Long-term comprehensive cardiopulmonary phenotyping of COVID-19

In our longitudinal post-COVID clinic, 71 patients (46 males, 25 females) consented to participation (DZL, German Center for Lung Research, database and biobank for research purposes, NCT04442789). The mean age at initial infection was 56 years (SD ± 12.4 years). 21 patients had been treated as outpatients, 24 patients had been treated on a general medical floor, and 25 patients had been treated in an intensive care unit (the treatment setting of 1 patient was not ascertainable). Out of these, 8 patients underwent non-invasive ventilation and 13 patients received invasive mechanical ventilation. The mean baseline Charlson Comorbidity Index was 1.9 (SD ± 2.2). A total of 150 encounters were available for analysis, an overview of the measurements available at the pre-defined intervals is available in Additional file 1 (Fig. 1).

Fig. 1figure 1

A total of 71 patients were seen for at least 1 follow-up encounter. The number of patients that showed up for each follow-up appointment is listed as well as the number of patients who underwent individual testing at the predefined intervals

Most patients who presented to the post-COVID clinic reported at least one symptom (91%). The most commonly reported symptom was decreased exercise tolerance (44%), followed by dyspnea (39%), and fatigue (26%) (Table 1). Quality of life was assessed by the EQ-5D questionnaire, the average visual analog scale (VAS) score was 72.8 (SD ± 17.7). The average EQ-5D VAS did not change significantly over the first year following the COVID-19 diagnosis (Fig. 2A). Patients treated in the ICU or hospital did not have a lower EQ-5D VAS compared with patients who only required outpatient treatment (average EQ-5D VAS from encounters in post-ICU patients 71.8 vs. 79.2 in post-general floor patients vs. 67.4 in previous outpatients).

Table 1 Patient characteristicsFig. 2figure 2

A EQ-5D Visual Analog Scale (VAS) over time. B Spaghetti plot of individual EQ-5D scores over time, dotted line represents the mean over time, shaded area represents ± 1 standard error. C Selected pulmonary function test results over time (blue dot – FEV1, red square FEV1/VC, green triangle TLC, purple triangle DLCO). D Spaghetti plot of individual exertional A-aDO2 values over time, dotted line represents the mean over time, shaded area represents ± 1 standard error. E Follow-up DLCO measurements as % predicted based on initial treatment setting at 3, 6, and 12 months, ± SD. Significance assessed by ANOVA and post-hoc Tukey test. F B-line severity score over time, ± SD. G Relative distribution of significantly abnormal (B-line severity score > 1) as % over time. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001

We found a rapid recovery of lymphocyte counts at 3 months after the initial infection (Additional file 1: Figure S1). There was no significant association between lymphocyte counts (total or individual CD4 and CD8 numbers) and symptoms as measured by EQ-5D VAS or exertional capacity as measured by maximal oxygen uptake (VO2max).

Patients underwent serial pulmonary function testing, ideally at the pre-defined 3, 6, and 12-month intervals from initial infection. There was no indication of significant obstructive airway disease with the FEV1/VC ratio (Tiffeneau-Pinelli index) stably ranging above 70% throughout the first year on average (Fig. 2C, Table 2). The DLCO was significantly reduced during at the 3-month time point (63% predicted, SD ± 16.9%) phase following infection with subsequent improvement, but remained slightly reduced throughout the first year. TLC was reduced acutely but remained within normal limits, on average, throughout the first year. In the patients (n = 6) that presented within 6 weeks of their initial infection, PFTs also revealed decreased FEV1 and FVC (54.6%, SD ± 12.5% and 46%, SD ± 13.6%) with a high FEV1/VC ratio (91.5%, SD ± 6.7%), suggestive of increased lung elastic recoil. A low TLC (59.90% predicted, SD ± 19.6%) demonstrates restriction. The DLCO maneuver was only available from 2 patients at this early time point (47.5 mL/min/mm Hg, SD ± 7.8 mL/min/mm Hg). Subsequent follow-up DLCO values were lower for patients who had been treated in the ICU compared with those treated as outpatients or the medical floor (Fig. 2E). Similar to the reduction in DLCO, the exertional alveolar-arterial oxygen gradient (A-aDO2) was measured and tracked over time, it was abnormal on 19 of 95 CPETs (abnormal widening defined as > 35 mmHg). The individual values were tracked over time (Fig. 2D). Beyond 6 weeks, average exertional A-aDO2 values were not significantly different over time (ANOVA, p = 0.155).

Table 2 Serial measurements

On lung ultrasound, initially all patients exhibited excessive B-lines (defined as a B-line score of > 1) if seen within 6 weeks of COVID-19 diagnosis. This dropped to 31% of patients at 3 months and further decreased to 6% after 12 months (Fig. 2F, G). The lung ultrasound score at 3 months demonstrated good correlation with the degree of gas exchange impairment as measured by contemporaneous exertional A-aDO2 during cardiopulmonary exercise testing (Spearman’s rho 0.738, p < 0.001) and DLCO (Spearman’s rho − 0.547, p = 0.003). This association was not observed on subsequent testing, where anatomic changes seen on ultrasonography were largely resolved, while mild impairment in gas exchange (as determined by exertional A-aDO2 and DLCO) persisted and remained generally stable from 3 months out.

