Two-year follow-up of patients with post-COVID-19 condition in Sweden: a prospective cohort study

Limitations include the lack of a control group as well as the self-reported format from a subgroup of patients that initially reported lingering symptoms. This may have resulted in a selection bias, by including individuals who were more likely to report symptoms than the general population. A strength of the study that counterbalances this to some extent is that the same individuals were interviewed at both points in time. Additionally, no further evaluations of these self-reported symptoms, such as measurements of lung function, were conducted. As two years had elapsed since hospitalisation, other factors than the initial SARS-CoV-2 infection may have influenced the outcomes. As this cohort represents individuals from the initial pandemic wave, whether the results are applicable in the context of current viral mutations and vaccination statuses before infection is uncertain. It should also be noted that the same structured interview protocol from the 4-month follow-up was used for the 24-month follow-up and thus some symptoms that we now know constitute part of PCC (e.g., postural tachycardia and other dysautonomic symptoms) were omitted. Also, information bias may arise from the fact that the interview protocol is not strictly validated for the study population. However, we have previously reported that a multi-disciplinary clinical assessment attended by the vast majority of the LinCoS cohort also included in this study corroborated self-reported symptoms from the 4-month telephone interview,

which is an indication of low risk of information bias. Furthermore, p-values should be interpreted with caution as potential confounders were not adjusted for, and no correction was made for multiple testing. Lastly, for some sub-analyses the sample size was low.