Table 3

Results of sensitivity analysis

OutcomeEstimates of treatment effects from different scenarios with 95% CIs
ITT, primary analysis
(n=119, nint=59, nobs=336)
Per protocol‡
(n=91, nint=36)
ITT, extended MAR multiple imputation*
(n=119, nint=59, 70 sets)
ITT, CIR multiple imputation†
(n=119, nint=59, 50 sets)
ITT, model including time from symptoms onset§
(n=119, nint=59, nobs=336)
Primary outcome: 6MWD (m)
 Post-treatment (6 weeks)65.45 (43.80 to 87.10)72.25 (47.54 to 96.97)62.23 (40.07 to 84.39)57.18 (35.42 to 78.95)65.12 (44.50 to 85.74)
 Follow-up (~28 weeks)68.62 (46.39 to 90.85)75.92 (51.21 to 100.64)61.99 (39.22 to 84.76)63.07 (40.87 to 85.27)67.99 (46.77 to 89.20)
Secondary outcomes
Squat time (s)
 Post-treatment (6 weeks)20.12 (12.34 to 27.90)22.67 (13.91 to 31.43)20.32 (11.72 to 28.91)17.81 (10.01 to 25.61)20.15 (12.44 to 27.86)
 Follow-up (~28 weeks)22.23 (14.24 to 30.21)25.94 (17.18 to 34.70)21.48 (12.73 to 30.24)20.07 (12.09 to 28.06)22.38 (14.45 to 30.31)
Pulmonary function
FEV1 (L)
 Post-treatment (6 weeks)0.08 (−0.08 to 0.25)0.09 (−0.1 to 0.28)0.07 (−0.11 to 0.25)0.06 (−0.11 to 0.24)0.08 (−0.08 to 0.25)
 Follow-up (~28 weeks)0.00 (−0.18 to 0.17)−0.03 (−0.22 to 0.16)−0.05 (−0.23 to 0.13)0.00 (−0.18 to 0.18)0.00 (−0.18 to 0.17)
FVC (L)
 Post-treatment (6 weeks)0.02 (−0.14 to 0.18)0.09 (−0.08 to 0.27)−0.01 (−0.18 to 0.16)0.03 (−0.13 to 0.20)0.02 (−0.14 to 0.18)
 Follow-up (~28 weeks)0.01 (−0.16 to 0.17)0.07 (−0.11 to 0.25)−0.06 (−0.23 to 0.12)0.01 (−0.16 to 0.18)0.01 (−0.15 to 0.17)
FEV1/FVC
 Post-treatment (6 weeks)0.03 (−0.02 to 0.07)0.02 (−0.02 to 0.07)0.02 (−0.02 to 0.06)0.02 (−0.02 to 0.06)0.02 (−0.02 to 0.06)
 Follow-up (~28 weeks)−0.01 (−0.05 to 0.03)−0.02 (−0.07 to 0.03)−0.01 (−0.05 to 0.03)0.00 (−0.04 to 0.04)−0.01 (−0.05 to 0.03)
MVV (L/min)
 Post-treatment (6 weeks)10.57 (3.26 to 17.88)14.3 (6.1 to 22.5)10.09 (2.11 to 18.07)10.32 (2.91 to 17.73)10.57 (3.3 to 17.85)
 Follow-up (~28 weeks)5.20 (−2.33 to 12.73)7.29 (−1.00 to 15.59)3.04 (−5.38 to 11.46)6.01 (−1.77 to 13.78)5.25 (−2.27 to 12.76)
PEF (L/s)
 Post-treatment (6 weeks)0.38 (−0.24 to 1.00)0.52 (−0.17 to 1.22)0.35 (−0.29 to 0.99)0.35 (−0.29 to 0.99)0.38 (−0.24 to 1.00)
 Follow-up (~28 weeks)−0.02 (−0.66 to 0.62)−0.16 (−0.86 to 0.55)−0.18 (−0.83 to 0.48)0.03 (−0.62 to 0.67)−0.03 (−0.67 to 0.61)
Quality of life
SF-12 PCS
 Post-treatment (6 weeks)3.79 (1.24 to 6.35)3.70 (0.76 to 6.63)3.68 (1.13 to 6.24)3.27 (0.69 to 5.86)3.75 (1.22 to 6.27)
 Follow-up (~28 weeks)2.69 (0.06 to 5.32)2.37 (−0.57 to 5.30)2.31 (−0.43 to 5.05)2.44 (−0.28 to 5.16)2.72 (0.12 to 5.33)
SF-12 MCS
 Post-treatment (6 weeks)2.18 (−0.54 to 4.90)1.92 (−1.14 to 4.97)2.17 (−0.57 to 4.91)1.65 (−1.07 to 4.38)2.18 (−0.53 to 4.89)
 Follow-up (~28 weeks)1.99 (−0.81 to 4.79)1.48 (−1.58 to 4.54)2.30 (−0.48 to 5.09)1.82 (−0.96 to 4.61)1.93 (−0.87 to 4.73)
mMRC-dyspnoea, to favourable outcome
 Interim 2 weeks1.27 (0.88 to 1.82)1.33 (0.90 to 1.98)1.26 (0.87 to 1.81)1.26 (0.88 to 1.82)1.28 (0.89 to 1.85)
 Interim 4 weeks1.08 (0.82 to 1.42)1.04 (0.76 to 1.42)1.06 (0.80 to 1.4)1.06 (0.80 to 1.40)1.09 (0.82 to 1.43)
 Post-treatment (6 weeks)1.46 (1.17 to 1.82)1.43 (1.14 to 1.80)1.42 (1.13 to 1.79)1.40 (1.11 to 1.77)1.47 (1.17 to 1.84)
 Follow-up (~28 weeks)1.22 (0.92 to 1.61)1.15 (0.84 to 1.57)1.21 (0.92 to 1.58)1.23 (0.92 to 1.63)1.23 (0.93 to 1.62)
  • Apart from estimates for mMRC perceived dyspnoea (favourable outcome), estimates are between-group differences in mean change from baseline derived from linear mixed effects models adjusted for study centre (fixed effect) and with baseline means constrained to be equal across comparison groups. Estimates for mMRC perceived dyspnoea are rate ratios derived from a general linear mixed model of the Poisson family with log link adjusted for centre and ln(number valid observations up to data point) as offset. CIs are estimated with cluster robust standard errors (cluster variable: participant ID).

