Abstract
Rationale To describe cardiopulmonary function during exercise 12 months after hospital discharge for COVID-19, assess the change from 3 to 12 months, and compare the results with matched controls without COVID-19.
Methods In this prospective, longitudinal, multicentre cohort study, hospitalized COVID-19 patients were examined with a cardiopulmonary exercise test (CPET) 3 and 12 months after discharge. At 3 months 180 performed a successful CPET, and 177 at 12 months (mean age 59.3 years, 85 females). The COVID-19 patients were compared with controls without COVID-19 matched for age, sex, body mass index, and comorbidity. Main outcome was peak oxygen uptake (V′O2peak).
Results Exercise intolerance (V′O2peak <80% predicted) was observed in 23% at 12 months, related to circulatory (28%), ventilatory (17%), and other limitations including deconditioning, and dysfunctional breathing (55%). Estimated mean difference between 3 and 12 months showed significant increases in V′O2peak % predicted (5.0 percent points (pp), 95% CI (3.1 to 6.9), p<0.001), V′O2peak·kg−1% predicted (3.4 pp, (1.6 to 5.1), p<0.001), and oxygen pulse % predicted (4.6 pp, (2.5 to 6.8), p<0.001). V′O2peak was 2440 mL min−1 in COVID-19 patients compared to 2972 mL min−1 in matched controls
Conclusions One year after hospital discharge for COVID-19, the majority, 77%, had normal exercise capacity. Only every fourth had exercise intolerance and in these circulatory limiting factors were more common than ventilatory. Deconditioning was common. V′O2peak and oxygen pulse improved significantly from 3 months.
Background
Severe coronavirus disease 2019 (COVID-19) may be followed by organ dysfunction and persisting symptoms [1, 2]. In hospitalized patients, the lung has been the organ primarily affected by COVID-19 infection, and consequently, respiratory symptoms and exercise intolerance are prevalent [3, 4]. Dyspnoea is the most frequently reported respiratory symptom after COVID-19, affecting about half of the patients 3 months after hospitalization for COVID-19 [5].
The cardiopulmonary exercise test (CPET) provides an integrated assessment of the cardiorespiratory system and is considered the gold standard for evaluating exercise capacity and dyspnoea on exertion. Hence, in patients who continue to experience dyspnoea after COVID-19, CPET is a valuable tool. Deconditioning has been considered the main limiting factor of exercise capacity 3 months after COVID-19, followed by circulatory and ventilatory limitations [5–7]. However, most studies have a short time interval between COVID-19 diagnosis and follow-up, usually 3 to 6 months [6, 7], which may not be long enough for pulmonary structural changes and exercise abnormalities to resolve. Whether or not these limitations to exercise persist 1 year after COVID-19 infection, is still unknown.
In a prospective study of patients hospitalized for COVID-19, we aimed to:
Determine cardiopulmonary exercise capacity at 12 months, including the impact of persisting dyspnoea and treatment in intensive care unit (ICU).
Assess the change in cardiopulmonary exercise capacity from 3 to 12 months, and
Compare the results from the post-COVID-19 population with a matched control group without a history of COVID-19.
We hypothesized that exercise capacity would improve from 3 to 12 months after discharge.
Methods
Study design and variables
The present study was a substudy of all patients undergoing CPET at 3 and/or 12 months in a prospective observational study of patients hospitalized for COVID-19 in Norway, the “Patient-Reported Outcomes and Lung Function after hospitalization for COVID-19” (PROLUN). The main study included participants ≥18 years with a discharge diagnosis of COVID-19 before 1 June 2020 from six hospitals in different parts of Norway. The patients were invited to follow-up visits 3 and 12 months after discharge, with pulmonary function, dyspnoea and CT findings as primary outcomes [5, 8]. Registration identifier number at Clinical Trials.gov was NCT04535154.
Among the 264 PROLUN patients providing consent, 256 attended at least one of the visits.
In the present substudy, CPET was performed in 190 patients at 3 months, and 187 at 12 months (figure 1). One of the centres performed CPET only at 12 months (n=23). All patients with valid CPET at either 3 or 12 months (n=210) were included in the analyses (figure 1).
