Abstract
BACKGROUND: Pulmonary rehabilitation (PR) is useful in survivors of COVID-19-associated acute respiratory failure (ARF). The aim of this retrospective study on in-patient PR was to report rehabilitative trajectories and effects of cycle training.
METHODS: According to the Short Physical Performance Battery (SPPB) score at admission (T0), participants were allocated to stage 1 (SPPB < 6), stage 2 (SPPB ≥ 6 and < 10), or stage 3 (SPPB ≥ 10) and performed increasing level of activities from passive exercises to free walking, balance exercises, strength exercises, and tailored cycle-ergometer endurance training. The primary outcome was SPPB. 6-min walk distance (6MWD), Medical Research Council score, Barthel dyspnea index, and rate of subjects able to cycling were also assessed.
RESULTS: Data of 123 participants were analyzed. At T0, 44 (35.8%), 50 (40.6%), and 29 (23.6%) participants were allocated to stages 1–3, respectively. At discharge, participants showed significant improvements in SPPB, independent of the initial stage, 81 (65.8%) improving more than its minimal clinically important difference. At T1, the proportion of participants in stages 1 and 2 decreased, whereas significantly increased in stage 3 (P = .003), (being 9.8%, 33.3%, and 56.9% for stages 1–3, respectively; P <.001). Sixty-nine of 123 participants (56.1%) underwent cycle exercise training. In participants able to perform it, 6MWD improved by 115 (65–240) m and 60 (40–118) m in participants with and without exercise-induced desaturation, respectively, with significant difference between groups (P = .044).
CONCLUSIONS: In-patient PR could be tailored and progressively increased to survivors of COVID-19-associated ARF; cycle training was feasible in half of the participants. Benefits were independent of initial stage of physical performance and allowed participants to move from lower to higher levels of activities.
Introduction
COVID-19 manifests a variety of clinical presentations ranging from no symptoms or light flu to pneumonia and acute respiratory failure (ARF) requiring admission to the ICU and possible death.1,2 In addition to the consequences on lung function,3 a high prevalence of impairment in physical performance is reported in survivors, who may suffer from fatigue and/or muscle weakness, exercise-induced dyspnea, sleep difficulties, and anxiety and/or depression up to 6 months after infection.4-7
Pulmonary rehabilitation (PR) including exercise training is a recognized cornerstone of comprehensive management of chronic respiratory diseases. In different settings, it improves symptoms such as dyspnea and fatigue, exercise capacity, and health-related quality of life.8,9 Clinical indications and modalities of PR have been proposed by international guidelines and recommendations,10,11 and uncontrolled studies have shown that this modality is possible and effective in survivors of COVID-19, including those requiring assisted ventilation or oxygen therapy.12,13 However, knowledge of the potential effects of programs tailored to the individual physiological characteristics is scarce; and the sometime safety, tolerability, and effects of moderately intense continuous training in people with ongoing symptomatic COVID-19 are totally unclear.
Therefore, this retrospective real-life study aimed to report in-patient rehabilitative trajectories and effects of cycle training programs in survivors of COVID-19-associated ARF.
QUICK LOOK
Current Knowledge
Pulmonary rehabilitation (PR) is a recognized cornerstone of comprehensive management of chronic respiratory diseases. Knowledge of the safety, tolerability, intensity, and effects of PR (including exercise training in people with ongoing symptomatic COVID-19) is unclear.
What This Paper Contributes to Our Knowledge
In subjects with ongoing symptomatic COVID-19, PR could be tailored and progressively increased. Cycle training was feasible in half of the participants. Benefits were independent of initial stage of physical performance and allowed participants to move from lower to higher levels of activities.
Methods
The study was approved by the Istituti Clinici Scientifici (ICS) Maugeri, IRCCS Ethics Committee (EC 2279; March 12, 2020). As a retrospective study, individuals did not provide any specific written informed consent; however, at admission (T0), they gave in advance their informed consent for the scientific use of their clinical data. As a retrospective analysis, the study was not registered.
The study was conducted on the records of survivors of COVID-19-associated pneumonia and ARF, with negative reverse transcription-polymerase chain reaction (RT-PCR) test for SARS-CoV-2, consecutively admitted, between April 1–August 15, 2020, to ICS Maugeri IRCCS Hospital of Lumezzane (Brescia), Italy, a referral institution for rehabilitation, diagnosis, and care of post-acute and chronic conditions.14,15 Participants had been transferred from acute care hospitals where they might have experienced invasive or noninvasive ventilation (NIV), high-flow oxygen therapy, or oxygen supplementation alone.16 Data were excluded from individuals who had a persistently positive RT-PCR test, were newly admitted to an acute care hospital, had a length of stay (LOS) in our institution < 6 d, or died.
