USA: The RECOVER Initiative - Long Covid research

Reminder that this starts in 1.5 hours.

 

They presented findings from two large observational studies of Paxlovid use during acute COVID, using a "Target Trial Emulation" methodology, which is an attempt to emulate an RCT as much as possible using observational data. They use statistical methods such as "Inverse Probability of Treatment Weighting" to control for as many possible confounders as possible.

PCORnet (see thread for study)
For the PCORnet study, they studied 497,499 patients. Paxlovid treatment was required to be within 5 days of infection.

159,659 high risk received treatment
5,597 low risk received treatment
272,048 high risk did not receive treatment
35,874 low risk did not receive treatment
​

When looking at incidence of long COVID over 180 days after infection, they found that about 3 fewer people developed long COVID per 100 people that were infected when taking Paxlovid. The hazard ratio was 0.88. (12% less likely to develop long COVID). Also found decreased incidence of hospitalization greater than 30 days after infection and decreased mortality with Paxlovid use.

They looked at the effect of Paxlovid on specific long COVID symptoms in various categories (neurologic, skin, pulmonary, etc). The hazard ratio is significantly less than 1 in almost every category (Paxlovid associated with decreased risk of LC). The two exceptions that were not significant were associations with hair loss and with receiving the long COVID diagnostic code U09.9.

They also looked separately at high risk and low risk patients. High risk means older age (≥ 50), smoker, or underlying conditions such as cancer or diabetes. There was a significant association of Paxlovid with decreased LC in high risk, but not low risk patients. When splitting by factors such as specific condition or age in the high risk population, the hazard ratios are still almost consistently significant when grouping based on these factors.

N3C (see thread for study)
This study on the N3C cohort also found decreased risk of long COVID with Paxlovid use. 1 fewer person per 366 infected when treated with Paxlovid [about 0.27 per 100 when comparing to 2.99 per 100 in PCORnet cohort].

There was a lower risk of long COVID when looking at the overall definition (RR=0.94, 95% CI: 0.92-0.97). When splitting by symptom clusters, there was significant decrease for cognitive phenotype (RR=0.86, 95% CI: 0.81-0.91) and fatigue/malaise phenotype (RR=0.92, 95% CI: 0.88-0.96), but not respiratory phenotype (RR=1.03, 95% CI: 0.97-1.09).

Also corroborated previously published findings about Paxlovid reducing risk of hospitalization and death.

-----

The findings were fairly close for relative risk between the cohorts showing a modest reduction in risk of long COVID in those who take Paxlovid during acute COVID infection, specifically in high risk patients.

The differences in absolute risk reduction were likely due to the different definitions used for long COVID. About 30% of people developed LC using rule-based PCORnet definition, versus only about 5% using machine learning-based N3C definition.

Edit: Fixed typo in relative risk value for fatigue/malaise phenotype.

 

Nature: "Identifying commonalities and differences between EHR representations of PASC and ME/CFS in the RECOVER EHR cohort"

 

Recording available

4/14/25, RECOVER Initiative: 'Effectiveness of Paxlovid in Protecting Against Long COVID: EHR Insights'

Speakers:
Dr. Fei Wang, Cornell University
Dr. Rainu Kaushal, Cornell University
Alexander (Sandy) Preiss, RTI
Dr. Abhishek Bhatia, University of North Carolina
Hannah Mandel, NYU Langone Health