Flow chart of the study population.
Informed consent was obtained from all participants. Regional Ethics Committee, South-Eastern Norway (no. 125384) and data protection officers at the participating hospitals provided ethical approval.
Comorbidity was based on both medical records and self-report, and included a previous diagnosis of chronic obstructive pulmonary disease, myocardial infarction, heart failure, cerebral vascular accident, or peripheral vascular disease.
Obesity was defined as body mass index (BMI) >30 kg m−2. The WHO Ordinal Scale for Clinical Improvement was used to score the severity of COVID-19 infection [9].
Dyspnoea and pulmonary function tests
The modified Medical Research Council (mMRC) scale (grade 0 to 4) was used to classify self-reported dyspnoea [10]; mMRC 0 was defined as no dyspnoea.
Spirometry, body plethysmography and diffusing capacity of the lung for carbon monoxide (DLCO) were performed (Jaeger Master Screen PFT Vyaire Medical GmbH, Germany) according to guidelines, using Global Lung Function Initiative (GLI) reference values [11–13].
CPET
Stepwise incremental treadmill exercise according to a modified Bruce protocol was applied for CPET (Vyntus CPX, Vyaire Medical), which included continuous measurement of electrocardiogram (ECG) and pulse oximetry (SpO2). Mouthpiece and nose clip were used for breath-by breath measurements of ventilation (V′E), oxygen consumption (V′O2), and expired carbon dioxide (V′CO2). Borg CR10 scale was used for the assessment of perceived exertion and dyspnoea [14]. V′O2·kg−1, oxygen pulse (V′O2peak /HR), respiratory exchange ratio (RER), V′E/V′CO2 slope, and ventilatory equivalents, were calculated. Ventilatory efficiency was assessed by the V′E/V′CO2 slope up to the ventilatory compensation point and by nadir ventilatory equivalent for CO2 (V′E/V′CO2nadir). Breathing reserve was calculated as (1- V′E/maximal voluntary ventilation (MVV))×100%, using an estimate of forced expiratory volume in 1 s (FEV1)×40 for MVV [15]. The anaerobic threshold (AT) was assessed by the V-slope method [16]. Post-exercise capillary blood samples were collected from the fingertip within 1 min and analyzed for lactate, pH, and carbon dioxide tension (PcCO2) (ABL 800 Flex, Radiometer Medical, Denmark). Norwegian reference values, from a healthy population, were used to calculate CPET values relative to expected for age and sex (% predicted) [17], except for V′E/V′CO2 slope and V′E/V′CO2nadir [18]. The prediction equation for V′O2peak (mL·min−1) [17] was used for assessment of exercise intolerance and V′O2 at AT % of predicted V′O2max. Exercise intolerance was defined as V′O2peak <80% predicted. Ventilatory limitation to exercise was defined when breathing reserve was <15% [15]. The Wassermann flowchart was used to define circulatory limitation in participants when it led to a circulatory category [16], including ECG changes consistent with ischemia or arrhythmia. Deconditioning was defined as V′O2peak <80% predicted with normal breathing reserve and no evidence of cardiocirculatory pathology (assessed by ECG, V′E/V′CO2 slope, and O2-pulse curve) with normal or low V′O2peak at AT.
Ventilatory inefficiency was defined as V′E/V′CO2 and/or V′E/V′CO2nadir z-score>1.645 [18]. Dysfunctional breathing was determined by random swings in ventilation due to chaotic changes in tidal volume and respiratory frequency, accompanied by hypocapnia and respiratory alkalosis. CPET was considered submaximal, and thus inconclusive and invalid, when exercise was restricted by non-cardiopulmonary factors, including back or leg pain, in patients with RER<1.0 and lactate <3.0 mmol L−1.
