It’s very rare we produce three instalments on one particular drug for one particular condition, and certainly not within such a short time frame.
However, tocilizumab has been worth it. Maybe not the one intervention to rule them all, but a therapy that despite courting initial controversy, now seems to be emerging with a definitive role to play in the management of COVID-19. If you haven’t seen previous discussions then take a look at our earlier posts.
Then, before you get hammered into submission by headlines and MHRA interim policy documents, have a good look at the newly available preprint data published by the RECOVERY trial investigator group. At St Emlyns, we would always encourage you to look at the primary evidence, rather than listen to or read a second-hand interpretation. This is even more vital at present, given the thirst for therapeutic options and the challenges around evidence-based medicine in a pandemic.
As a brief recap, the IL-6 antagonist tocilizumab (TCZ) has now been studied as a therapeutic intervention for COVID-19 in at least 7 trials, to our knowledge. Up until recently, significant uncertainty remained regarding clinical use. The two most recent trials were stopped early following recommendations by the trial steering committee, but for conflicting reasons; REMAP-CAP (published today in the NEJM, following release via preprint) reported a significant clinical benefit in critically ill patients, yet Veiga et al reported a potential increase in mortality with the intervention. There were limitations to both studies and consequent challenges interpreting the results. An excellent summary of the previous disparate findings has already been published in this BMJ editorial.
There is also some limited evidence to suggest that that increasing IL-6 levels are associated with higher mortality in this disease, which provides us with biological plausibility.
So what’s new?
On Thursday 11th Feb, the RECOVERY trial investigators reported findings from their adaptive platform RCT comparing tocilizumab to standard of care in patients hospitalised with COVID-19. In order to be randomised, patients also needed to be hypoxic (defined as requirement for supplemental oxygen or SpO2 <92% on room air) and have evidence of systemic inflammation (defined as CRP>75mg/L). At 4116 reported participants, this trial contained more than double the patients of all previous studies combined.
And it was good news according to the press release and subsequent media attention; patients receiving tocilizumab were more likely to survive to 28 days and less likely to clinically deteriorate, regardless of the level of respiratory support. These benefits were in addition to the clinical improvements seen with corticosteroid use. No wonder some news outlets got excited, reporting the above combination (corticosteroids and TCZ) could halve the risk of death.
Good news is always welcome – but needs to be critically appraised and rationally applied. This drug is a potent immunosuppressant and I think we would all have some anxiety about widespread use throughout the hospital. So grab a brew, let’s go through the preprint and try to work out why the authors report a 4% mortality reduction but the Daily Mail shouts about halving the death rate…..
Tell me about the design.
Routine inclusion criteria for the RECOVERY trial are published on the relevant website and I think we are all fairly comfortable with this methodology now. Essentially, anyone hospitalised with suspected or confirmed SARS-CoV-2 infection is eligible for initial recruitment to the trial platform.
Further eligibility criteria for the tocilizumab domain are as described above (hypoxia and systemic inflammation). Specific domain exclusion criteria are described in the preprint as hypersensitivity, evidence of tuberculosis or clear evidence of active bacterial, fungal, viral or other infection. This latter one is always subjective.
Of note, there are specific issues within this trial design (and REMAP-CAP) that may introduce bias and confounding to the results. Any medical history that might ‘put the patient at significant risk’ in the opinion of the attending clinician, is considered a barrier to recruitment. Patients were also ineligible for the tocilizumab arm of the study if the managing clinician felt it was either definitely indicated or definitely contraindicated. All these criteria are again very subjective and could theoretically create a recruitment cohort of lower risk subjects, through selection bias. However, you could also argue that the investigators are trying to select out the subpopulation where there is genuine equipoise and evaluate the effectiveness within a pragmatic usual care setting. Tricky.
In addition, RECOVERY is an open label trial. Lack of blinding can cause further confounding in downstream medical management. If you know your patient has had tocilizumab, your clinical management could conceivably change in ways outside the study protocol, through additional antimicrobial prescribing, more aggressive medical management etc.. We may get some assurances about this within a main paper after peer review, but at present there is limited information in the preprint.
What was the intervention?