The results of PR of some participants reported in the present study had been included in a previous multi-centric study.12
During their stay in our institution, participants underwent a progressive reduction in oxygen supplementation and weaning from NIV (when necessary) in addition to therapy for their comorbidities.14,16 A consensus audit among senior respiratory therapists (RTs) and pulmonologists used tailored programs according to the modified Italian position paper.11 Using safety procedures and wearing appropriate personal protective equipment,17 at T0 the RT evaluated the subjects’ physical function by the Short Physical Performance Battery (SPPB) test.18,19 The SPPB total score results from the sum of 3 components: standing balance, 4-m walk test, and 5 repetitions standing from sitting position. The total SPPB score ranges from 0–12, with 0 corresponding to inability to perform and 12 to normal. According to their SPPB total score at T0, participants were included in stage 1 (SPPB < 6), stage 2 (SPPB ≥ 6 and < 10), or stage 3 (SPPB ≥ 10) to perform programs of increasing activities.
Changes between one stage to another were possible thanks to the application of a protocol-guided and follow-ed by an RT-based on a common battery of evaluations and including also exercise training to tailor to the improved ability of survivors of COVID-19-associated ARF during the rehabilitation program.
Stage 1. An RT supervised a single participant (ie, RT/participant ratio [RT/P] 1:1) to perform, once a day, one or more of the following activities: passive mobilization, active exercises, free walking, peripheral limb muscle activities, and shoulder and full arm circling.
Stage 2. With an RT/P 1:4–5, once a day participants underwent a group session of exercises including calisthenics, balance exercises, and paced walking. All exercises were performed either without devices or using tools such as balls, canes, balance boards, or light weight bands.
Stage 3. The RT supervised a group of 4–5 participants undergoing 2 daily sessions of cycle-ergometer endurance training under continuous heart rate, arterial blood pressure, and SpO2 monitoring. Each session consisted of 40-min cycling (5-min warm-up, 30 min at a constant load, 5-min cooldown without load). The constant workload started from 20 W, and at the end of each session, participants were assessed to prescribe the workload intensity for the next session according to the schedule shown in Figure 1. In brief, when participant’s heart rate was < 85% of maximal predicted20 or perceived dyspnea and/or leg discomfort was ≤ 3 on a Borg scale,21 the workload was increased by 10 W; if participants’ heart rate was > 85% of maximal predicted or they had dyspnea and/or muscle fatigue ≥ 7 on the Borg scale, the workload was decreased by 10 W. When the Borg scale was scored between 4–6, the workload was unchanged. When necessary, oxygen supplementation was administered to maintain SpO2 > 92% during the training sessions.
Pathways of exercise training intensity. Panels A-C refer to prescription criteria of workload according to dyspnea, leg fatigue, and heart rate, respectively.
The following data were recorded at T0: demographics, anthropometrics, number and diagnosis of comorbidities by the Cumulative Illness Rating Scale,22 arterial blood gases, and FIO2 needed at rest and during exercise. The FIO2 was estimated from the oxygen flow according to the formula FIO2 = 0.20 + (4 x oxygen liter flow).23
The PaO2/FIO2 was also recorded. LOS and use of mechanical ventilation, either invasive or NIV in referring hospitals, were recorded as well.
Within the first 48 h since T0, when allowed by the participant’s clinical conditions and level of collaboration, dynamic lung volumes were assessed according to standards24 using the predicted values of Quanjer.25 Moreover, the following outcome measures were also assessed within the first 48 h from T0 and at discharge (T1):
Physical function was evaluated by SPPB18-19 (see above) as primary outcome. One point was reported as the minimal clinically important difference for SPPB.26
Exercise tolerance was assessed by the 6-min walk test (6MWT)27 under continuous heart rate and SpO2 monitoring (Vyntus WALK, Vyaire Medical, Mettawa, Illinois) using the predicted values of Enright and Sherrill.28 For participants unable to perform the 6MWT, a 0 m value was computed in the analysis. Only participants in stage 3 were divided according to the presence of exercise-induced desaturation (EID) defined as baseline SpO2 − mean SpO2 during the 6MWT > 4%. The SpO2/FIO2 was also monitored before and during the test and recorded.
