Changing COVID-19 outcomes in patients with rheumatic disease—are we really getting better at this?

The COVID-19 pandemic has continued to impact the world, and we are now on track to exceed two million deaths worldwide. Patients with rheumatic disease were an immediate concern, but research to date has not convincingly suggested that having a rheumatic disease per se increases the risk of poor outcomes. Instead, poor outcomes seem to be driven by comorbidities and certain medications, such as chronic glucocorticoids and rituximab.1, 2

An important question to ask is whether we are getting better at treating COVID-19. In The Lancet Rheumatology, April Jorge and colleagues³ address this by exploring temporal trends in patients with rheumatic disease, comparing an early cohort (Jan 20 to April 19, 2020) with a late cohort (April 20 to July 19, 2020). They used a large network of hospitals and health systems across the USA from the TriNetX database with over 8500 patients with rheumatic disease, and completed both an unmatched and matched analysis to try to reduce confounding factors. They found that the risk of hospitalisation for COVID-19 decreased in the late cohort compared with the early cohort (874 [32·4%] of 2701 patients vs 1227 [45·4%] of 2701 patients; relative risk [RR] 0·71, 95% CI 0·67–0·76). Outcomes such as intensive care unit admission, mechanical ventilation, kidney injury, and death were also reduced in the late cohort compared with the early cohort. Among those patients that were hospitalised for COVID-19, the risks of intensive care unit admission, mechanical ventilation, and death were lower in the later part of the pandemic compared with the earlier part of the pandemic (334 [30·7%] of 1089 patients vs 450 [41·3%] of 1089 patients; RR 0·74, 95% CI 0·67–0·83). These improvements mirror recent findings within the general population, with reductions in COVID-19 mortality during the later months of the pandemic compared with the first few months.⁴ The strengths of this study include its large sample size and detailed patient information, including comorbidities.³ The authors also did appropriate sensitivity analyses, such as restricting analyses to hospitalised patients and accounting for a washout period.

What are some of the potential explanations for these findings? First, artefacts should be considered, such as expanded testing capacity detecting milder cases in the later months of the pandemic. There is also the potential for factors that were not captured in the analysis, such as background glucocorticoid dose and rheumatic disease severity and activity being unbalanced between the early period and the later period, to bias the outcomes. Additionally, it has been shown that those infected in the later part of the pandemic had a different risk profile, leading to differing background risks of poor outcomes.⁵ Insufficient information about outcomes across facilities and geographies, which were not included in propensity score matching, raises the issue of health-care facility differences driven by resource availability or hospital overload. For example, hospitals that were overwhelmed at the start of the pandemic in April might have lowered their threshold for hospitalisation in June as a result of increased bed capacity. By not adjusting for such features, the findings might be explained by these epidemiological and health-care system factors.⁶

Therapeutics have also changed over the course of the COVID-19 pandemic, with treatments being used in the later cohort that were not routinely used in the earlier cohort. Agents such as remdesivir and glucocorticoids have become the standard of care in many health-care settings and, as Jorge and colleagues³ have pointed out, non-pharmacological treatment has also changed, including avoidance of mechanical ventilation in favour of non-invasive ventilation,⁷ altering ventilation strategies,⁸ prone positioning,⁹ and anti-coagulation treatment.¹⁰ It is unlikely that our acquired knowledge of specific risks in patients with rheumatic disease affected results, as comorbidities and therapies such as glucocorticoids and rituximab are difficult to alter in the short term, and in the absence of active infection the recommendations have been to not alter therapy. Therefore, clinical improvements in the treatment of COVID-19 might explain the differences in outcomes over time.

It is difficult to determine how much of the improvement seen in Jorge and colleagues’ study is due to clinical factors, such as therapies and practices, versus a selection bias of patients in earlier cohorts compared with later cohorts. A key finding of this study is that use of historical controls might overestimate the effect of treatments. As the pandemic evolves and we continue to measure patient outcomes, it will be important to appropriately account for changes over time and place in longitudinal studies. So, on one hand we must remain vigilant to consider the limitations of outcome studies published during the pandemic, but on the other hand, be hopeful that we are making progress.

© 2021 Shutterstock