In a population of COVID-19 patients who were hospitalised for severe pneumonia, lung CT abnormalities were still visible in almost 80% patients after 2-3 months from disease onset, and in almost 40% after 6-7 months. CRP as a measure of systemic inflammatory response predicted residual CT abnormalities. The best predictors were CRP peak for residual abnormalities at 2-3 months and CRP integral for persistent abnormalities at 6-7 months. Other measures of disease severity, such as LOS, the lowest PaO2/FiO2, and the need for mechanical ventilation were mediators in the relationship between CRP and persistent CT lung abnormalities. While D-dimer was associated with persistent lung abnormalities, it was not a mediator of the inflammation effect.
Our results are consistent with prior follow-up studies, which reported late resolution or residual abnormalities at 6 months in up to 72% of subjects who had had COVID-19 pneumonia.[4–7] Most patients with residual CT abnormalities did not have overt CT features of lung fibrosis. Furthermore, more than 40% of patients who showed residual abnormalities at 2-3-month achieved complete resolution at 6-7 months. These findings corroborate the hypothesis that residual CT abnormalities are the expression of slow recovery in most cases, despite data on longer-term follow-up being needed to confirm that persistent fibrotic changes are not common. In fact, non-fibrotic abnormalities could be the expression of immature fibrosis remodelling.
In the few available studies on the topic, several measures of disease severity predicted the presence of persistent lung abnormalities in COVID-19 patients,[4, 5] but the role of inflammatory burden throughout the disease course has never been investigated.
Baseline CRP has been widely studied as a predictor of COVID-19 severity in the acute phase of the disease.[10–12] Some studies have also evaluated CRP trajectories during hospitalisation, showing that rapidly rising CRP levels may predict respiratory failure and that CRP peak is higher in patients who require intubation or die.[12–17] We show that persistent lung abnormalities are also linked to inflammatory burden, with most important predictors being CRP peak for persistent damage in a subacute phase (2-3 months), and CRP integral, reflecting both the intensity and the duration of the inflammatory reaction, for persistent abnormalities at 6-7 months.
Given the known association between CRP and lung damage in the acute phase, it is not surprising that much of the effect of inflammatory burden on persistent lung abnormalities is mediated by proxies of lung failure severity, such as the need for mechanical ventilation or the lowest PaO2/FiO2 as a measure of hypoxemia. The lack of a mediation effect for the lowest PaO2/FiO2 when considering 6-7-month abnormalities is explained by the lack of an association between this severity measure and the outcome, probably due to the fact that acute hypoxemia not requiring mechanical ventilation does not lead to medium-term damage.
Higher D-dimer levels have been widely associated with COVID-19 severity and mortality,[19–21] as the infection induces a coagulopathy with secondary hyperfibrinolysis in severe cases. In our study, D-dimer peak was associated with persistent lung abnormalities, adding predictive value to the models without subtracting from the effect of CRP. Accordingly, its role as a mediator of the effect of inflammation on persistent abnormalities was modest, if any. Apparently, even if inflammation and coagulopathy are linked in COVID-19 patients,[23–25] CRP and D-dimer effects on persistent lung damage are independent, acting through different causal pathways.
This study has some limitations. First, only COVID-19 survivors who experienced severe pneumonia (extension >40% at hospital admission and/or respiratory failure during hospitalisation) were routinely evaluated with follow-up CT scans and included in the study. Second, CRP curves were not always complete, and extrapolations were done. Many values were missing for D-dimer. Furthermore, we had no information on other potential confounders, such as pre-existing lung damage. Nevertheless, our model building strategy was guided by the attempt to describe causal pathways, and we showed that several factors are on the same causal pathway of inflammation. Our findings suggest that future studies trying to understand the etiopathogenesis of COVID-19 should carefully consider which variables should be placed in multivariable analyses. Otherwise, the risk is of hiding the causal links with upstream determinants, such as hyperinflammation markers, and of introducing effect mediators. Moreover, our study focused on persistent morphological abnormalities evaluated by means of CT scan, but the evaluation of the impact of residual CT abnormalities on symptoms and lung function tests is ongoing and is beyond the scope of the present study. Finally, our attempt to classify the pattern of CT abnormalities may be questionable. In fact, the non-fibrotic CT pattern may include features (e.g., GGO), that require more time to become overtly fibrotic on CT.
In conclusion, higher inflammatory burden is associated with short- and medium-term lung damage. The peak value of CRP is the most important determinant of damage at 2-3 months, while both the duration and intensity of the inflammatory response become important for more long-lasting damage (6-7 months). Other measures of disease severity seem to be mediators of the inflammatory effect, except for D-dimer peak, confirming that the two pathways of inflammation and coagulopathy are partially independent, although both contribute to lung damage.
Data availability statement
Participant data that underlie the results reported in this manuscript will be shared after de-identification, beginning 6 months and ending at least 7 years after article publication, to researchers who provide a methodologically sound proposal with objectives consistent with those of the original study. Proposals and data access requests should be directed to the Area Vasta Emilia Nord (AVEN) Ethics Committee at [email protected] as well as to the Authors at the Epidemiology Unit of AUSL–IRCCS di Reggio Emilia at [email protected], who are the data guardians. To gain access, data requestors will need to sign a data access agreement.