Our data concur with other lung function reports of COVID-19 survivors [2–5] by showing a restrictive pattern with a reduction in overall gas transfer (TLCO) in a proportion of COVID-19 survivors. However, we also note the relatively preserved or slightly raised KCO in 78.1% of these patients. Our data suggest that this pattern of physiological dysfunction may be more prevalent in COVID-19 patients admitted to ITU who were treated on the ward alone (and weren’t intubated or mechanically ventilated). This pattern was also reported by Mo et al  but was not sufficiently explained. We have considered several possible explanations for this pattern including (i) Intussusceptive angiogenesis (remnants of lung vascular damage from COVID-19) or pulmonary haemosiderosis and (ii) Extrapulmonary restriction (obesity, pleural issues or muscle weakness).
The most common cause of this pattern is extrapulmonary restriction, with obesity being the most likely cause . Indeed given the body habitus of all our patients to be predominantly overweight /obese, this seems a likely hypothesis. However, reviewing the literature suggests that this pattern is usually only observed in severe obesity (BMI > 40) . Only 4 ward and 9 ITU patients were above this threshold, so excessive obesity is unlikely to be the sole reason for this pattern. Although extrapulmonary restriction is more associated with upper body fat , we didn’t collect this data in our patients.
Several studies suggest that the reduction in volume caused by obesity was insufficient to explain the increase of KCO found in patients with small lung volumes [18, 21]. Usually TLCO in healthy subjects’ decreases as the inspired volume reduces at a rate of about 3.3% per 10% decrease of vital capacity. In this current study, the inspired volume (Vi) was 94.8% (3.6%) of the vital capacity (VC) indicating a close match of Vi with VC. On average, VC was reduced by about 15% which would mean Vi should have been reduced by about 5%, but it was reduced by around 15% which suggests more than an obesity effect. It is understood that there is an increase in capillary blood volume in obesity that is thought to lead to the increase in gas transfer .
Whilst the COVID-19 survivors in the Mo et al  study also showed the same reduced TLCO with raised KCO pattern, their population all had a mean BMI below 25, which suggests that the phenomenon is not related to obesity causing extrapulmonary restriction.
Muscle weakness causing extrapulmonary restriction would show a reduction in muscle pressures. Although we didn’t measure respiratory muscle function, Huang et al recently showed normal maximal inspiratory/expiratory muscle pressures in post-COVID-19 survivors , which argues against this hypothesis. There is no suspicion from the recent COVID-19 literature that respiratory muscle weakness is a feature of COVID-19 recovery, although neuropathy and general fatigue have been noted as a key symptom in sick and recovering patients.
A high KCO indicates a predominance of pulmonary capillary volume (Vc’) over alveolar volume, which may arise for different reasons. Incomplete alveolar expansion but preserved gas exchange unities frequently lead to KCO > 120–140% or even higher (i.e. extra-parenchymal restriction, such as pleural, chest wall or neuromuscular disease) [21–24]. Pleural changes have been identified in COVID-19 patients using imaging [25–29] and at autopsy . Alternatively, an increase in pulmonary blood flow from areas of diffuse (pneumonectomy) or localised (local destructive lesions/atelectasis) loss of gas exchange units to areas with preserved parenchyma can lead to more modest increases in KCO. However, a high KCO can also be seen with normal or near-normal VA when there is increased pulmonary blood flow or redistribution (e.g. a left-to-right shunt or asthma). Intussusceptive angiogenesis  as a result of chronic infection is a dynamic intravascular process that can modify the structure of the microcirculation. Recently, this has been shown to be present at autopsy in COVID-19 patients  and could explain some or most of the rise in KCO observed.
Haemosiderosis, which is the deposition of extra-vascular haemoglobin, may also be a cause. Alveolar haemorrhage is a possible mechanism given the vascular destruction reported in active COVID-19 and fits with the lung function results. In haemosiderosis, macrophages convert the iron in haemoglobin into haemosiderin within 36–72 h [33–34] and the haemosiderin-laden macrophages can reside for up to 4–8 weeks in the lungs. Pulmonary haemosiderosis is usually considered to be from persistent or recurrent intra-alveolar bleeding, which may explain the symptoms of “long COVID” and the time course of improvement in symptoms. The effect of haemosiderosis on interpretation of the gas transfer test has been highlighted by Hughes [18, 35]. However, there was only mild anaemia (haemoglobin < 120 g/L) in 7 (14.9%) of the ward patients and 12 (25.5%) of ITU patients and all gas transfer tests were corrected for haemoglobin. This makes haemosiderosis an unlikely explanation for most of the abnormalities we observed in gas transfer.
Another explanation for the reduced TLCO/raised KCO pattern could be the development of necrotising pulmonary capillaritis occurring in isolation . This arises from diffuse interstitial neutrophilic infiltration with cell fragmentation and, because of apoptosis, cellular accumulation within the lung tissue, filling the interstitial space. This can lead to expansion and fibrinoid necrosis. As a result of these processes, the integrity of interstitial capillaries is damaged, allowing red blood cells to pass through the alveolar capillary basement membranes, freely enter the interstitial compartment and flood alveolar spaces. Clinically, this diffuse alveolar microhaemorrhage enables the CO in the gas transfer test to combine with this “occult” blood or haem from haemosiderosis and effectively raise the KCO. However, global gas transfer (TLCO) is not as affected because of the counteractive restrictive defect that causes a decrease in lung volume and, hence, alveolar surface area which, in turn, has a greater effect on decreasing the TLCO than the rise due to the diffuse local haemorrhage.
A similar TLCO and KCO pattern seen in SARS  was thought to be the result of muscle wasting and corticosteroid induced myopathy. We have insufficient data to prove or disprove this hypothesis currently so, even though it is unlikely, it cannot be excluded.