Matched controls (HUNT4 HOPE)
The matched controls were recruited from the HUNT4 HOPE, part of the large population-based Norwegian study HUNT (The Trøndelag Health study), where CPET and echocardiography were performed in 2461 participants between 2017 and 2019 [19]. After matching individually for comorbidity and sex, matching on group level was done for age, BMI, and blood pressure. HUNT4 HOPE CPET treadmill protocol increased inclination and/or speed every minute until voluntary exhaustion. Continuous gas analysis was performed with the MetaLyzer II (Cortex Biophysik Gmbh, Leipzig, Germany) mixing chamber system with patients wearing an oro-nasal mask.
In total 177 patients and 207 controls were included in the analysis.
Statistical methods
Data are presented as mean (sd), median (25- and 75-percentiles), or frequency (%), as appropriate. Normality of data and residuals was checked by inspection of histograms and QQ-plots and Shapiro-Wilk's or Anderson-Darling tests.
The change in outcome variables from 3 to 12 months and potential interactions with ICU stay or dyspnoea were analyzed by linear mixed models (LMM). A subject-specific random intercept accounted for within-subject correlations. Models with and without interaction between ICU stay or dyspnoea and the categorical time variable (3 and 12 months) were fitted. Since the interaction effect was not statistically significant, results for the effect of time on ICU stay or dyspnoea from main effect models are presented. All models included sex, comorbidity (present or not present), BMI, and age, all measured at 3 months, as additional covariates, and a fixed effect for the hospitals to adjust for a potential centre effect. To explore other potential predictors of change in the outcome variables, LMMs including interactions of time with obesity, comorbidity, age, sex, in addition to ICU stay and dyspnoea, were fitted similarly. The lmer function and the models in the lme4 package were fitted in the R version 3.4.4 [20, 21].
A subset of CPET variables were compared between the patients with COVID-19 and the controls using multiple regression analysis, adjusting for age, sex, BMI, resting systolic blood pressure, COPD, diabetes, previous heart failure, and previous myocardial infarction. After matching for comorbidity and sex, matching on group level was done for age, BMI, and blood pressure. Because of the partly individual matching of controls (see Methods), LMM were first fitted to account for potential within-pair correlations. Because these correlations were very small, we used ordinary regression models. For the compared CPET variables, the normality assumption for the residuals was considered reasonable. Other assumptions for regression analyses were checked by correlations between the variables, variance inflation factor and inspection of plots of residuals versus predicted and found to be satisfactory.
The main study, PROLUN, was an observational study with the prevalence of reduced lung function after hospitalization and interstitial lung findings after 3 and 12 months as primary outcomes. There were no a priori sample size calculations for these outcomes, and the study included all eligible patients in the six hospitals until 1 June 2020.
p-values<0.01 were considered statistically significant to give some protection against false positive results.
Results
Study population characteristics
The 12-month visit was completed at a median (25th-75th percentile) of 376 (309–472) days after discharge from the hospital. The mean age was 58.1(13.8) years, 41% were female (n=85) and mean BMI was 28.5(4.8) kg·m−2. The patients were hospitalized for a median of 6 (3–11) days, 41 patients (20%) were treated in an intensive care unit (ICU) for a median of 10 (4–15) days, and 27 (13%) were intubated and mechanically ventilated for median 10 (7–15) days (supplementary table 1). Comorbidity at baseline was present in 26 patients (13%) and obesity in 59 patients (29%). Figure 2 summarizes the main findings of the study. Supplementary table 1 summarizes the descriptive data of the study population.
Estimated changes in CPET variables in COVID-19 patients from 3 to 12 months in estimated values from linear mixed models and observed values at 3 and 12 months
Central illustration.
At 12 months 41 (22%) had supervised rehabilitation. The majority attended in-patient rehabilitation (n=27), fewer attended community-based (n=8) and out-patient (n=6) rehabilitation.
The patients lost to follow-up were slightly older, had a higher degree of obesity, were female, fewer were born in Norway, and had lower V′O2peak. They had similar rates of ICU admission, comorbidity, and dyspnoea.
Descriptive results
Dyspnoea
mMRC was≥1 in 86 patients (47%) at 12 months compared with 89 patients (51%) at 3 months (supplement table 1).