TCZ was given open label as an initial weight adjusted IV infusion over 60 minutes. The protocol allowed a further dose after 12-24h, based on subjective clinical impression of limited improvement. Just to put the first question to bed, less than a third of patients recruited who completed follow up got a second dose (29%). As such, we should look at these results as primarily delivering the intervention of a single dose.
Looking at the numbers does throw up some early questions though. The supplementary webtable reports that 2022 patients were randomised to receive TCZ. Only 1602 had ‘compliance data available’. I am not sure what happened to the other 21% patients here. Furthermore, of those patients with ‘compliance data available’, only 83% had a follow up form completed and analysed at the time of reporting. This is a further loss of 17%. As such, we are left with a reported randomised allocation of 2022 TCZ patients with verified follow up for only 1333 (66%). This is slightly alarming.
Usual care got some TCZ as you might expect during a pandemic, but not loads (3%). Otherwise the groups appear well matched in terms of what we might consider standard of care; 70% in both groups received corticosteroid therapy, with an interesting amount of azithromycin (32-35%), remdesivir (27-29%) and convalescent plasma (20-22%).
What was the primary outcome?
There can be only one of these – and it was the right one; all cause mortality. This is essential to look at with regard to an intervention that could potentially have significant adverse events (such as iatrogenic infection). The 28 day cut off is arbitrary, but accepted. It will be very important to look at the planned 6 month follow up when released, to see if there is any later signal of downstream issues following TCZ.
Secondary outcomes included time to discharge alive from hospital; progression to mechanical ventilation (MV), ECMO or death on those not receiving MV at randomisation; use of non-invasive respiratory support; time to successful cessation of MV; use of renal replacement therapy. Adverse events were captured ‘in an expedited fashion’ and pre-specified safety outcomes ensured that cause specific mortality was also captured within the data.
Tell me about the statistical jiggery pokery – was this one of those Bayesian things?
No, I don’t think so. The authors report an abbreviated power calculation determined by the TSC (which is a little odd), but was at least pre-specified. The numbers seem predicated on the assumptions that if 28-day mortality in the usual care group was >25%, recruitment of around 4000 patients total would provide 90% power at a 2 sided alpha level of 0.01 to detect a proportional mortality reduction of at least 1 fifth (20% relative risk reduction). As such, 4000 was the golden number, hence cessation of the domain at 4116 patients.
This is more understandable to us simple folk, but begs the question of why the result was therefore so profound when the sample size of patients with ‘compliance data available’ for this interim report was only 3266 and the mortality reduction reported was actually less than one fifth (4% absolute risk reduction on a control group mortality of 33% would be 12% relative risk reduction, to my mind). I am sure someone cleverer than me can explain this. But I can’t.
Are these patients similar to mine?
Only you can answer that. But they certainly look like the ones I am seeing in hospital. A review of baseline characteristics in table 1 suggests that this cohort was predominately white males. Mean age was 63.6 years (SD 13.7). BMI was not reported, but more than half of recruited patients had a significant comorbidity.
This was a similar cohort to REMAP-CAP, in that they were randomised at an early stage of illness, a median of 9 days since symptom onset and 2 days since hospitalisation. This would fit with the theory of significant cytokine driven inflammatory reaction causing a secondary deterioration at days 7-10. However, the level of respiratory support differed significantly from REMAP-CAP. The majority of recruits (1868) were receiving supplemental oxygen only, with 9 of these on no oxygen at all. 1686 patients were receiving non-invasive ventilation, defined as high flow oxygen, CPAP or other NIV. Only 562 patients were receiving MV at the point of randomisation. Median CRP levels were similar to the TCZ population in REMAP-CAP (140 to 150mg/L) and the vast majority of patients were already receiving corticosteroid medication.
And the key results?
So. An intention to treat analysis on the primary outcome. All above board and prespecified. A reported mortality of 33% (694/2094) in the standard care arm, compared to 29% (596/2022) if randomised to TCZ. An absolute risk reduction of 4%. A risk ratio of 0.86 (95% CI 0.77 to 0.96, p=0.007). Happy days.