  • *Based on multiple imputation using chained equations assuming data were missing at random. The imputation model included all outcomes and the following auxiliary variables with complete baseline information: gender, age, smoking history (no vs yes), presence of any comorbidity (no vs yes), COVID-19 severity (non-severe vs severe), body mass index, time from admission to hospital to baseline assessment in days.

  • †Used reference-based controlled multivariate normal imputation for each outcome assuming increments in both comparison groups followed the observed pattern of the control group where data points were not observed. Auxiliary variables employed were as in the MAR model. After imputation adaptive rounding was used for imputed values of mMRC-dyspnoea (favourable outcome).

  • ‡Based on per-protocol sample, that is, participants who completed all assessments and completed the intervention programme at or above minimum intensity and duration required for compliance as defined in the protocol (intervention group).

  • §Same as main analysis but providing additional adjustment for time from onset of symptoms to measurement point in days and time from onset of symptoms to measurement point in days squared. A likelihood ratio test confirmed superior fit of the model that also included the squared term.

  • CI, confidence interval; CIR, copy increments from reference; ITT, intention to treat; MAR, missing at random; MCS, mental component score; mMRC, modified Medical Research Council; MVV, maximum voluntary ventilation; 6MWD, six min walking distance; PCS, physical component score; PEF, peak expiratory flow; SF-12, Short Form Health Survey-12.