rando2
commented
4 years ago Questions to answer about each paper:
Please provide 1-2 sentences introducing the study and its main findings
Over 6000 hospitalized COVID_19 patients in the UK were split into 2:1 trial arms of standard care vs dexamethasone (a steroid), some of whom were ventilated, on non-invasive oxygen, or breathing independently. Analysis of mortality, discharge, and ventilation over a 28-day period suggests that dexamethasone reduced mortality in patients who were ventilated at the start of the study (by 35%) or on oxygen at the start of the study (by 20%). It also may reduce progression to ventilation in patients who were on oxygen at the start of the study.
Study question(s) being investigated:
Does dexamethasone reduce mortality (or time to/probability of discharge or progression to ventilation) in hospitalized UK COVID-19 patients compared to standard care?
How many/what drugs/combinations are being considered?
Dexamethasone, oral or intravenous administration, 6 mg per day for up to 10 days Some patients also took azithromyacin
What are the main hypotheses being tested?
Does dexamethasone reduce mortality in hospitalized UK COVID-19 patients compared to standard care?
Study population:
What is the model system (e.g., human study, animal model, cell line study)?
human patients
What is the sample size? If multiple groups are considered, give sample size for each group (including controls).
- Standard care: 4321
- Dexa: 2104
This corresponded to 9355 out of 11320 patients enrolled. The rest of the patients didn't meet the inclusion criteria and were randomized to a different treatment arm.
For human studies:
What countries/regions are considered?
United Kingdom, 176 NHS sites Apparently it's evaluating about 15% of all UK Covid-19 patients!
What is the age range, gender, other relevant characteristics?
Age: mean is 66.1 years Sex: 36% female Laboratory confirmation of SARS-CoV-2: 82% (9% results outstanding, 9% not tested)
Comorbidities:
- Diabetes: 24%
- Heart disease: 27%
- Chronic Lung Disease: 21%
- At least one major comorbidity = 56%
Oxygen supplementation at randomization:
- 16% invasive mechanical ventilation
- 60% oxygen only
- 24% neither
What is the setting of the study (random sample of school children, inpatient, outpatient, etc)?
inpatient
What other specific inclusion-exclusion criteria are considered?
Dexa could not be specifically indicated or counterindicated for patients Dexa had to be available in the hospital they were in
Treatment assignment:
How are treatments assigned?
randomized by web service that took baseline information -- interventional
Is the study randomized?
Yes
Provide other relevant details about the design.
This includes possible treatment stratification (e.g., within litters for animal studies, within hospitals for human studies), possible confounding variables (e.g., having a large age range of individuals), possible risks of bias and how they are addressed (e.g., is there masking in a clinical trial? how are individuals chosen in an observational study?).
- Analysis stratified based on pre-specified characteristics at baseline: "age, sex, level of respiratory support, days since symptom onset, and predicted 28-day mortality risk."
- Very few patients received other potential COVID-19 treatments (e.g., remdesivir) and comparable percentages between the two conditions received azithromycin
- open label
Outcome Assessment:
Describe the outcome that is assessed and whether it is appropriate.
For example: Is the outcome assessed by a clinician or is it self-reported? Is the outcome based on viral load or a functional measurement (e.g., respiratory function, discharge from hospital)? What method is used to measure the outcome? How long after a treatment is the outcome measured?
Primary outcome: all-cause mortality within 28 days Secondary outcomes: time to discharge, receipt of mechanical ventilation (for those not ventilated at start), death Also tracked: cause of death, renal hemodialysis or hemofiltration, cardiac arrhythmia, and receipt/duration of ventilation
Endpoint is: discharge, death, or status at 28 days after enrollment in trial Information recorded: "adherence to allocated study treatment, receipt of other study treatments, duration of admission, receipt of respiratory or renal support, and vital status (including cause of death)" Collected via "web-based case report" which suggests physician recorded
Are outcome measurements complete?
For example, are there individuals lost to follow up?
- 4.8% of individuals were not at 28 days by the time the study ended (10 June)
- 95% of patients had complete info
- 95% of those in dexa group received at least one dose
- 7% in usual care received dexa
Are outcome measurements subject to various kinds of bias? For example, a lack of masking in randomized clinical trials.