Dyspnea was assessed using the Medical Research Council (MRC) dyspnea scale,29 on a 0–5 scale, with 0 corresponding to the absence of dyspnea and 5 to the most severe level of dyspnea, as well as using the Italian version of the Barthel dyspnea index.30 The total score ranged from 0 (no)–100 (the most severe level of dyspnea). A decrease in score represented an improvement, whereas an increase represented a worsening in symptoms. The test-retest and inter-observer reliability and the concurrent validity of the Barthel dyspnea index have recently been proposed,30 and the minimal clinically important difference has been defined as −9 points for patients with COPD without chronic respiratory failure and −12 points for patients with COPD and chronic respiratory failure.31
Statistical analysis was performed with statistical software Stata 13 (StataCorp, College Station, Texas). Variables were described as mean ± SD, median (first–third quartile) for continuous variables, and numbers and frequency distribution for categorical and ordinal ones. Normality of distribution for continuous physiological variables was assessed with a Shapiro-Wilk test. The differences in continuous outcome measures between hospital T0 and T1 were evaluated by paired t test or by Wilcoxon matched-pairs signed-rank test when appropriated. For binary and categorical variable, a chi-square test was used. Differences between groups were evaluated by a one-way ANOVA test or Kruskal-Wallis test. If significant, a post hoc analysis was delivered by pairwise t test evaluation using Bonferroni correction. In case of nonparametric analysis, we performed a multiple comparison using Dunn test; and a 2-stage linear step-up correction procedure of Benjamin, Krieger, and Yekutieli was applied. Time course in workload, oxygenation, heart rate, and symptoms during training sessions was analyzed by 2-way mixed-effect ANOVA with time and group as factors. A correlation evaluation by Pearson test was performed to assess factors associated with SPPB changes. All statistical tests were considered significant when P < .05.
Results
Figure 2 shows the flow chart trial profile. Of 161 participants admitted, 123 fulfilled inclusion criteria. Data were excluded from participants for the following reasons: persistently positive RT-PCR (n = 10), died (n = 5), transferred to an acute care hospital (n = 10), and LOS < 6 d in our institution (n = 13).
Flow chart. RT-PCR = reverse transcription-polymerase chain reaction.
Table 1 shows the characteristics of the included participants. At T0, 44 (35.8%), 50 (40.6%), and 29 (23.6%) participants were included into stages 1−3, respectively. Participants initially included in stage 3 were significantly younger than those included in other stages and had fewer comorbidities and levels of dyspnea severity than in stage 1. However, even in the most severe stages, participants showed mild levels of dyspnea as assessed by Barthel dyspnea index. Participants showed mean body mass index (BMI) in the overweight range, with 25% showing a BMI > 30 kg/m2, without significant differences among stages. The 6-min walk distance (6MWD) increased significantly with stages. No significant differences among stages were observed in LOS, use of oxygen, or NIV in acute care hospitals or lung function, whereas participants included in stage 3 spent significantly fewer days in our institution than those included in the other stages.
Anthropometric, Demographic, Physiological, and Clinical Characteristics of Participants at Admission
As shown in Figure 3, panel A, at T1 the proportion of participants in stages 1 and 2 decreased, whereas significantly (P = .003) increased in stage 3. At T1, the rate of subjects was 9.8%, 33.3%, and 56.9% for stages 1–3, respectively, P < .001. Figure 3, panel B shows the proportion of participants moving from one stage to another or remaining in the initial stage.
Trajectories of rehabilitation programs. A: Proportion of participants in different stages at T0 and T1; B: Proportion of participants moving from one stage to another (oblique dashed lines) or remaining in the initial stage (horizontal lines). T0 = hospital admission; T1 = hospital discharge. * P = .003.
As shown in Figure 4, at T1 participants showed significant improvements in SPPB, independent of the initial stage. Eighty-one (65.8%) participants improved their SPPB more than the minimal clinically important difference. However, only participants of stage 3 reached mean SPPB values in the normal range ≥ 10. The improvement in SPPB was inversely and slightly correlated with PaO2/FIO2 (r = −0.3108, P =.03), baseline SPPB score (r = −0.5265, P =.003), and baseline 6MWD (r = −0.2700, P < .001). In 57 participants with acute hospital LOS ≤ 30 d, the mean SPPB score improved from 6.0 (2.0–9.0) to 10.0 (7.0–11.0) (P < .001), whereas in 66 participants with LOS > 30 d it improved from 4.0 (1.0–8.0) to 9 (6.0–11.0) (P < .001), without any significant between-group difference (P = .69).
Changes in the Short Physical Performance Battery (SPPB) (admission-discharge) for the whole sample and according to initial stages. A: All subjects, B: stage 1, C: stage 2, D: stage 3. Boxes extend from the 25th to 75th percentiles; bars denote the median value, and the whiskers extend from the 10th to the 90th percentiles.* P < .001.
