Critique of hydroxychloroquine studies to date

Written by Associate Professor Steven Tong, ASCOT Coordinating Principal Investigator

There has been a lot of controversy about hydroxychloroquine (HCQ) as a therapeutic intervention for COVID-19. Currently, HCQ is one of the interventions being trialled in ASCOT. We will be using a starting dose that is higher than is typically used for other conditions. Given the recently available studies suggesting little or no benefit of HCQ and potential harm we have received questions about:

Should we continue to include HCQ?

Should we be using a high dose of HCQ?

The two key types of studies of HCQ have been observational studies and randomised controlled trials. Observational studies ‘observe’ what has happened, and in our context are describing what happened to patients who received HCQ, often in comparison to patients who did not receive HCQ. Although we can learn some things from observational studies, the comparison between patients who do and don’t receive HCQ may not be valid. The main issue is that clinicians choose to prescribe a drug for various reasons, including because a patient may be more unwell. So the two groups may be quite different and would have different outcomes compared to each other regardless of whether they received HCQ.

The best way to remove these biases is to conduct a randomised controlled trial (RCT). In this case, patients would be randomly assigned to receive HCQ or not. The decision is taken out of the hands of the treating clinicians and the randomisation process means that patients should be similar between the two groups. Well designed and conducted RCTs are considered the best way to determine whether a drug such as HCQ is effective.

Three further things to keep in mind.

One, we should look out for benefits of HCQ – in particular, does it reduce the risk of dying or needing to go onto a ventilator – because these are the outcomes we care most about. We call these ‘clinically relevant’ outcomes.

Two, we should look out for harms of HCQ – does it cause serious adverse effects? And then we can balance the benefits against the harms. For example, if HCQ reduces the risk of death from 30% to 20%, even if it causes 5% more serious heart rhythm abnormalities, then the benefit outweighs the harms.

Three, the dose of HCQ may be important. Too low a dose may mean it has no effect, but too high a dose may cause more harm. There may be a middle ground which maximises the benefits and minimises the harms. We call this middle ground the “therapeutic window”.

With these points in mind, here is a brief recap of relevant papers.

Early on, Wang et al showed in the laboratory that chloroquine had antiviral activity against SARS-CoV-2 at a concentration that can be achieved in people. Notably, this letter to Cell Research has already been cited 46 times, accessed more than 1 million times, and tweeted over 9000 times according to the journal website.

Based on this, we made a decision to include hydroxychloroquine as one of the interventions to test in ASCOT. Hydroxychloroquine is very similar to chloroquine in chemical structure and was chosen over chloroquine due to ease of access in Australia.

Next, was the infamous publication from the Raoult group that claimed the benefit of HCQ, particularly when combined with azithromycin, in reducing the viral load. Azithromycin is a commonly used antibiotic that can also have an anti-inflammatory effect. This paper has been roundly criticised, not least for issues with selective reporting, small sample size as an observational uncontrolled study, and lack of clinically relevant outcomes. 

The Raoult group has now described their experience with 80 patients receiving HCQ plus azithromycin. Although this report is of better quality, it is an observational study with no control group and we cannot determine whether there is truly any benefit of the intervention over standard of care. Most of the included patients were of low severity and treatment was commenced early. Only one patient died. There were no concerning safety signals.

Across four other hospitals in France, HCQ looked less promising. Again in an observational study, among 181 patients of whom 84 received HCQ within 48 hours of admission and 97 who did not, there was no difference in death or need for intensive care admission. Eight of the patients receiving HCQ developed ECG abnormalities (an electrocardiogram or ECG measures the heart rhythm) resulting in cessation of the HCQ.

The dose used in the French studies was 600mg per day.

Then came a few small RCTs.

  1. Among 30 patients in Shanghai, China there was no difference in viral shedding or clinical outcomes between patients treated with and without HCQ 400mg per day. This study is clearly limited by the small sample size.

  2. Among 62 patients in Wuhan, China patients with mild illness, the 31 patients treated with HCQ 400mg per day had a shorter time to resolution of fever, remission of cough, and improvement of pulmonary imaging findings compared to 31 without HCQ. Major limitations of this study are the uncertain value of the primary endpoint of time to clinical recovery and the small sample size.

  3. In the largest RCT to date, 150 patients across 16 hospitals in China were randomised to HCQ 1200mg for three days followed by 800mg daily or to standard of care. This study is notable for the higher dose of HCQ used. There was no difference in the primary endpoint of viral clearance at day 28 and also no difference in viral clearance at any earlier time points. Clinically relevant outcomes were not reported.

Up to this point, the balance of admittedly poor quality data was either of no or some benefit of HCQ. Unfortunately the studies were limited in their design of being observational studies with no appropriate control group or being small randomised controlled trials. Most of these studies chose endpoints with very limited clinical relevance and were not adequately powered to even show a difference in these endpoints. We are unable to conclude whether there is a benefit with the use of HCQ. At least there was no signal of harm.

This changed with two further studies: one an observational study from the Veteran Affairs hospitals in the US, and another, a RCT from Brazil that was stopped early due to a safety concern.