More males than females (as reported elsewhere in COVID-19) required assisted ventilation/oxygenation on ITU and, therefore, probably had worse infections. These hospitalised patients were predominantly overweight /obese, which is another known risk factor in severe COVID-19 for an increased likelihood of hospitalisation, ITU admission and morbidity. Our data also show that more never-smokers were admitted to hospital and ITU greater but this may be because smokers with COPD either shielded from COVID-19 and never got the disease or died on ITU and weren’t followed up. Ethnicity is also known to be a factor associated with an increase in incidence and severity of COVID-19 in patients from black, Asian and minority ethnic (BAME) communities in the UK . However, we did not note any significant differences in ethnicity between ITU and ward patients, so this does not appear to be causal in the outcome of the gas transfer tests.
There were no major differences in lung function between ward and ITU patients, despite a statistically lower KCO on average in ITU patients. It might have been expected that patients who had had mechanical ventilation for severe covid pneumonitis to have had worse lung function but this wasn’t the case at 3 + months.
Our data also show that few of the patients had any evidence of airflow obstruction on spirometry (FEV1/FVC SR) and that never-smokers showed greater hospitalisation but this may be because smokers (more likely to have COPD) either shielded from COVID-19 and never got the infection, or died on ITU and weren’t followed up. Some may interpret this as evidence of the “protective effect” of smoking in COVID-19 .
The anticipated alteration in breathing patterns was not evident when compared with reference values. The ITU patients had no more dysfunctional breathing patterns than the ward patients. Whilst many post-ITU patients display dysfunctional breathing immediately on leaving ITU, it appears to improve rapidly in most, so by 3 + months there are only 20% showing abnormality.
These abnormal SLP values were both lower and greater than the normal range with no consistent pattern. We had wondered whether SLP could have been used to detect dysfunctional breathing patterns in COVID-19 survivors, linked to the severity of impaired gas transfer and, therefore, lung damage. However, this relationship wasn’t strong, so screening for lung function impairment should continue to use traditional spirometry and gas transfer in patients who have symptoms compatible with post-COVID-19 lung changes.
The reasons for abnormal breathing patterns could be the result of (a) obesity, (b) COVID-19 itself causing pneumonia, leading to sepsis, and producing delirium, (c) the effects of sedation and medication on breathing centres, or (d) mechanical ventilation and oxygen therapy.
It is well established that the work of breathing is increased and the total respiratory compliance is decreased in obesity . This could be a cause of altered breathing patterns, although there is no obvious link between the two in this data.
The potential errors with lung function testing have been minimised since all testing was performed on calibrated equipment and was measured by experienced, well-trained personnel. In addition, all equipment was monitored with a stringent quality control protocol, including both physical and biological quality control, within tests quality checks and review of all tests by senior physiologists [7–8]. Furthermore the VA/TLC ratio in the sub group who had lung volumes measured showed VA to be on average within 7% of TLC, which indicates good consistency and test quality. Unfortunately, we were unable to perform lung volume measurements in all patients due to limited lung function timeslots.
Our population and their treatment may be different from other centres who have published lung function data in COVID-19. Certainly the body habitus of the data from Mo et al  shows normal BMI values, unlike our population who were predominantly obese/overweight. However, the ventilation regimens adopted in the UK and at our centre were based around the WHO guidance for COVID-19 following experience from Wuhan early in the pandemic .
We didn’t measure muscle pressures as this wasn’t a prospective study. However, Huang et al  found no abnormalities in respiratory muscle function. We also didn’t measure abdominal obesity as this may have a different effect on gas transfer compared with upper body obesity.
Changes we have seen may not just be due to COVID-19 directly but, also, therapeutic insults/interactions (e.g. corticosteroids, oxygen, and mechanical ventilation) or other pathophysiological events such as delirium or sepsis. However, similar regimens were used across world after the Wuhan experience was published. Nevertheless, the physiological changes in this population will remain a legacy for many patients who have had COVID-19 and will add further demands to already over-subscribed, limited in performance lung function facilities worldwide . Consequently, because of aerosol-generating properties of lung function testing and the difficulties delivering testing , it has not been possible to test all patients at the same time since hospital discharge.
The reference values for breathing patterns (using SLP) are a recently derived and validated set of references values and may not be good at discriminating normal from abnormal.
Future work should measure TLNO or Dm/VC in COVID-19 so that the vascular component and membrane components of the gas transfer processes can be better understood. It would be expected that a pattern consistent with altered capillary blood volume may become evident.
Some patients in our post-COVID-19 clinics will have further follow-up lung function after another 3 + months, so it will be interesting to see if the changes we have found (particularly in gas transfer) are related to any change in body habitus or to the repair of the suspected lung damage we have highlighted here.
We found similar restrictive patterns (reduced vital capacity and alveolar volume) in survivors with moderate and severe covid pneumonitis whether admitted to wards or ITU. There is a mild reduction in gas transfer (TLCO) but a preservation/ relative rise in transfer coefficient (KCO). These results can be explained partly by (i) obesity causing extrapulmonary restriction but perhaps also by (ii) haemosiderosis from lung damage and (iii) localised microvascular changes in lung capillaries. Potential respiratory muscle fatigue/weakness is unlikely to be a causal factor in our study.
Abnormal breathing patterns (outside 1.64 SRs) of reference data showed 20% of subjects displayed one or more abnormality of breathing in duty cycle, phase angle and respiratory rate. However, no consistent breathing pattern abnormalities were evident. The use of breathing patterns to screen post-COVID-19 patients for those who require more extensive lung function testing isn’t borne out in our population.
We conclude that the residual changes in lung function and breathing patterns observed at 3 + months are similar whether patients attended wards or were mechanically ventilated on ITU.