In the whole group, the 6MWD significantly improved from 165 (0–290) m to 300 (210–390) m (P < .001). There was no significant difference in improvement in 6MWD according to LOS in acute care hospitals: 190 (0–290) m to 305 (230–390) m (from 28.2% predicted [0–43.6] to 44.6% predicted [33.5–54.4]; P < .001) and from 100 (0–270) m to 280 (180–375) m (16.4% predicted [0–39.2] to 44.3% predicted [28.0–52.7]; P < .001) in participants with LOS ≤ and > 30 days, respectively, (P = .68 for between-group differences).
The dyspnea scores significantly improved at T1 (MRC: from 3.0 [2.0–4.0] to 2.0 [1.0–3.0], P < .001; twice a day, from 22.0 [10.0–29.0] to 8.0 [3.0–18.0], respectively; P < .001).
Sixty-nine of 123 participants (56.1%) underwent cycle exercise training. Fifty (30 and 20 with and without EID, respectively) of 69 participants performed at least 12 consecutive training sessions, and their data were analyzed. At the end of the program, 6MWD increased by 115 (65–240) m and 60 (40–118) m in participants with and without EID, respectively, with a significant between-group difference (P = .044).
The time course of session workload, performed by these participants according to the presence of EID, is shown in Figure 5. Training intensity increased in both groups from sessions 1–7 or 8 and plateaued after the eighth session at 39.3 ± 13.4 W and 44.3 ± 17.0 W in participants with and without EID, respectively, with a 98.9 ± 112.0% and 145.2 ± 107.0% increase from the start of training, respectively, without any significant between-group difference (P = .15). Figure 5 shows also the time course of physiological parameters and symptoms assessed at the end of each session. There was no significant difference between participants with and without EID, except for the expected differences in SpO2/FIO2.
Time course of training workload and physiological variables assessed at the end of each session. Data as mean ± SD. A: Workload; B: Fatigue; C: Heart Rate (% predicted); D: SpO2/FIO2; E: Dyspnea; and F: Mean Blood Pressure. EID = exercise-induced desaturation. HR= heart rate. P = ANOVA analyses: t x g = interaction time × groups; t = time factor, g = group factor. Grey circle = participants with EID; Black circle = participants without EID.
Discussion
This real-life study shows that a protocol including exercise training based on a common battery of evaluations and tailored to survivors of COVID-19-associated ARF was feasible and effective. Benefits were independent of the physiological and clinical conditions at T0 and allowed participants to move from lower to higher levels of activities.
The first interesting result of our study is that survivors of COVID-19-associated ARF showed a wide range of physical performance as assessed by means of the SPPB. In a previous study,4 we also observed a high prevalence of muscle weakness and physical performance impairment in individuals without any previous motor limitation, recovering from less severe COVID-19 pneumonia not requiring prolonged ICU stay, and/or mechanical ventilation, with LOS in acute care hospitals far less than participants of this study (10 vs 40 d).
What is the cause of the limitation of physical performance in these individuals? It may be argued that immobilization due to LOS in acute care hospitals may have influenced this result. However, in our participants, the LOS in acute care hospitals was not different among the stages and was not significantly correlated to the outcome measure such as SPPB. At the same time according to our results, limitation in physical performance cannot be ascribed to the need for supported ventilation (either invasive or NIV) as we did not find any difference in the use in acute care hospitals of such tools among participants of different initial stages. It can be also argued that physical performance may have been influenced by post-infection lung function. A systematic review3 showed that survivors of COVID-19 may suffer from impaired lung function, with diffusion capacity being the most affected. We did not find any significant difference in dynamic lung volumes among participants included in initial stages who showed a common mild restrictive pattern (mean FVC 70% predicted in all stages), and unfortunately, we were unable to assess diffusion capacity. Therefore, we cannot draw conclusions from these observations.
Despite treatment recommendations by various professional and international scientific societies,10,11,32 the early phase of the pandemic was characterized by continuous organizational changes requiring program adaptations, often more empiric than evidence based. On the other hand, we could not wait for well-designed prospective randomized controlled trials to be published before starting these interventions in daily clinical practice. Our program was provided according to the Italian position paper,11 with appropriate modifications to tailor type, intensity, timing, and modality of intervention to the individual characteristics. The use of a selection procedure based on SPPB allowed profitable and therapeutic trajectories of PR to also be used in the post-acute phase of the disease.33,34
The program resulted in improvements in assessed outcome measures, namely SPPB and 6MWT, independent of the initial stage of inclusion. However, only stage 3 participants reached post-program mean values of 6MWD > 40% predicted, indicating the need for strict follow-up and post-T1 programs for a full recovery.35 In the study by Huang et al,6 in most individuals who recovered from severe COVID-19, dyspnea scores and exercise capacity improved over time; but in that study,6 individuals who required intubation and mechanical ventilation were excluded given the potential for the consequences of mechanical ventilation itself to influence the factors under investigation.