No placebo
Statistical Methods Assessment:
What methods are used for inference?
hazard ratio from Cox regression also used log-binomial regression model to estimate the risk ratio for mechanical ventilation chi-square to compare trends among subgroup categories (age, sex, disease severity) p-values are all 2-sided
Are the methods appropriate for the study? For example, are clustered data treated independently or are clusters adjusted for, such as different hospitals or litters?
Not my area of expertise but it seems like the methods were prespecified prior to data collection and they seem consistent with other clinical trials I've read. Intention-to-treat principle used. Enrollment numbers were determined based on power estimates by external (blinded) advisory board.
Are adjustments made for possible confounders? For example, adjustment for age, sex, or comorbidities.
Due to chance, 1.1 year age difference between groups. They added methods to control for it.
Results Summary:
What is the estimated association?
For example, is it an estimated odds ratio, a median difference in detected cases, etc?
- 28 day mortality is lower in dexa group than SC group (21.6% vs 24.6%, p <0.001)
What measures of confidence or statistical significance are provided?
For example, confidence intervals, p-values, and/or Bayes factors.
- confidence intervals and p-values
Interpretation of results for study population:
Can we make a causal interpretation for the individuals in the study of drug -> outcome, such as "taking drug A improves likelihood of survival twofold over taking drug B."
For example, with a well-performed animal study or randomized trial it is often possible to infer causality. If is an observational study, does it match up with some of the Bradford Hill criteria? https://www.edwardtufte.com/tufte/hill https://en.wikipedia.org/wiki/Bradford_Hill_criteria
It seems like patients on oxygen or mechanical ventilation benefit from dexamethadone administration. The lack of placebo is something of a limiting factor but the fact that they didn't find an association in patients who were not on oxygen supplementation makes it seem like this might not have had a huge impact on outcome.
Are there identified side effects or interactions with other drugs?
For example, is the treatment known to cause liver damage or to not be prescribed for individuals with certain comorbities?
I'm not able to answer this question fully. It is an immunosuppressant so it seems like administrating it later in the course of the disease to combat the cytokine storm is preferable, rather than while the immune system is helping fight off the virus. I don't know enough about steroids to know what other concerns would exist. As the authors note, there is a concern that it might slow clearance of the virus. There was, as the authors note, the possibility of harm in patients who were not in need of oxygen supplementation (which makes sense because you would not want to give an immunosuppressant prior to immunopathy).
Are there specific subgroups with different findings?
For example, do individuals with a specific baseline seem to do particularly well? Be particularly cautious with respect to multiple testing here.
- More severe respiratory support at randomization was associated with a greater effect of dexa -- 28-day mortality was reduced by 35% in patients with mechanical intervention at randomization compared to 20% in patients receiving oxygen without invasive ventilation vs no benefit in patients who were not receiving oxygen or ventilation (with and without age adjustment)
- longer duration of symptoms was also associated with a greater benefit of dexa on mortality
- secondary outcomes showed similar patterns, with dexa reducing time to and probability of discharge, with the largest effects in the group receiving mechanical ventilation at randomization
- for the patients who were not on mechanical ventilation at baseline, dexa decreased the probability of proceeding to mechanical ventilation (although this was only just significant in overall (p=0.049), but seemed more robust in the oxygen-receiving group)
Extrapolation of conclusions to other groups of individuals not specifically included in the study:
If the study is an animal study, which animal and how relevant is that model?
Is the model system appropriate? Is there evidence from past use that it's highly-relevant to therapeutic design in this context?
If it is a human study, what characteristics of the study population may support/limit extrapolation?
-
Can results extrapolate easily to other similar groups? (e.g., same country, similar age groups)
-
What would happen if conditions are extended in terms of dose or duration?
-
Can results be extrapolated to other populations or in very different settings? (e.g., different age group, primary care setting vs emergency department etc)
-
They do not provide a demographic breakdown, which to go on a personal rant I think these types of studies should because corticosteroids are closely linked to the stress response, and therefore trauma, and that would absolutely intersect with the exacerbated effects of COVID-19 on minoritized populations, but given that this study analyzed 15% of the UK's COVID-19 patients, it would seem likely that they sampled a diverse group.
