Tamiflu and Paxlovid: when antivirals actually help, and by how much

The two antivirals people most often ask about — oseltamivir (Tamiflu) for influenza and nirmatrelvir-ritonavir (Paxlovid) for COVID-19 — are frequently spoken of in the same breath, as though "an antiviral" were one kind of thing with one kind of benefit. The evidence says otherwise. They differ not just in effect size but in what they were shown to do and for whom. Understanding that is the difference between a drug that lives up to expectations and one that quietly does not.

Tamiflu: a real but small effect on flu

Oseltamivir was, for years, stockpiled by governments against pandemic flu on the strength of claims that it reduced complications and hospitalisations. A 2014 Cochrane review, produced after a long campaign to obtain the full trial data from the manufacturer, took a much more sober view. Pooling the randomised evidence, it found that in adults, oseltamivir shortened the time to first alleviation of influenza symptoms by around 16.8 hours — from roughly seven days to a little over six [s1]. In children the picture was mixed, with a benefit in otherwise healthy children but not in those with asthma [s1].

More striking was what the review did not find. It found no good evidence that oseltamivir reduced hospitalisations, and the data on serious complications such as pneumonia were unreliable because of how the original trials had recorded them [s1]. Meanwhile the drug caused measurable harms: it increased the risk of nausea and vomiting, and, used preventively, of headaches and psychiatric and renal effects [s1]. The reviewers' conclusion was that the benefits were smaller, and the harms better established, than the stockpiling case had assumed [s1]. The debate over that review was fierce and is not fully settled — some observational data and later analyses argue for larger benefits in sicker, hospitalised patients — but the core finding for ordinary flu stands: on average, oseltamivir shortens symptoms by well under a day.

Paxlovid: large benefit, narrow group

Nirmatrelvir-ritonavir tells a very different story, with a sharp lesson about who is studied. Its pivotal trial, EPIC-HR, enrolled unvaccinated adults with COVID-19 who were at high risk of severe disease and treated them within days of symptom onset. The result was dramatic: among those treated within three days, 0.77% were hospitalised or died by day 28, against 6.31% on placebo — a relative reduction of about 89% [s2]. For a high-risk, unvaccinated patient caught early, that is a large and clinically meaningful effect, and it is why the drug was authorised and widely used [s2].

But the population matters enormously, and a second trial made the point. EPIC-SR tested the same drug in standard-risk adults — a group that included vaccinated people and those without major risk factors — and did not find a statistically significant benefit on its primary endpoint of sustained symptom relief [s3]. In other words, the striking 89% figure belongs to a specific group at substantial baseline risk; extrapolating it to a young, vaccinated, otherwise-healthy person with a mild case is exactly the error the second trial guards against [s3]. When the baseline risk of a bad outcome is already low, cutting it by a large percentage removes very little in absolute terms.

The general lesson

Put side by side, the two drugs teach the same two things. First, "an antiviral works" is an incomplete sentence: the honest questions are how much, measured how, and against what baseline. A 16.8-hour reduction in flu symptoms and an 89% reduction in hospitalisation are both "it works," and they are not remotely the same claim [s1][s2]. Second, benefit depends on risk and on timing — antivirals are generally most useful when started early and when the person has enough baseline risk for a large relative reduction to translate into a real absolute one [s2][s3]. Who is at that level of risk, and whether a given drug is appropriate, is a clinical judgement. This explainer sets out what the trials measured; it is not a recommendation to take, avoid, or dose any medication, and it does not describe dosing.