In the patients (n = 41) who underwent left heart echocardiography at least once, the initial 3D LV ejection fraction was 58% (SD ± 4%) with 35 patients (85%) having a normal LV ejection fraction (> / = 55%). 39 patients also underwent assessment by global longitudinal LV strain, which was normal (< − 18%) in 88% of patients, borderline (− 16 to − 18%) in 5 patients, and abnormal (> − 16%) in no patients. LV ejection fraction did not change significantly across follow-up time points (Table 2). Given the limited availability of 3D echo data at 12 months, we additionally assessed LVEF by E-point septal separation (EPSS) in those patients for whom only limited 2D echo images were available and calculated the LVEF. This also did not reveal any significant impairment or change in LVEF across the first year.

Dedicated right heart echocardiography was available for 66 patients. Mean values TAPSE values were 22 mm (SD ± 3 mm). This was consistent over time (Table 2). The TAPSE/sPAP ratio as a measure of RV-to-PA coupling was similarly within normal limits (mean 0.8 mm/mmHg, SD ± 0.25 mm/mmHg). 4 patients presented within 6 weeks of their infection and demonstrated a normal LVEF (57%, SD ± 1.8%), normal LV longitudinal strain (− 24%, SD ± 4.3%), normal TAPSE (21.9 mm, SD ± 5.5 mm), and normal TAPSE/sPAP ratio (0.69 mm/mmHg, SD ± 0.36 mm/mmHg).

In those patients who underwent cardiac MRI (n = 41), imaging demonstrated an average LV ejection fraction of 60.6% (SD ± 8.8%) with 34 patients (83%) having a normal LV ejection fraction on cMRI. Late gadolinium enhancement and regional wall motion abnormalities were uncommon, occurring in only 7% of patients. Ultimately, imaging review by a dedicated radiologist yielded an imaging pattern consistent with myocarditis in just 3 (7%) of patients, with 1 equivocal finding.

CPET was available for 57 patients. On average, VO2max was reduced at 79% of predicted (SD ± 19%, Fig. 3A). VO2max remained mildly decreased on average across the first 12 months (Fig. 3B). The initial VO2max increased slightly but significantly from the 3-month (n = 29) to the 6-month (n = 40) interval (75% predicted vs. 82% predicted, p = 0.04, partially overlapping t-test). Subsequent VO2max values on testing beyond 6 months revealed values of 82% (9 months, n = 5) and 81% (12 months, n = 20). Similarly, while 76% of patients demonstrated an impaired VO2max (< 85% predicted) at 3 months, this number decreased to 60% by 6 months. Notably, at 12 months, 60% of patients still had a decreased VO2max. Heart rate reserve (HRR) was increased on average (25.6%, normal: ≤ 20%), while ventilatory reserve (43.3%, SD ± 16.4%), max O2 pulse (97.9% predicted, SD ± 26.6%), AaDO2max (24.2, SD ± 11.1%), ventilator threshold (63.1% of VO2max, SD ± 13.6%) and VE/VCO2 slope (29.7, SD ± 4.956) were within normal limits (Fig. 3C). Regarding only the subset of CPET studies with decreased exercise tolerance, the main limitation was again decreased HRR with preserved ventilatory reserve, O2 pulse, anaerobic threshold, and VE/VCO2 slope. 26.8% of patients were receiving beta blockers, however, beta blockade was more common in patients previously treated in the ICU (54%, compared with 9% of previous outpatients and 16% of previous floor patients).

Fig. 3figure 3

A Distribution of all VO2max values (measured in % predicted) derived from CPET. B VO2max values over time, ± SD. C mean select individual CPET values, ± SD. D Association between A-aDO2 on exertion and VO2max (top panel) as well as the association between DLCO and VO2max. E VO2max values based on initial treatment setting over time ± SD. Significance assessed by ANOVA and post-hoc Tukey test. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001

In comparison, 2 patients underwent CPET testing within 6 weeks of their initial infection, at which time VO2max was significantly reduced (48.5% predicted, SD ± 6.4%). These two analyses demonstrated an impaired ventilatory reserve of 15% (SD ± 17%) with a HRR of 19.5% (SD ± 0.7%) and an excessively widened AaDO2max (54.6 mmHg, SD 2.1 mmHg), indicating a ventilatory/pulmonary pattern of limitation.

Overall, the degree of VO2max reduction did not correlate with the severity of gas exchange impairment as measured by DLCO or exertional A-aDO2 (Pearson’s r 0.174, p = 0.116 and Pearson’s r − 0.064, p = 0.547, respectively, Fig. 3D). Similarly, the degree of objective reduction in VO2max did not correlate with the subjective quality of life of the patients (as measured by EQ-5D, Pearson’s 0.211, p = 0.084). When splitting the data based on treatment setting, it can be shown that the persistent decrease in VO2max is mainly driven by the patients previously treated in the ICU (Fig. 3E).

For patients who underwent contemporaneous measurement of QOL by EQ-5D and CPET (68 observations from 46 unique patients), there was no correlation between the severity of subjective symptoms and the objective exercise capacity (Pearson’s r 0.083, p = 0.499). Patients who had required ICU-level care tended to have a lower VO2max that persisted, but this did not correlate with a lower QOL score as measured by EQ-5D.

When restricting the scope only to those patients who had been treated in the ICU, the pattern of limitation was similar to the that seen in the overall cohort (data not shown). When plotting the VO2max relative to the comorbidities by CCI, there was no significant association (Spearman’s rho − 0.115, p = 0.265).

留言 (0)

沒有登入
gif