Pulmonary function tests at 12 months
Mean (sd) FEV1 was 94 (15)% predicted, forced vital capacity (FVC) 97 (13)% predicted, total lung capacity (TLC) 97 (17)% predicted, and diffusion capacity of the lung for carbon monoxide (DLCO) 92 (17)% predicted. Results below lower limit of normal (z-score <-1.645) were observed in 12 (7%) for FEV1, in 14 (8%) for FVC, and in 25 (15%) for DLCO. V′O2peak % predicted correlated with TLC % predicted (r=0.38, p< 0.001), but not with FEV1% predicted (r=0.01, p=0.94), or DLCO % predicted (r=0.01, p=0.95).
Cardiopulmonary exercise test at 12 months
Observed CPET variables at 12 months are presented in table 1.
Exercise limiting factors
V′O2peak <80% predicted was observed in 40 patients (23%). The exercise limiting factors were circulatory limitations in 11 (28%), ventilatory limitations in 7 (17%), and other factors in 22 (55%). Among the 22 patients with other limiting factors, 3 satisfied our definition of dysfunctional breathing, and 19 satisfied the definition of deconditioning.
Ventilatory inefficiency
Ventilatory inefficiency was observed in 30 patients (17%) and was related to ventilatory factors (n=6), circulatory factors (n=10), and dysfunctional breathing (n=13). The cause of ventilatory inefficiency could not be established in one participant. Patients with ventilatory inefficiency had lower mean (sd) V′O2peak % predicted (74 (19) versus 97 (17) %, p<0.001), end-tidal CO2 (PETCO2) at maximal exercise (4.1 (0.4) versus 4.7 (0.5) kPa, p<0.001), and lactate (6.9 (3.6) versus 9.7 (3.7) mmol·L−1, p<0.001) compared to those with normal ventilatory efficiency. Among 27 patients with ventilatory inefficiency, 17 (63%) reported dyspnoea by mMRC. Among 85 patients reporting dyspnoea, 17 (20%) had ventilatory inefficiency.
Changes from 3 to 12 months and determinants of change
Exercise intolerance was observed in 23% at 12 months, compared to 34% at 3 months. V′O2peak, oxygen pulse, lactate, and PCO2, as well as V′O2 at AT % of predicted V′O2max, were significantly higher at 12 months compared to 3 months after hospital discharge (table 1). Estimated mean increases in V′O2peak % predicted and V′O2·kg−1% predicted were 5.0 percent points (pp) (95% CI 3.1 to 6.9) and 3.4 pp (95% CI 1.6 to 5.1), respectively (Table 1).
There was little or no evidence of any interactions between time and age, sex, obesity and comorbidity (figure 3a, supplementary tables 2 and 3).
V′O2peak % predicted and oxygen pulse % predicted, and V′E/V′CO2 slope according to dyspnoea and ICU status at three and 12 months.
V′O2peak % predicted and oxygen pulse % predicted, and V′E/V′CO2 slope according to obesity and comorbidity status at three and 12 months.
Estimated effect of dyspnoea and ICU stay on CPET variables from linear mixed models (n=210)
CPET variables compared between controls and COVID-19 patients at 12 months follow-up.
SpO2 was 98 (1)% at rest and 95 (4)% at maximal load at 12 months. Desaturation (defined as SpO2 desaturation >5pp) was not observed during CPET at 12 months compared to in 34 patients (23%) at 3 months.
Impact of dyspnoea or ICU treatment on cardiopulmonary function
Patients reporting dyspnoea at 3 months were more females, had a higher BMI and more comorbidity compared to patients without dyspnoea, but there were no differences in pulmonary function or number treated with non-invasive ventilation or mechanical ventilator (Supplementary table 1). Patients reporting dyspnoea had lower V′O2peak and higher V′E/V′CO2 slope at 12 months compared to those with dyspnoea (table 2, figure 3b). However, the changes in CPET variables from 3 to 12 months were the same for patients with and without dyspnoea (table 2, figure 3b).
