rando2
commented
4 years ago What did they analyze?
- All adult patients admitted to hospitals in Wuhan with confirmed COVID-19 who had reached a definite outcome (death or discharge) by January 31, 2020 (N=191). The first patients included were admitted on December 29, 2019. The two hospitals analyzed were the only hospitals in Wuhan handling COVID-19 cases until February 1, 2020, so this should be all hospitalized patients in the area.
- Assessed clinical course with respect to potential risk factors (e.g., comorbities)
- Examined viral shedding over period of hospitalization
What methods did they use?
Design of Study
- Retrospective cohort study using electronic health records
- Used multivariable regression to assess association between possible risk factors and outcome
Data Collection
- At intake, recorded several demographic characteristics and measured health status (e.g., pulse, blood pressure, temperature), symptoms (e.g., cough, fatigue, nausea) and biomarkers (e.g., lymphocyte count, platelet count, D-dimer, and many others).
- Tested for presence of SARS-CoV-2 virus in throat swab samples every other day using either next-generation sequencing or real-time PCR (RT-PCR) methods
- "Routine" blood examinations for: complete blood count, coagulation profile, serum biochemical tests (including renal & liver function, creatine kinase, lactate dehydrogenase, and electrolyes), myocardial enzymes, interleukin-6, serum ferritin, and procalcitonin (list copied verbatim from manuscript) (variable frequency)
- Chest radiographs or CT scans (variable frequency)
Does this paper study COVID-19, SARS-CoV-2, or a related disease and/or virus?
- COVID-19 as confirmed by presence of SARS-CoV-2 virus in throat swab sample
What is the main finding (or a few main takeaways)?
- Increased odds of death with increased age (per-year increase = 0.0043), higher Sequential Organ Failure Assessment score, and d-dimer score (higher than 1 µg/mL), per the multivariate model
- Other variables analyzed using univariate tests were significantly different between the survivor and non-survivor groups -- e.g., comorbitities except carcinoma, SOFA score, white blood cell count, lymphocyte count, lactate dehydrogenase, high-sensitivity cardiac tropnin I, d-dimer, IL-6, serum ferritin, procalcitonin, etc. -- but see concerns below about the statistical testing.
- Viral shedding continued until death in deceased patients. The median length of viral shedding was 20 days in survivors, with the longest observed period 37 days from onset of disease. I believe viral shedding was assessed based on the testing of throat swabs for the SARS-CoV-2 virus, and thus patients were not discharged until they tested negative twice at least 24 hours apart. However, this is not clearly defined.
What does this paper tell us about the background and/or diagnostics/therapeutics for COVID-19 / SARS-CoV-2?
- The most common symptoms at intake were cough and fever. Sputum production and fatigue were also common. The most common complications were sepsis, respiratory failure, ARDS, heart failure, and septic shock. Complications were associated with non-survival.
- Several of the biomarkers measured are linked to putative mechanisms of pathogenesis or pharmacological management, e.g., IL-6 and cytokine storm.
Do you have any concerns about methodology or the interpretation of these results beyond this analysis?
- Statistical testing: No correction for multiple testing was used, at least as far as methods indicate, despite the fact that they did many Mann-Whitney, chi-square, and Fisher exact tests. It seems we just have to trust that tests were used "as appropriate" (per their methods), because they don't seem to specify which test was used where (e.g., hopefully males and females split among survivors and non-survivors was analyzed with chi-square and not two separate Fisher exact tests, but it's not specified). They used a p-value of 0.05, which is very liberal given the number of tests that don't seem to have been otherwise corrected. They also don't discuss how having one group much larger than the other might have affected their results (seems worth noting whether the variances were comparable between groups?)
- Patients were defined as recovered if they received two negative tests 24 hours apart. However, this indicates that the SARS-CoV-2 tests were not always given 48 hours apart. There are several small issues like this where the methodology is not totally clear.
- Because of the small size of the deceased group, they say the model was actually only run using five variables based on "previous findings and clinical constraints" but it's not clear at what point in the analysis they selected/ranked which variables would be used (e.g., was this a post hoc decision?) The five variables were: lymphocyte count, d-dimer level (µg/mL), SOFA score, coronary heart disease status, and age. Other variables of interest (e.g., cytokine-related levels) were therefore excluded from the central analysis.
- Treatment was not included in the model! 95% of patients received antibiotics and 21% received the HIV antivirals lopinavir/ritonavir. Some patients received corticosteroids or intravenous immunoglobulin and administration of these drugs differed between survivors and non-survivors. No contingency table is provided for outcomes x treatment (beyond death vs discharge) nor is treatment factored into the analysis of biomarkers.
Any comments or notes?
- Measurements were taken at admission, but the patients could have been at different points in the disease course. They mention this in their discussion. Interestingly, the average across the survivor and non-survivor groups was exactly the same (11 days from onset of disease).
- Having EHR data from the first wave of COVID-19 hospitalizations provides a valuable dataset, and naturally these analyses were retrospective. However, I have serious reservations about the statistical analyses. The sample size may just be too small and treatments too non-standardized to test so many variables. "Significance" here should be interpreted cautiously.
dziakj1
Title: Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study
General Information
Please paste a link to the paper or a citation here:
Link: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)30566-3/fulltext
What is the paper's Manubot-style citation?
Citation: @doi:10/ggnxb3
Is this paper primarily relevant to Background or Pathogenesis?
Please list some keywords (3-10) that help identify the relevance of this paper to COVID-19
Which areas of expertise are particularly relevant to the paper?
Summary
Suggested questions to answer about each paper:
What did they analyze? They analysed a cohort of 191 adult inpatients at two hospitals in Wuhan, China, who had been found via laboratory testing to have COVID-19 and who either were discharged (137) or died in hospital (54) between December 29, 2019, and January 31, 2020.
What methods did they use? They treated this as a retrospective sample and used statistical methods (logistic regression and multivariable logistic regression, as well as the univariate tests Mann-Whitney U, chi-squared, and Fisher's exact tests) to study what factors predicted who would die and who would recover. They also estimated how many days a patient would continue to have a detectable viral load, and how many days the patient would continue shedding virus.
Does this paper study COVID-19, SARS-CoV-2, or a related disease and/or virus? It studies COVID-19 caused by SARS-CoV-2.
What is the main finding (or a few main takeaways)? Predictors for death included older age, higher Sequential Organ Failure Assessment ( score, and higher levels of d-dimer. Survivors continued to shed the virus for a median of 20 days and a maximum of at least 37 days, which the authors say is relevant to isolation of survivors.
What does this paper tell us about the background and/or diagnostics/therapeutics for COVID-19 / SARS-CoV-2? This paper reports findings which have partly become well-known, but the possible prognosis usefulness of d-dimer and some other biomarkers such as lactate dehydrogenase and cardiac troponin I may be interesting, as is the long duration of viral shedding. They noted that the relationship of mortality rate to age had also been found in the lab in macaques.
Do you have any concerns about methodology or the interpretation of these results beyond this analysis? Of course, these are very early results (the first two months or so in which the disease was known to science), which is both their appeal and their limitation. They also noted that some tests had some missing data for early patients, because not all were initially being done. Also, not all patients received the same medical treatment or adhered to it equally well.
Any comments or notes?