Randomized Controlled Trials
Definition: This is an experiment that has a test group and a control group. The word 'randomized' in the title means that all study participants have and equal chance of being in either the experimental group or the control group. The word 'controlled' means that there is a control group to compare against.Why Randomize?
We randomize so that the experimental and the study groups are similar. If we don't randomize, we may end up having an imbalance in the groups. The experimental group may have healthier patients than the control group or vice versa.
Let's See An Example
In this NEJM study, hospitalized COVID-19 patients were randomized to either receive steroids (dexamethasone), or the usual treatment
This is how the groups ended up after randomization. Ignore the total number in each group just now but just see the percentages, the numbers inside the parentheses. They match up between groups which is what we want. Note that this is before the experiment was done.
| Characteristic | Dexamethasone Group (N = 2104) |
Usual Care (N = 4321) |
|---|---|---|
| Age | ||
| Mean - yr. | 66.9±15.4 | 62.3 ± 9.2 |
| < 70 | 1141 (54) | 2505 (58) |
| 70 - 79 | 469 (22) | 859 (20) |
| > 80 | 494 (23) | 957 (22) |
| Sex - no (%) | ||
| Male | 1338 (64) | 2749 (64) |
| Female | 766 (36) | 1572 (36) |
| Race or Ethnic Group - no (%) | ||
| White | 1550 (74) | 3139 (73) |
| Black, Asian, or minority ethnic group | 364 (17) | 783 (18) |
| Median no. of days since symptom onset (IQR) | 8 (5–13) | 9 (5–13) |
| Median no. of days since hospitalization (IQR) | 2 (1-5) | 2 (1-5) |
| Respiratory support received — no. (%) | ||
| No Oxygen | 501 (24) | 1034 (24) |
| Oxygen only | 1279 (61) | 2604 (60) |
| Invasive Mechanical Ventilation | 324 (15) | 683 (16) |
All RCTs publish this baseline table, we should carefully look at it. Sometimes studies leave out important baseline characteristics. Your domain expertise should be able to catch that.
The difference in that baseline parameter that was left out might have influenced the results of the study.
What Was The Population Studied?
This is an important question to ask. We need to look at the demographics.
Was the study done only on males that it cannot be generalized to females?
Is the mean age of the population is much smaller than what I would encounter in my practice?
Are they of a different race or ethnic group?
What Was the Outcome of Interest?
The primary outcome that was being measured was the all-cause mortality within 28 days.
This seems like a reasonable outcome in this study as it is meaningful to patients.
Sometimes you may encounter studies that have a radiological outcome.
For example, the study may be trying to see if a tumor shrinks in size with treatment. But this may not be a patient-centered outcome as it may not be clear that tumor shrinkage corrrelates with better survival. It may sound plausible that shrinkage correlates with better survival but when these patients are followed up, this may not be the case.
This is like a start-up that is trying to improve their Search Engine Optimization metrics when it may not be clear that improving these metrics might make them more profitable, although a study could well prove that improvement of these metrics correlates with profits. But until this study is done, this is still uncertain.
What Did The Study Find?
Mortality at 28 days was significantly lower in the dexamethasone group than in the usual care group, with deaths reported in 482 of 2104 patients (22.9%) and in 1110 of 4321 patients (25.7%)
An astute reader of the study may try to question this and you should. Can the difference in outcomes between the two groups be explained by something else?
This is when you should look at the baseline table that I showed above. Have the study authors deliberately not explained (and this happens) an importan baseline variable?
This was a Practice-Changing Study
Dexamethasone is a cheap drug and is widely available. This study proved that sick Covid-19 patients are less likely to die while taking this drug. Many doctors and hospitals have now adopted this drug.
What is the Funding Source of the Study?
A profit-based funding might bring some bias in the study. If the study was conducted by the same body that sells the product/medicine, it might be a huge bias.
Don't Just Read the Abstract
Don't just read the extract. Abstracts are like PR for the study. They are mostly click-bait. Time and again authors leave out important nuance in the study. Nuances that are big enough to invalidate the study. Also reading the abstract might bias your view so just skip the abstract and read the details.