Patients admitted to an ICU at the index hospitalization had lower V′O2peak and oxygen pulse compared to patients not treated in an ICU (table 2, figure 3b). However, the changes in CPET variables from 3 to 12 months were the same for patients with and without ICU treatment (table 2, figure 3b).
Comparison between COVID-19 patients and matched control group
At 12 months, the COVID-19 patients had lower V′O2peak and V′O2peak·kg−1 than matched controls (table 3). Maximal heart rate, breathing frequency and V′E were lower in the COVID-19 patients compared to the matched controls (table 3).
Mean RER at maximal load was 1.10 for the controls and 1.07 for the patients, which was a significant difference in the adjusted analysis (supplementary table 4). However, there was only little evidence of differences in CPET variables between controls and patients, when RER in patients was dichotomized to greater or less than 1.10 (supplementary table 4).
Discussion
The main findings in this study were that the majority of COVID-19 patients had normal exercise capacity at 12 months, exercise intolerance was reduced, and V′O2peak and oxygen pulse improved from 3 to 12 months after hospitalization. The frequency of ventilatory limitation was low at 12 months. Patients with dyspnoea or ICU treatment had lower values of V′O2peak at 12 months, but similar improvement from 3 to 12 months, compared to patients without dyspnoea or ICU-treatment. The study patients had lower V′O2peak at 12 months compared to matched controls.
Exercise capacity and limitations
Exercise capacity improved from 3 to 12 months after hospitalization, and the increase in V′O2peak was considered sufficient to have a positive impact on activities of daily living. At 12 months, the majority had regained normal exercise capacity and the prevalence of exercise intolerance was reduced to every fourth patient.
Circulatory limitations were more frequent than ventilatory limitations in patients with exercise intolerance. Mean values of pulmonary function tests were within normal limits at 12 months, few had abnormal values. Except for TLC, there were no correlations between V′O2peak and pulmonary function tests, which support that exercise capacity for most patients is limited by factors other than the lungs.
The majority of patients with exercise intolerance were limited by other than circulatory and ventilatory factors. This group included patients with deconditioning and dysfunctional breathing, but other virus induced limitations may also have been present. Our study was limited to non-invasive methods, thus we cannot explain all aspects of the mechanisms interfering with exercise capacity. However, deconditioning due to inactivity seems to be the most prevalent exercise limitation. Naeije and colleagues grouped together 581 COVID-19 patients from 11 studies and found a CPET profile of deconditioning in the recovery phase of an acute inflammatory process [22].
As stated by the Fick equation, V′O2peak=cardiac output x arteriovenous oxygen difference, a low V′O2peak may be related to either reduced cardiac output or reduced peripheral oxygen extraction. Both these mechanisms may apply in patients with deconditioning [23, 24]. Furthermore, reduced peripheral oxygen extraction has been shown in COVID-19 patients with small fibre neuropathy, complicating evaluation of exercise limitation even more [25, 26].
Dysfunctional breathing with large disharmonic variations in tidal volume and respiratory frequency, accompanied by hypocapnia and respiratory acidosis, was limiting exercise capacity in a few patients. Similar dysfunctional breathing patterns have also been observed in other studies [27, 28].
Dyspnoea
Dyspnoea was reported by half of the patients, consistent with findings in other studies [29]. Among patients with dyspnoea, there were more females, more obesity, and more comorbidity compared to patients without dyspnoea. Patients with dyspnoea had lower V′O2peak·kg−1% predicted compared to those without dyspnoea. However, in the patients reporting dyspnoea, few had circulatory or ventilatory limitations. This is similar to observations in a CPET study of COVID-19 patients with prominent dyspnoea, where only mild physiological abnormalities were found [30].
Patients with dyspnoea had reduced ventilatory efficiency, with dysfunctional breathing as the most frequent cause. Although ventilatory inefficiency and hyperventilation may account for some of the reported dyspnoea in our study, only one-fifth of the patients with dyspnoea showed ventilatory inefficiency. Perceived dyspnoea is often multifactorial [31], complicating the interpretation of this symptom. Given the magnitude of the COVID-19 pandemic, it will be essential to differentiate symptoms caused by COVID-19 from dyspnoea due to other etiologies.
