Study population and setting
This retrospective study considered serial laboratory measurements from patients who had received PCR testing for SARS-CoV-2 infection at Mayo Clinic hospitals in Minnesota, Arizona, and Florida between February 15 to May 28, 2020. Analysis was restricted to patients who had at least three serial observations from one of 194 laboratory tests. The study included 246 patients with PCR confirmed SARS-CoV-2 infection (mean age 61 years, 56% male) and 13,666 patients who tested negative (mean age 64 years, 52% male), 2,460 of whom were selected as controls via propensity matching (using age, gender, race, anticoagulation/antiplatelet medication use, history of coagulopathies, and hospitalization status prior to PCR testing). Differences in laboratory measurements of coagulation-related markers were assessed between the two groups at pre-specified time intervals centered around the date of PCR testing. The broader cohort of 2,232 patients who tested positive for SARS-CoV-2 was used for a supplementary analysis of thrombotic events: a neural network was used to classify clinical “sentiment” from electronic medical records regarding diagnosis of coagulopathies as “yes” (confirmed diagnosis), “no” (diagnosis ruled out), “maybe,” and “other” (alternate context, e.g., family history of disease).
Summary of Main Findings
Across nine windows of time defined relative to the date of PCR testing, there were 130 comparisons (66 unique laboratory values) that significantly differed between those testing positive and negative for SARS-CoV-2 infection (at a Cohen’s D > 0.35 and a Benjamini-Hochberg-adjusted p-value < 0.05). Relative to controls with available measurements, mean plasma fibrinogen levels were higher in patients with SARS-CoV-2 infection at the time of PCR testing (529 vs. 361 mg/dL, p<0.01); this difference waned and resolved over the next 7 days. Mean pre-diagnosis platelet counts were lower in the positive group relative to controls (185 vs. 226 x 109/L, with thrombocytopenia observed in 29% of those with SARS-CoV-2 infection vs. 21% without) but increased over the next 10 days to levels that were significantly higher than those in controls. In the full cohort of 2,232 patients with a positive SARS-CoV-2 result, 101 (4.5%) experienced a clinically diagnosed thrombotic phenotype within 30 days after the PCR test; 53 of these patients had a deep vein thrombosis. The main analytic cohort of 246 SARS-CoV-2-positive patients with serial laboratory measurements contained 76 of the 101 patients with thrombosis. Among this group, platelet counts at SARS-CoV-2 diagnosis were not associated with subsequent development of thrombosis, and the degree of mean platelet increases over the next 10 days was similarly unassociated with thrombosis. Five of the 2,232 (0.2%) in the full cohort of SARS-CoV-2-positive patients exhibited symptoms consistent with disseminated intravascular coagulation (DIC).
This study employed serial measurements of coagulation markers in patients infected with SARS-CoV-2, and compared trajectories to those observed in propensity-matched controls who were similar on average with respect to several determinants of coagulopathy.
The sample of patients testing positive for SARS-CoV-2 with serial coagulation marker measurements was small, and most were white (63%), making it difficult to generalize to the broader population of COVID-19 patients. Those with serial measurements available are likely different from the wider population; for example, they probably represent more severe disease and are at higher risk for thrombotic events. The occurrence of thrombotic events among the control group of patients testing negative for SARS-CoV-2 was not reported, preventing any inference about the role of SARS-CoV-2 infection in thrombosis. The timing of PCR testing may have varied considerably with respect to the date of infection or symptom onset, which could have muddied the observed trends and comparisons. Finally, there may have been additional unmeasured determinants of serial testing that were related to the probability of experiencing coagulopathy; if so, results could be biased in an unpredictable manner. Analysis of laboratory markers was limited to individuals with available measurements, which were often a small subset of cases and controls. This selection likely undermines any benefit of propensity-score matching for control of confounding. Controls were selected at the time of their first negative SARS-CoV-2 test, as long as they did not have a positive test later, whereas cases were selected at the time of their first positive test, regardless of how many negative tests they had previously. This means that individuals with more testing for COVID-19 would be more likely to be selected as cases.
This study is one of the few published longitudinal analyses of coagulopathy in COVID-19 that also considered a control group of comparable patients who tested negative for SARS-CoV-2.
This review was posted on: 22 January 2021