-
The age groups defined in Table 1 seem like they are not fine enough to depict the variance of the sample, since the average was 66 and the table breaks groups in < 70, 70- 80, and 80+
Summary of reliability
1-2 sentences on concluding remarks, including summary of strengths, weaknesses, limitations.
This study has no glaring statistical issues, and was randomized. Obviously placebo-controlled is preferable, but overall these results seem reasonable. I would be interested to hear what someone with more experience in clinical trials thinks.
dziakj1
Title: Effect of Dexamethasone in Hospitalized Patients with COVID-19: Preliminary Report
Please paste a link to the paper or a citation here:
Link: https://doi.org/10.1101/2020.06.22.20137273
What is the paper's Manubot-style citation?
Citation: doi:10.1101/2020.06.22.20137273
Please list some keywords (3-10) that help identify the relevance of this paper to COVID-19
Please note the publication / review status
Which areas of expertise are particularly relevant to the paper?
Questions to answer about each paper:
Please provide 1-2 sentences introducing the study and its main findings
Study question(s) being investigated:
How many/what drugs/combinations are being considered?
What are the main hypotheses being tested?
Study population:
What is the model system (e.g., human study, animal model, cell line study)?
What is the sample size? If multiple groups are considered, give sample size for each group (including controls).
For human studies:
What countries/regions are considered?
What is the age range, gender, other relevant characteristics?
What is the setting of the study (random sample of school children, inpatient, outpatient, etc)?
What other specific inclusion-exclusion criteria are considered?
For example, do the investigators exclude patients with diagnosed neoplasms or patients over/under a certain age?
Treatment assignment:
How are treatments assigned?
For example, is it an interventional or an observational study?
Is the study randomized?
A study can be interventional but not randomized (e.g., a phase I or II clinical trial is interventional but often not randomized).
Provide other relevant details about the design.
This includes possible treatment stratification (e.g., within litters for animal studies, within hospitals for human studies), possible confounding variables (e.g., having a large age range of individuals), possible risks of bias and how they are addressed (e.g., is there masking in a clinical trial? how are individuals chosen in an observational study?).
Outcome Assessment:
Describe the outcome that is assessed and whether it is appropriate.
For example: Is the outcome assessed by a clinician or is it self-reported? Is the outcome based on viral load or a functional measurement (e.g., respiratory function, discharge from hospital)? What method is used to measure the outcome? How long after a treatment is the outcome measured?
Are outcome measurements complete?
For example, are there individuals lost to follow up?
Are outcome measurements subject to various kinds of bias?
For example, a lack of masking in randomized clinical trials.
Statistical Methods Assessment:
What methods are used for inference?
For example, logistic regression, nonparametric methods.
Are the methods appropriate for the study?
For example, are clustered data treated independently or are clusters adjusted for, such as different hospitals or litters?
Are adjustments made for possible confounders?
For example, adjustment for age, sex, or comorbidities.
Results Summary:
What is the estimated association?
For example, is it an estimated odds ratio, a median difference in detected cases, etc?
What measures of confidence or statistical significance are provided?
For example, confidence intervals, p-values, and/or Bayes factors.
Interpretation of results for study population:
Can we make a causal interpretation for the individuals in the study of drug -> outcome, such as "taking drug A improves likelihood of survival twofold over taking drug B."
For example, with a well-performed animal study or randomized trial it is often possible to infer causality. If is an observational study, does it match up with some of the Bradford Hill criteria? https://www.edwardtufte.com/tufte/hill https://en.wikipedia.org/wiki/Bradford_Hill_criteria
Are there identified side effects or interactions with other drugs?
For example, is the treatment known to cause liver damage or to not be prescribed for individuals with certain comorbities?
Are there specific subgroups with different findings?
For example, do individuals with a specific baseline seem to do particularly well? Be particularly cautious with respect to multiple testing here.
Extrapolation of conclusions to other groups of individuals not specifically included in the study:
If the study is an animal study, which animal and how relevant is that model?
Is the model system appropriate? Is there evidence from past use that it's highly-relevant to therapeutic design in this context?
If it is a human study, what characteristics of the study population may support/limit extrapolation?
Summary of reliability
1-2 sentences on concluding remarks, including summary of strengths, weaknesses, limitations.
Progress
Check off the components as they are completed. If the component is not applicable, check the box as well.