A new study by an international team of researchers and published on the preprint server medRxiv* in June 2020 describes the possible reasons for inconsistent or null findings in clinical studies of antiviral drugs to date. It proposes a new approach to decide the sample size in light of its findings.
Why Are Effective Antivirals Not Being Found?
Against the background of the current COVID-19 pandemic, the need to develop an effective antiviral against the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has become an urgent priority for global health. Both new and existing (repurposed) antivirals are being tested for efficacy. While some FDA-approved drugs have shown good results both in vivo and in vitro, clinical studies have failed to show correspondingly convincing efficacy.
Novel Coronavirus SARS-CoV-2 This transmission electron microscope image shows SARS-CoV-2, the virus that causes COVID-19, isolated from a patient in the U.S. Virus particles are shown emerging from the surface of cells cultured in the lab. The spikes on the outer edge of the virus particles give coronaviruses their name, crown-like. Image captured and colorized at NIAID's Rocky Mountain Laboratories (RML) in Hamilton, Montana. Credit: NIAID
The reasons for this apparent loss of consistency could include the trial design, which is sometimes not rigorous enough to capture the true effectiveness of the antiviral, and the setting in which the repurposed drug is used, such as a compassionate use program. Study designs typically require much preparation in the form of deciding the dosage, the clinical outcomes, required sample size, and safety assessments – all of which may not have been critically assessed in the background of the urgent demand for an effective drug.
Compassionate use programs are by default observational studies, and the use of the antiviral is decided by the doctor along with the patient or next of kin of the patient. This discretionary design allows for a host of confounding factors, including the stage of the disease, underlying or accompanying comorbidities, and dosage used. These could lead to inaccurate conclusions even when observable confounding factors are accounted for.
Modeling Changes in Viral Load
The current paper uses mathematical modeling to describe the viral dynamics within the host and applies it to current clinical studies to demonstrate its utility. The quantitative approach used by the researchers shows that there are at least two factors that can confound the true effectiveness of the antivirals.
The design of the study was aimed at exploring the quantitative differences in the way the virus behaves in different patients and the reason for these variations. To do this, the researchers made use of data on the infection from 38 patients from different countries, over time, to estimate the viral load since the earliest symptoms.
By comparing the viral load between patients, the differences between them were visualized at statistically significant levels, at two levels: the initial viral load at the onset of symptoms and the rate at which cells containing replicating virus died per day (speed of viral load decay). This led to the formation of three groups, namely, rapid, medium, and slow viral load decay.
Good immune responses would lead to rapid viral load decay and a shorter period of virus production, which is clinically observed as a shorter period of detection of viral particles in respiratory samples. This has been associated with less severe disease.
The modeling findings led to in silico experiments to find out how the virus dynamics would change when drugs blocking virus replication are administered. From this, the therapeutic possibilities are determined.
How The Timing of Initiation Affects Outcome
Clinical outcomes are typically related to the timing of initiation of the antiviral treatment, especially for influenza. The dose and level of immunity are critical factors deciding the size and direction of antiviral effects.
The researchers, therefore, simulated several scenarios with varying times of initiation, from 0.5 to 5 days following symptom onset, inhibition rates of 99% or 50%, and the three patient groups mentioned earlier in terms of viral load decay. The variation in days corresponds to the period before and after the peak viral load, as estimated by the current model.
The study shows that starting antiviral treatment early affects the rate of viral load decay favorably, whether the inhibition rate is high or low. On the other hand, beginning treatment once the peak has been reached is rather fruitless in all three groups and irrespective of the rate of inhibition.
The investigators also introduced the duration of virus shedding as one of the outcomes since this is often used in clinical studies to evaluate the efficacy of antiviral treatment for SARS-CoV-2. They found that as expected, the duration was shortest in the group with rapid viral load decay, and longest with slow decay.
How Should Sample Size Be Decided?
The researchers then focused on finding the appropriate sample size that would have 80% power to detect a difference in efficacy between controls and treatment groups in a randomized clinical trial, assuming initiation of antiviral therapy as soon as the patients were hospitalized. They found that the typical clinical trial in which patients are treated irrespective of the time of initiation of treatment requires a much larger sample size than is usually seen.
The average point of initiation is at about 4.6 days from the symptom onset after the peak viral load is attained. The time from the onset of symptoms to the initiation of treatment was then used to decide an inclusion criterion on which the new sample size was then based. This was much smaller – 98 vs. 2720 with the use of broad and strict (within 0.5 days of symptom onset) criteria, respectively.
If the area under the curve (AUC) is used as the outcome, the pattern is similar. Thus, the time of initiation and the rate of decay of the viral load must be taken into account while assessing clinical efficacy. Most studies calculate the required sample size in terms of the distribution of the outcomes in the treatment and control groups, in general. In the case of the SARS-CoV-2, this is not valid due to the non-linear association of the treatment effect with the timing of initiation of the antiviral.
Examples, Limitations and Future Directions
For instance, hydroxychloroquine was reported to be effective against the virus in one study, but other researchers could not replicate this. The lopinavir-ritonavir combination was also not associated with consistent efficacy, but in most cases, the drugs were begun more than ten days from symptom onset.
The study is limited by its failure to fully replicate physiological differences between target cells, which could affect the susceptibility of the cell to infection and by the lack of modeling of immunomodulatory effects of the drugs used. However, it passes on an important message, according to the researchers: The failure to identify effective antivirals against the virus “might not be because the antivirals are not effective, but because of the imperfect design of the clinical studies. The timing of treatment initiation and virus dynamics should be accounted for in the study design (i.e., sample size and inclusion-exclusion criteria) to identify effective antivirals.”
medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.