ICU treatment
Patients treated in an ICU had the same improvement in V′O2peak and oxygen pulse from 3 to 12 months compared to patients without ICU treatment. However, they still had lower V′O2peak despite more frequent rehabilitation.
Patients and matched controls
Even though the patients in our study improved their exercise capacity from 3 to 12 months, it was still not normalized compared to the matched controls. Maximal heart rate and ventilation were lower among the COVID-19 patients compared to matched controls, indicating slightly submaximal performance. This could have influenced the comparison between patients and matched controls, but subgroup analyses show that patients with RER greater or less than 1.1 both have lower V′O2peak compared to the matched controls.
Limitations
As all study patients were hospitalized in the first phase of the pandemic when vaccines were unavailable, our results may not apply to a vaccinated population. The study was performed in hospitalized patients during acute COVID-19 infection and the results may not apply to the subjects with long COVID who were not hospitalized.
Unlike the COVID-19 patients, the controls have not been hospitalized. However, the only purpose of the controls is to account for pre-existing comorbidity when evaluating if the patients have recovered their expected exercise capacity. Timely change in exercise capacity cannot be compared, as the controls only had one assessment.
CPET was performed using different equipment and protocol in the COVID-19 population and the matched HUNT control group. There have been reports of higher V′O2peak in the HUNT fitness population compared to other population cohorts and difference between patients and other controls might have been smaller [17, 19).
CPET was performed on treadmill which gives 5–10% higher V′O2peak compared to cycle ergometer. Cardiac output was not measured during exercise, and muscle biopsies were not performed, thus evaluation of deconditioning is hampered with some uncertainty.
The studýs strength is the inclusion of most patients hospitalized for COVID-19 in the study's catchment areas in Norway at the beginning of the pandemic, representing an unselected, thus representative, hospital population.
Conclusions
Exercise capacity was normal in 77% of the patients 1 year after hospital discharge for COVID-19. In patients with exercise intolerance, circulatory limitation to exercise was more common than ventilatory limitation. Deconditioning seemed to be the most prevalent exercise limitation, but other, unknown mechanisms may have contributed to exercise intolerance. V′O2peak and oxygen pulse improved significantly from 3 to 12 months, but V′O2peak was lower compared to matched controls. Even though patients with dyspnoea or ICU treatment had lower V′O2peak at 1 year, they still had similar improvement from 3 months, compared to patients without dyspnoea or ICU treatment.
Acknowledgements
The Trøndelag Health Study (HUNT) is a collaboration between HUNT Research Centre (Faculty of Medicine and Health Sciences, Norwegian University of Science and Technology NTNU), Trøndelag County Council, Central Norway Regional Health Authority, and the Norwegian Institute of Public Health.
Footnotes
Source of funding: This work was supported by the National Association for Heart and Lung Diseases, Akershus University Hospital and Norwegian Health Association. Akershus Universitetssykehus; DOI: http://dx.doi.org/10.13039/501100012446; Nasjonalforeningen for Folkehelsen; DOI: http://dx.doi.org/10.13039/501100013263; National Association for Heart and Lung Diseases.
Conflict of interests: C.B. Ingul has received lecture fees from Bayer AS, unrelated to the current study.
Conflict of interests: I. Skjørten has provided lectures for doctors’ education paid by Norwegian Directorate of Health and Norwegian Medical Association.
Conflict of interests: G. Einvik has received research grants from AstraZeneca to perform the current study.
Conflict of interests: A. Edvardsen has received payment or honoraria for lectures, presentations or educational events from GlaxoSmithKline.
Conflict of interests: Chiesi. K. Stavem has received consulting fees from UCB Pharma and MSD, unrelated to the present study.
Conflict of interests: All other authors have nothing to disclose.
- Received April 8, 2022.
- Accepted August 27, 2022.
- Copyright ©The authors 2022
This version is distributed under the terms of the Creative Commons Attribution Non-Commercial Licence 4.0. For commercial reproduction rights and permissions contact permissions{at}ersnet.org