HOWEVER. Those numbers require a little unpicking. Given the nature of this preliminary report and the lack of completed follow up, we are actually talking about an intervention cohort of 1602 patients with a completed follow up form and a control cohort of 1664. Indeed, only 1333 patients in the TCZ arm of the study actually received the intervention. That is only 65.9% of patients in the intervention arm of the trial. As such, it’s hard to get a clear idea of what the figures within a per protocol analysis would be. This is important, as if only 1333 patients in the intervention arm received TCZ, but the outcomes are reported for around 2000 patients, the mortality within the entire cohort could be entirely outside of the intervention (making TCZ even better than reported) or entirely within (leading to less evidence of benefit). I’m a little surprised they didn’t attempt this. They did perform an exploratory analysis restricted to patients with a positive SARS-CoV-2 test result; I think they could have attempted an early per-protocol analysis limited to those receiving the intervention and with follow up data available.
The authors report in the stats section that 92% patients had information available on the primary outcome for this preliminary report. In the results section, they present the figures above, with completion of follow up reported in <80% for both intervention and control groups. It’s a bit tricky to understand exactly which of these figures they are working with, and how they came to them.
I see (or maybe not, but think I get what your concerns are). What about secondary outcomes?
Several of these supported a beneficial effect of TCZ, within the reported trial context. Randomisation to TCZ was associated with a greater probability of discharge from hospital alive within 28 days and a reduction in the risk of progressing to the composite of MV or death.
Interestingly, no significant effect was seen on progression to NIV or MV, amongst patients on no respiratory support at randomisation. In addition, no significant effect was seen regarding cessation of MV, for those patients who were on it at randomisation.
Preliminary data is reassuring for those of us concerned about immunosuppression and iatrogenic infection. Cause specific mortality did not show evidence of excess deaths from other infections and there were minimal serious adverse reactions reported within the study.
What to do now?
What to do, what to do.
I suppose the first thing to do, is to ask for the completed trial results. This is a preliminary analysis and we need to remember that, as we apply broad risk reduction estimates to our protocols and guidelines.
However, this is compelling and promising data. The composite forest plot demonstrating a rapid meta-analysis of all 8 trials in figure 4 shows an overall statistically significant reduction in 28 day mortality with the use of TCZ, in >7500 patients. A risk ratio of 0.87. Again, happy days.
If I wanted to be a grinch though, I might point out that the absolute risk reduction in the meta-analysis is more like 2.7%, which is of course a NNT of 38 (rounded up). And is a far cry from a halving of death rates we mentioned earlier (although I expect the mail are lumping all beneficial interventions together in a best case scenario and comparing to an early baseline mortality, without steroids or other agents).
I might also raise the question of who to give this to, and when? REMAP-CAP was leading us towards prioritising use in critically ill patients. I would suggest RECOVERY leans towards an almost opposite stance – greater benefit was seen in patients who received the intervention at <7days from onset of symptoms for example and in those who were on supplemental oxygen or NIV, compared to those receiving tocilizumab at a later stage or who were on MV at randomisation (figure 3).
I would also point out that according to the subsidiary clinical outcome data presented here, this drug does not seem to improve your chances of dodging non-invasive ventilation and/or MV if you were on neither when you started it. Or your chances of coming off a ventilator if you were on one when you started it. It may reduce your overall chances of MV from a combined baseline of supplemental oxygen or NIV at the point of randomisation; this was a secondary outcome result and the absolute difference reported is small (3%), but it is an important one.
You haven’t answered my question – what shall I do?
Well, like so much in this pandemic, I think this is likely to be taken out of (y)our hands. The MHRA have already issued interim guidance in the UK advising use. I suspect your trust/organisation will rapidly develop policy on this and then the limiting steps will be focussed on supply/delivery.
Having said that, I think there is a real role for clinicians to perhaps bring everyone back down to earth on this and ensure Tocilizumab is used wisely. The overall data now support use in COVID-19, but only when given early in the disease course to patients who mirror the trial inclusion criteria; I would suggest these are patients without any concern for concomitant bacterial infection, who have clear evidence of systemic inflammatory process who are in a higher risk clinical group for deterioration/progression of disease. I don’t think we should be giving this agent for those who are likely to recover anyway with conservative management. I don’t think we should be giving this drug to delayed presentations as a last ditch rescue strategy. I personally don’t think we should be giving it at the same time that we give broad spectrum antibiotics for community acquired pneumonia, but I accept that in the early phase of clinical presentation clinicians may feel they want to do everything possible to improve the chances of recovery. If you want to read another editorial, Dr’s Rubin, Longo and Baden discuss the continued challenges with TCZ use in this NEJM editorial released today, with particular mention of timing.
