Sick in Bed
© Sebastian Rushworth.com
Most of us probably take fever lowering drugs, like paracetamol (a.k.a. acetaminophen, tylenol, panadol, alvedon), aspirin, or ibuprofen (a.k.a. advil, motrin, ipren), when we get a high temperature. The technical term for these drugs is antipyretics. After half an hour or so, we start to feel better and maybe don't have to spend the whole day in bed. But it is well understood among researchers studying the immune system that the fever is in itself an important part of the body's defence against infection. Our immune system works better at a higher temperature, and many pathogens have trouble replicating at a higher temperature. So, does taking antipyretics increase the risk of a more severe infection, or even of dying? And does it delay recovery?

A systematic review was carried out in 2010 seeking to answer this question, at least in part. It collected randomized controlled trials looking at the effect on mortality of treating influenza with antipyretics. Unfortunately, the authors of the review weren't able to find any trials involving humans, but they did find seven trials using mice, and one that used chickens. The review didn't receive any outside funding (i.e. the researchers who put the review together did it on their own time). In total 1,116 animals were included, with 697 in the antipyretic group and 419 in the control group. The animals in the antipyretic group were variously treated with aspirin, paracetamol, or diclofenac (a drug that is similar to ibuprofen).

41,8% died in the group that received antipyretics, compared with 38,1% in the control group. This was a 3,7% absolute difference in mortality, and it was statistically significant. If you think the number of animals that died in both groups was surprisingly large, that is because the strain of influenza they were infected with had been modified to be as lethal as possible.

So what conclusion can we draw from this? If you have a pet mouse, don't give it an antipyretic if it gets sick!

Of course, animal data is only ever suggestive; we can't know for certain whether the same effect exists in humans until we do the studies on humans. There are plenty of examples of drugs that have one effect in animals, and a different effect in humans.

A randomized double-blind controlled trial was carried out in 2015 looking at the effects of giving paracetamol to humans with influenza. It measured both the amount of viral shedding (a sign of how active the infection is and how infectious the patient is) and the speed of illness resolution. These endpoints aren't as important as mortality of course, but it's relatively rare for humans to die of influenza (about one in one thousand), so you would need a much larger study to show effect if that was your endpoint, which is probably why the researchers chose to look at these endpoints instead. The study was funded by the Health Research Council of New Zealand. 80 participants between the ages of 18 and 65 with symptoms of influenza (fever and at least one of the following: cough, sore throat, muscle pain, tiredness, or a runny nose) and a positive influenza rapid antigen test were randomized to receive either paracetamol or placebo. The average age of the participants was 26 years old. Many more patients in the placebo group had a respiratory co-morbidity (43% vs 10% in the paracetamol group), for example asthma or chronic bronchitis, which is a problem because it meant that the placebo group was more prone to bad outcomes from the start.

Let's talk results. When tested for PCR-positivity, only 46 of the 80 patients actually had influenza (24 in the paracetamol group and 22 in the placebo group), which shows how prone the rapid antigen test that was used was to giving a false positive result. Most of the rest of the patients were confirmed through further testing to have other respiratory viral infections. When the viral load was measured at baseline, day 1, day 3, and day 5, there was no difference between the groups. In terms of reported symptoms and temperature, there wasn't any difference either, so paracetamol didn't seem to be doing much to relieve the patients' symptoms. This is a bit odd, since I was pretty convinced paracetamol could at least do that much, based on personal experience.

When looking at the median time to resolution of symptoms, the paracetamol group saw complete resolution after an average of ten days, while the placebo group saw resolution after an average of seven days. However, this difference was not statistically significant. This is interesting, because there is a pretty big difference between seven and ten days. The reason it's not possible to tell if the difference is statistically significant or not is because of the small size of the study, which results in wide confidence intervals for the results.

Ok, this study wasn't very high quality, but unfortunately, the human data we have to work with is quite limited. The study had three big weaknesses. First of all, it was small. That means that each individual patient had a big effect on the result, and there is a big risk that any real difference that does exist drowned in random variation. Second, the patients in this study were young, so we can't extrapolate to older patients. And older patients are really the ones we're most interested in, since they're the ones with the weakest immune systems, who are most likely to die or suffer severe complications from an infection. A third weakness with the study was, as described above, that the groups varied significantly in terms of respiratory co-morbidities, which could really be messing up the data in favour of paracetamol.

Let's look at some higher quality data. A systematic review was published in 2017 looking at the use of antipyretics in the treatment of critically ill patients with sepsis (a bacterial infection in the blood stream). More specifically, the review looked at whether antipyretics had an effect on mortality. If there is a meaningful difference in mortality, these types of studies that look at really sick people should have better odds of finding it, since there are going to be many more deaths in them than in studies of people with milder infections. The authors were funded by the National Institutes of Health, a couple of private foundations, and Washington University in St Louis. The review looked at both observational and randomized studies. Eight randomized studies involving 1,507 patients were included, as were eight observational studies involving 17,432 patients.

Neither the randomized trials nor the observational studies showed any statistically significant difference in mortality at 28 days. The randomized trials showed a slight decrease in mortality in the group treated with antipyretics, but the difference was nowhere close to statistically significant.

So, what conclusions can we draw from all this? In animal studies, it is clear that treating with antipyretics increases the risk of serious complications and death. In humans, there is as yet no good evidence either way.

My personal take-away is that if you have a sore throat or a fever and are otherwise healthy, you can probably get away with taking antipyretics to relieve your symptoms without any marked effect on the length of illness or risk of complications. If you're old and have co-morbidities, then maybe you should consider being more careful, although that is really a speculative conclusion based on the animal studies.

One thing that concerns me is that a large portion of the elderly are on paracetamol all year round for chronic pain conditions. Considering that the elderly already have a weakened immune system, I wonder how much of an additional hit they are taking from their paracetamol, and whether that is increasing their risk of dying when they get an illness. That is a subject which I think definitely merits further study.