1. Introduction
A new pandemic respiratory infection caused by the coronavirus disease 2019 (COVID-19), the first cases of which were reported in Wuhan, China [
1], killing approximately 0.5% of confirmed case mostly by acute respiratory failure [
2]. On 15 August 2020, the French government reported that 30,406 patients died from COVID-19 on the national territory [
3]. During the epidemic’s first months, several treatments were rapidly recommended, mainly as symptomatic treatments. Numerous anti-infectious treatments were rapidly under investigation [
4], including remdesivir [
4], lopinavir/ritonavir [
5,
6], or hydroxychloroquine [
7]. Angiotensin-converting enzyme inhibitors during COVID-19 infection were also discussed [
8,
9].
Coffee contains 1,3,7-trimethylxanthine (TMX), which is thus the most commonly used psychoactive drug in the world. Recent studies suggested an efficacy of 1,3,7-trimethylxanthine (TMX) on the tolerance to central hypovolemia [
9] and the prevention of hypotension-related syncope [
10]. In a study with 13 patients, the exposure to lower body negative pressure led to a decrease of blood pressure that was significantly inferior in participants who took caffeinated coffee. Syncopes are typically caused by a decrease in blood flow, generally from low blood pressure. Consequently, improved blood pressure control allows a better perfusion of the brain and a reduced risk of syncope.
Due to the interest in drug repurposing as an avenue for research in COVID-19 crisis [
11] and the large number of clinical studies aiming to evaluate treatments using inappropriate design and methods, we conducted a pedagogic comparative study aiming to evaluate how the lack of proper methodology can suggest the efficacy of coffee (1,3,7-trimethylxanthine) for the treatment of patients with COVID-19.
4. Discussion
Patients receiving the TMX treatment had a significantly shorter hospital stay (9.5 vs. 15 days, p < 0.05). Moreover, TMX decreased the use of antibiotics and the extension of pulmonary lesions in the CT-scans. These results lead to the conclusion that the 1,3,7-Trimethylxanthine (TMX), the active molecule of the patients’ morning coffee beverage, is a potential treatment for COVID-19 infection.
In fact, we cannot conclude to a direct effect of TMX on COVID-19 because of a huge lack of methodological considerations [
15]. The aim of this study was not to prove the efficacy of coffee but to reveal how anyone can prove anything with a non-rigorous study. For instance, the multiple bias, the selection criteria, the clinical relevance of the endpoints, and the methodological structure of our study weakened the external and internal validity.
First, the lack of accurate selection and exclusion criteria led to the production of two incomparable groups. Even if those two groups seem to be similar according to demographic characteristics (
Table 1), the TMX group was clearly advantaged by its construction. Patients who could not orally take TMX were considered only in the control group and so were all the patients who needed mechanical ventilation. The same trick was used to move patients treated with corticosteroids from TMX group to the control group. This has been observed in several published articles evaluating candidate drugs for the treatment of patients with COVID-19 [
16]. Confounding factors must not be overlooked as they are strongly associated with both exposure and outcomes.
Second, a constant lack in observational studies is the indication bias [
17]. Why does a patient receive this treatment instead of another? A statistical pathway, such as a propensity score, can be discussed even if it is not a magical solution and has many limitations, such as it does not take into consideration unnotified factors [
17]. Indication bias can be avoided with the use of randomization.
Third, the primary outcome defined in the methods as the mortality at day six after hospital admission is not reported in the results. Indeed, mortality was not different between the control and the treatment group (6% vs. 7.7%,
p = 0.671,
Table 2) but the results only described secondary outcomes. Focusing on secondary outcomes is frequent in controlled trials and is a type of spin defined as “exaggerating the clinical significance of a particular treatment without the statistics to back it up”. Spin has been widely described in biomedical literature and is a questionable research practice [
18,
19,
20]. Moreover, there was no estimation of the number needed to treat, the power, or the expected effect of the treatment described in the methods. Most of the recommendations of the STROBE guidelines for reporting observational studies were not followed in the current study [
21].
Fourth, temporality is one of the keys to determining a causality between exposure (or treatment) and outcome. To conclude that TMX decreased the extension of lung lesions is inappropriate and does not fit with Bradford Hill criteria for causality [
22].
Finally, the multivariable analysis suggested that TMX is associated with a higher odd to be hospitalized for less than seven days. Adjustment with actual known prognostic factors (age, obesity, and gender) [
18,
23] does not mean causality. For instance, selection of the covariates for an appropriate analysis of causal relationship can be done with a directed graph [
19]. This approach can eventually reduce the degree of the bias in the estimated model [
20]. In this study, including initial clinical status would change all the results. We should stay cautious with the use of statistical models and the results as the knowledge in this pandemic evolves day after day, particularly with small sample sizes. The strategy of analysis depends on the conception of the study and unfortunately, without a relevant methodology, prognosis and causality can be confounded [
21].
Well-conducted studies are important to make reliable conclusions. Limitations of the studies must be considered from their conception to their conclusion.
Controlled randomized clinical trials are considered as the gold standard to evaluate treatment effects [
24]. To control biases, randomized trials provide a fair allocation of both known and unknown risk factors to each group. Both blinded follow-up and outcome evaluation sustain comparability between the groups [
25]. As a result, in a randomized trial with intention-to-treat analyses, conclusions on outcomes can be assigned to the causal effect of the treatment [
26]. Well-conducted observational studies can be useful to describe a scientific idea but must be considered with cautiousness as they can easily be biased [
27].
Unfortunately, the COVID-19 pandemic crisis situation conducted to an increasing use of methodological-lacking observational studies or poorly-designed clinical trials to evaluate the effects of treatment [
28,
29]. Emergency statements have conducted to a significant belief bias as “people require less evidence to endorse a syllogism as valid when it has a believable conclusion” [
30]. However, no treatment of COVID-19 has been found and so grows the desperation to find a “magic pill” [
31,
32,
33,
34,
35,
36,
37].
The fundamental issue is to identify a factual clinical benefit in the general population.
We cannot conclude of any association between coffee or TMX and COVID-19. We hope that nobody will use caffeine without medical surveillance and more generally, nobody will use caffeine against coronavirus. Overconsumption of caffeine, as with the overconsumption of any drug or molecule, could result in adverse events [
38].
This study is not about a potential therapy in COVID-19 infection but a pedagogic study with real data from real patients treated for COVID-19 in two centers. We voluntarily conducted and analyzed this study to highlight that methodological-lacking studies can lead to overinterpreted data and give the wrong conclusions. We hope that an educational message is best conveyed through humor [
39].