We also need to remember the SPC contraindications, monitoring requirements and subsequent downstream impact. CRP is essentially switched off for several weeks, for example, and will no longer help guide decisions on therapy (if it ever did). Baseline microbiology panels should be considered, to allow guided treatment strategies in case of consequent deterioration. According to the SPC, women of childbearing age should be advised to use effective contraception up to 3 months following treatment. Who knows if this therapy is safe in pregnancy; “do not use during pregnancy unless clearly necessary” is the unhelpful company line…
What next for RECOVERY? The recovery trial has moved on and we are now recruiting to arms that include Baricitinib (a JAK-2 inhibitor), Dimethyl Fumarate (an anti-inflammatory drug used in multiple sclerosis), Aspirin, Colchicine and REGN auto-antibodies. One concern expressed by colleagues is that we may end up with patients on dexamethasone, tocilizumab, and baricitinib which is s LOT of immunosuppression. Anecdotally, ICU colleagues have identified patients with opportunistic infections and it is possible that the effect of these may not be identified within the 28 day follow up of RECOVERY or the 21-day follow up in REMAP-CAP. It is an area that we need to watch closely.
We also need to ensure patients and their relatives understand tocilizumab is not a wonder-drug. It is great news that we have another agent that can potentially help with this awful disease. But the data overall would suggest it is not a panacea. And the most positive data comes from a study that still has 20% patients to follow up properly, has not reported long term impact of this therapy and has several methodological limitations which have not yet been subject to peer review. Any decision to use it should come with a frank discussion regarding the risks/benefits in line with GMC guidance on the prescribing of unlicensed medications.
We must all keep our eyes out for publication of the full papers for REMAP-CAP / RECOVERY and not close the Tocilizumab/IL-6 chapter of this COVID-19 book quite yet. Until we have read, digested and discussed these findings in full, we must be honest with patients about the reasons why we are offering this medication; early data, face validity and hope. And we must highlight the continued uncertainties and the potential downsides.
Here’s hoping the papers are published soon, and the results with completed follow up are even better.
- RECOVERY Collaborative Group. Tocilizumab in patients admitted to hospital with COVID-19 (RECOVERY): preliminary results of a randomised, controlled, open-label, platform trial | medRxiv 2021.02.11.21249258; doi: https://doi.org/10.1101/2021.02.11.21249258
- Evidence-based medicine and COVID-19: what to believe and when to change | Emergency Medicine Journal Carley S, Horner D, Body R, et al Evidence-based medicine and COVID-19: what to believe and when to change Emergency Medicine Journal 2020;37:572-575.
- The REMAPCAP Investigators. Interleukin-6 Receptor Antagonists in Critically Ill Patients with Covid-19 – Preliminary report | medRxiv Interleukin-6 Receptor Antagonists in Critically Ill Patients with Covid-19 – Preliminary report | medRxiv 2021.01.07.21249390; doi: https://doi.org/10.1101/2021.01.07.21249390
- Veiga Viviane C, Prats João A G G, Farias Danielle L C, Rosa Regis G, Dourado Leticia K, Zampieri Fernando G et al. Effect of tocilizumab on clinical outcomes at 15 days in patients with severe or critical coronavirus disease 2019: randomised controlled trial BMJ 2021; 372 :n84
- McCreary Erin K, Meyer Nuala J. Covid-19 controversies: the tocilizumab chapter BMJ 2021; 372 :n244
- The REMAPCAP Investigators. Interleukin-6 Receptor Antagonists in Critically Ill Patients with Covid-19. NEJM 2021; online first DOI: 10.1056/NEJMoa2100433
- Rubin, Longo and Baden. Interleukin-6 Receptor Inhibition in Covid-19 — Cooling the Inflammatory Soup. NEJM 2021; Online first DOI: 10.1056/NEJMe2103108