term fluctuation: changes smaller than the proposed MID are likely to reflect typical variability, whereas changes meeting or exceeding 8 points are more likely to represent clinically relevant change at the group level.

However, within-person day-to-day variability in EQ-VAS was itself substantial: the median wpSD over the 7-day baseline week was 6.4 points, with the 2.5th–97.5th percentile range spanning 1.8 to 16.8, and maximum within-person ranges reaching up to 65 points in the highest-variability sextile (figure 2). This magnitude of natural fluctuation is comparable to, or in many participants exceeds, the proposed MID of 8 points. This finding reinforces that the MID is intended for interpreting mean differences across a population or between study arms, rather than as a threshold for classifying whether an individual patient has experienced meaningful improvement; applying it at the individual level risks misclassifying natural fluctuation as genuine change, particularly among participants with higher within-person variability. The value of the scheduled, repeated assessment design used in this study lies precisely in mitigating this risk by deriving weekly means rather than relying on single-timepoint measurements.

In addition to supporting clinical interpretation, an empirically derived MID can support the planning of future studies by guiding sample size calculations and defining clinically relevant targets.

Notably, the anchor-based analysis was derived from a single anchor assessing fatigue (FACIT-Fatigue). Fatigue is a frequent symptom in PCC^{15 44 45} and was highly prevalent in our study population with 83% of our participants having a FACIT-Fatigue score <34, indicating clinically relevant fatigue.²² Although direct comparisons with other studies are challenging due to differences in study populations, PCC definitions, and duration of persistent symptoms, fatigue prevalence in our cohort, who experienced persistent symptoms for almost 2 years, was relatively high compared with previously reported estimates in PCC populations.^{15 44 45} Since PCC is characterised by a heterogeneous symptom profile, including post-exertional malaise, dyspnoea, cognitive impairment, and psychological symptoms, which may independently affect overall health status, our MID estimate may not be universally applicable to all individuals with PCC. In contrast, our anchor-based MID may better reflect meaningful change for individuals whose health status is predominantly driven by fatigue and may underestimate or overestimate meaningful change for those whose functional impairment is primarily driven by other symptom clusters. This should be considered when applying our MID estimate to PCC populations or subgroups with a different symptom profile.

Taken together, the proposed MID of 8 points is best interpreted as a group-level benchmark for meaningful change in EQ-VAS among PCC populations with a chronic, predominantly fatigue-driven symptom profile similar to our cohort. Given the substantial short-term fluctuation in EQ-VAS and reliance on a single anchor, caution is warranted when applying this threshold to populations assessed earlier in the disease course, or to those whose functional impairment is primarily driven by symptoms other than fatigue.

A major strength of this study is the scheduled daily assessment design. By collecting EQ-VAS ratings on 7 predefined consecutive days before baseline and follow-up, we were able to minimise self-selection in response timing (‘good day’ bias) and reduce the influence of day-specific fluctuations on the measurement of overall health status. This design ensures that weekly mean values more accurately reflect typical health status rather than isolated good or bad days.⁴⁶ Another strength is the heterogeneity of our sample. We included participants with a broad range of disease severity, from individuals who were only mildly affected to those who were severely affected and unable to attend the study centre. To ensure their participation, we offered home visits, which increased inclusiveness and improved the representativeness of the cohort. In addition, people with lived experience were actively involved in the preparation of the study design, including feedback on primary outcome selection (EQ-VAS). They expressed that the EQ-VAS captures and reflects their symptom burden, supporting the relevance and acceptability of the chosen PROMs. Finally, the use of complementary anchor-based and distribution-based methods, combined with prespecified performance checks, represents a methodological strength, although the anchor-based component was ultimately derived from a single retained anchor (see Limitations). The use of multiple complementary approaches provides a more robust and patient-relevant estimate of meaningful change in EQ-VAS values.³⁰

This study has several limitations. First, our study population was recruited from a single-centre, randomised, placebo-controlled trial investigating the effects of Pycnogenol versus placebo on EQ-VAS in adults with PCC. Eligibility criteria required the presence of at least one of the four self-reported symptoms (postexertional malaise, fatigue, breathing difficulties or brain fog), with a confirmed SARS-CoV-2 infection or physician diagnosis. Exclusion was limited to safety considerations rather than symptoms. Accordingly, our cohort spanned a wide range of disease severity, from participants with minimal symptoms to those unable to leave their home. Nonetheless, baseline health status was comparatively low (median EQ-VAS 48.7) and the average duration of persistent symptoms in our cohort was almost 2 years, both of which may reflect self-selection into a treatment-seeking trial population. Both factors may limit applicability relative to broader, unselected PCC populations, including individuals earlier in the disease course or with a short duration of symptoms. The female predominance is consistent with the known epidemiology of PCC, where female sex is a recognised risk factor,^{47 48} and is therefore unlikely to substantially limit generalisability. Second, only FACIT-Fatigue met the prespecified correlation threshold for use as an anchor in the analyses. Consequently, our triangulation approach therefore relied on a single retained anchor rather than multiple independent anchors; as such, the resulting MID reflects agreement between one anchor-based estimate and a distribution-based range, rather than convergence across independently derived anchor-based estimates, which may otherwise strengthen confidence in the final value. It also means that the anchor-based component may not fully capture meaningful change related to other potentially relevant PCC constructs, such as post-exertional malaise, dyspnoea or psychological symptoms. Third, this is a secondary analysis of pooled data from a randomised placebo-controlled trial comparing the effects of Pycnogenol versus placebo on change in EQ-VAS over 12 weeks. Overall, both groups showed improvements in health status over time (mean change in the Pycnogenol and placebo groups was 5.4 units and 7.9 units, respectively); however, there was no evidence of a difference in change between the groups (β=0.54, 95% CI −3.45 to 4.54, p=0.79). We, therefore, do not believe that pooling the treatment arms meaningfully biased our MID estimates. Fourth, a larger sample would have allowed for greater precision of anchor-based estimates and enabled subgroup analyses (eg, by symptom phenotype), which were beyond the scope of the current analysis. Finally, the daily EQ-VAS data were collected over a 7-day window at baseline and follow-up, which captures short-term variability but may not reflect longer-term cyclical symptom patterns commonly observed in PCC.

Conclusions

Adults with PCC report low health status on the EQ-VAS, and EQ-VAS values demonstrated considerable day-to-day variability, underscoring the need for repeated measurements rather than single-day assessments. Using daily EQ-VAS data and a combination of anchor-based and distribution-based approaches, we estimated an EQ-VAS MID of 8 points, representing change that exceeds normal day-to-day variability and can be interpreted as meaningful improvement at the group level. The MID was derived from a chronic, fatigue-predominant PCC population and offers a practical benchmark for clinical interpretation, supporting more robust outcome definition in future research. Its applicability to earlier-stage or non-fatigue-predominant PCC populations warrants further evaluation. Implementing repeated assessments and MID-based thresholds may strengthen both routine clinical monitoring and the methodological quality of intervention studies in PCC.

Data availability statement

Data are available on reasonable request.