The Brazilian RCT was published on 24 April, 2020 (Borba et al,). In this study 81 patients with suspected COVID-19 infection were randomised to either high dose chloroquine (600mg twice daily for 10 days) or low dose chloroquine (450mg twice daily on day 1, then 450mg once daily for 4 days). The high dose of chloroquine is a significantly larger dose than in other studies and may have led to increasing blood levels.

The study planned to enrol 440 patients with a primary endpoint of mortality at day 28 (a highly relevant clinical endpoint). At an unplanned interim analysis on 5 April, 13 days after enrolment had begun (23 March), the mortality in the high dose group (7/11 patients died) was noted to be higher than the low dose group (4/11 patients died). The data and safety monitoring board recommended stopping the high dose arm, and it appears that recruitment was closed on 5 April. By this time, 81 patients had been randomised, of whom 62 had confirmed COVID-19 infection.

Patients in the high dose group were older and more had underlying cardiac disease. All patients received azithromycin and the majority received oseltamivir (a drug used to treat influenza infections). At the time of publication, the reported mortality at day 13 was 39% in the high dose group (16 of 41 patients) and 15% (6 of 40 patients) in the low dose group.

Patients in both groups were found to have the important cardiac rhythm abnormality of a prolonged QTc > 500ms at day 13 (7 of 37 [19%] for high dose vs 4 of 36 [11%] for low dose). The QTc measures the time from the initial electrical activity of the heart ventricles contracting to expel blood into the circulation to the time the ventricles are ready for the next contraction. If the QTc becomes too prolonged, it can lead to ventricular tachycardia (a very serious heart rhythm abnormality that if not recognised and treated can lead to death). Two of 37 patients in the high dose group experienced ventricular tachycardia.

This was a difficult study to read through and I have had to infer the timing of a number of the events at critical time points. It suffers from a number of issues that cloud the conclusions and the generalisability:

  • The dose of chloroquine chosen was likely to be too high and is not being used elsewhere clinically nor in clinical trials at this dose. Therefore, the toxicity concerns may not be applicable with more appropriate dosing.
  • Almost all patients received azithromycin and oseltamivir, which can both prolong QTc. Therefore, the toxicity concerns may not be applicable to HCQ alone (even at the high dose).
  • Stopping the study after 13 days and with 11 deaths was very early. The primary endpoint of mortality at 28 days for even the first patient enrolled may not have been met. Unfortunately, this means the sample size is small and we cannot draw any robust conclusions based on this data. 

The VA study was an observational study of data from the US Veterans Health Administration medical centres. Of 368 evaluated patients, 97 were administered hydroxychloroquine alone, 113 hydroxychloroquine plus azithromycin, and 158 no hydroxychloroquine. There were 27 deaths (28%) in the HCQ group, 25 deaths (22%) in the HCQ + azithromycin group, and 18 deaths (11%) in the no HCQ group. Careful analysis revealed a higher risk of death in the HCQ group compared to the no HCQ group.

This was a robustly conducted study but still suffers from the retrospective observational design. Patients receiving HCQ may have been different from those who did not, in both measured and unmeasured parameters. For example, more patients in the HCQ group had low oxygen levels at baseline, were hypertensive, and had lower protein levels, all risks for worse outcomes. Although adjustments were made in the analysis, unmeasured parameters cannot be adjusted for. Other limitations were that only men were included in this dataset and there is no detail about the doses of hydroxychloroquine used.

So back to our initial questions.

Should we continue to include HCQ?
Yes, but cautiously. The existing data do not provide sufficient evidence to conclude that there is a clear benefit, lack of benefit or harm. The observational datasets are prone to bias. The RCTs are too small and have not reported on clinically relevant outcomes. Only adequately powered clinical trials with clinically relevant endpoints such as mortality or need for advanced respiratory support can provide the answers important to patients and clinicians. Nevertheless, concerns about toxicity have been appropriately raised and it will be important for sufficient safety monitoring to occur within the framework of a clinical trial infrastructure.

Should we be using a high dose of HCQ in ASCOT?
Yes. And indeed we think it is important to get the dosing right. HCQ may not be effective if adequate levels are not achieved early on and then maintained. A critical question is whether adequately dosed HCQ provides a meaningful clinical benefit. Studies using a lower dose may not be achieving a level adequate for the antiviral effect. In ASCOT, we are using 800mg twice a day for the first day, then 400mg twice a day for the next six days. Importantly, we have built in safety monitoring with ECG at baseline, four hours post initial dose, and 24 hours post initial dose. We will be excluding patients with already prolonged QTc, and carefully looking for evidence of prolongation of QTc once started on therapy. We will be actively checking for other drugs that may prolong the QTc and recommending caution with use of any such agents.

As data continues to appear, we will be monitoring all these closely as ASCOT proceeds. If it becomes obvious that HCQ has a very clear benefit, then we will provide HCQ to all patients. If HCQ leads to very clear harm, then we will stop using HCQ in ASCOT.