In our study, 60% of participants undergoing exercise training showed EID as compared to 43% and 27% reported in previous studies.36,37 Such differences may be explained by differences in participants’ selection and the definition of EID. As also previously reported in individuals with stable COPDs,38 the intensity of training of our participants undergoing exercise training increased progressively from sessions 1–7 or 8 and plateaued after the eighth session with a relevant increase as compared to the start of training, without any significant differen-ce between participants with or without EID. The time course of the assessed physiological parameters and symptoms did not differ either. These results suggest that there is an indication for exercise training in these individuals independent of the presence of EID.
Both tools used to assess perceived dyspnea, namely MRC and Barthel dyspnea index, were able to discriminate among stages and paralleled the stage severity, confirming previous reports in other diseases.29,30 However, even in the most severe stages, participants showed mild levels of dyspnea. The sensation of dyspnea is multifactorial, including psychological factors, and in different chronic respiratory diseases is based on different pathophysiological abnormalities with different qualities of respiratory discomfort, as defined by specific verbal descriptors. In addition, as shown in Table 1, there was no significant difference in lung function among groups, indicating that the severity of the symptom cannot be predicted from lung function; therefore, dyspnea must be measured specifically. Several instruments are commonly used to measure different domains of dyspnea such as sensory-perceptual experience, affective distress, and symptom impact or burden. In individuals with chronic respiratory diseases, the twice a day based on a modified Barthel index was shown to be a reliable, sensitive, and adequate tool for measuring the level of dyspnea perceived in performing activities of daily life. It has been shown to be responsive to PR in individuals with COPD with and without chronic respiratory failure. Other tools such as those assessing health status or activities of daily life often include items related to symptoms but are not specific.39
The mean BMI of our participants was in the overweight range, being > 30 kg/m2 in 25% of the subjects, independent of the initial stage allocation. A BMI within the overweight (25–30 kg/m2) or obese (> 30 kg/m2) categories and cardiometabolic morbidity have been associated with a more serious illness needing hospitalization and invasive ventilation.40
Only subjects with negative real-time PCR test for SARS-CoV-2 were included in our study. In absence of such evaluation, there is no consensus on how long patients should be self-isolating. It has been suggested that local infection prevention recommendations should be followed with significant adaptation of the program, eventually with the adoption of tele-rehabilitation.41
This was a retrospective single-center study with the flaws of this type of study, like missing data. The reported comorbidities occurrence in our study must not be considered as a real prevalence as our subjects did not undergo any specific diagnostic test. Furthermore, we did not assess the impact of the disease on health-related quality of life. At T1, participants showed improvements in outcome measures. The results of an uncontrolled study may be difficult to interpret because we can suppose a positive effect in the long-term follow-up of these individuals without any rehabilitative intervention. A control population not performing any activity would be unethical given the undisputed benefits of PR or simple physical activity and the results of preliminary studies.12 However, our results of participants moving from a more severe stage of physiological compromise performing lower levels of activities to stages of more complex training may contribute to the dilemma solution.
This protocol is based on a thorough assessment of baseline physical performance and of physiological data and symptoms on exercise. It may be useful for the assessment and prescription of training as part of the approach for personalized management of survivors of COVID-19-associated ARF. Protocols driving the program prescription may be included in the routine practice to administer mobilization and exercise, ensuring adherence to the guidelines, and customizing the treatment.
Conclusions
Based on T0 evaluations, our protocol to tailor in-patient rehabilitation programs to survivors of COVID-19-associated ARF was feasible and effective, allowing cycle training in half of the participants. Our findings may be useful to guide clinicians on how to take care of these individuals. Future studies are needed to demonstrate whether our protocol may play a role in achieving the aims of health care including better outcomes, lower cost of care, and improved experience for patients and staff.
Acknowledgments
The authors thanks all physiotherapists of ICS Maugeri IRCCS, Respiratory Rehabilitation Division who treated the subjects enrolled in this study.
Footnotes
- Correspondence: Michele Vitacca MD, Respiratory Rehabilitation Department, Istituti Clinici Scientifici Maugeri IRCCS, Via Salvatore Maugeri 4, 27100 Pavia, Italy. E-mail: michele.vitacca{at}icsmaugeri.it
The authors have disclosed no conflicts of interest.
This work was supported by the “Ricerca Corrente” funding scheme of the Italian Ministry of Health.
- Copyright © 2022 by Daedalus Enterprises
