In this first EMC Journal Club (where we take the “boring” out of journal clubs and deliver clear, concise, practical critical appraisal knowledge based on an Emergency Medicine journal article that may have flown by your radar – not too detailed and not too brief), Dr. Rohit Mohindra, an Emergency Physician at North York General in Toronto and SREMI researcher works his critical appraisal magic on the article “Fever therapy in febrile adults: systematic review with meta-analyses and trial sequential analyses” by Holgersson et al. Plus, for the EBM keeners, we have Dr. Shelley McLeod, clinical epidemiologist at SREMI give us a research methodology hot take on the difference between a traditional meta-analysis and a network meta-analysis and why it matters. Does treating fever in adults make a difference?

To sign up for an email blast with the EMC Journal Club of the month delivered to your inbox go to our subscribe page

Fever therapy in febrile adults: a systematic review with meta-analyses and trial sequential analyses

Case

A 55-year-old male with diabetes, hypertension and CKD presents with 5 days of persistent fever and a positive COVID rapid test at home.

Your initial impression is that the patient has uncomplicated COVID. However, the patient is reluctant to keep taking ibuprofen for his symptoms because of his co-morbidities. He wants to know if it is dangerous not to treat his fever.

Paper: Fever therapy in febrile adults: systematic review with meta-analyses and trial sequential analyses | The BMJ

Summary

Patients: any adult patient with a fever

Intervention: any antipyretic therapy (medication and/or physical cooling)

Comparison: placebo

Outcomes: mortality, serious adverse events

Results: Fever therapy did not reduce risk of death (risk ratio 1.04, 95% confidence interval 0.90 to 1.19; I2=0%; P=0.62; 16 trials; high certainty evidence) or the risk of serious adverse events (risk ratio 1.02, 0.89 to 1.17; I2=0%; P=0.78; 16 trials; high certainty evidence)

Study author’s conclusions: Fever therapy does not seem to affect the risk of death or serious adverse events

Critical appraisal

The quality of meta-analysis results depends on a few key things:

  1. Studying the same patient population
  1. Studying the same outcome
  1. Avoid bias while doing the study (things like blinding, randomization, matched cohorts etc can help reduce this)

There are many ways to evaluate this: heterogeneity scores, risk of bias tables, and the tables of study inclusion and outcome data are the most important to read carefully.

Although the results from this paper are interesting, the results are truly meaningless. Just look at the wide disparity in the populations included in the study: 3007 were critically ill, 1892 were non-critically ill, 3277 had an infectious fever, and 1139 had a non-infectious fever; 3062 participants were admitted to the hospital, and 2078 were outpatients. From those numbers alone, one can already guess that there will be “no difference” in outcomes. As well, if you add in the multitude of treatments they included and it’s pretty clear that there was no chance they would see a difference anyway.

Other issues are that half of the studies have a high risk of bias, and the authors did not do pre-specified sub-group analysis to help understand if specific populations benefit from antipyretic treatment.

Take home message

Meta-analysis and systematic reviews work best on focused populations and treatments/interventions. Broad inclusion criteria are just increasing the risk of finding no difference and making a false negative conclusion (that there is no difference).

You tell your patient that we genuinely do not understand the mortality benefit of treating fevers, but the treatments are generally safe, well-tolerated, and may make illness more tolerable. However, in his case, considering CKD you would advise against the routine use of NSAIDs given their well-established harm in CKD and take acetaminophen if it makes him feel better.

Research Methodology Hot Take

Meta-analysis is the statistical procedure for combining data from multiple studies with the goal of providing a single, best summary estimate of the treatment effect. It is used when studies are similar enough to make combining the results a sensible thing to do. Judging combinability requires mostly clinical common sense, but you have to assess both clinical and methodological heterogeneity. As Dr. Mohindra mentioned, to assess clinical heterogeneity, look to see if the inclusion/exclusion criteria, intervention, comparison, and outcome measures are similar between studies. To gauge the statistical heterogeneity, look at the I2 statistic. If the I2 is more than 50%, data shouldn’t be pooled. While the I2 in this meta-analysis was 0%, the clinical heterogeneity suggests this data should never been pooled.

In the 16 (n = 2415) RCTs included in meta-analyses, although all patients were hospitalized, 85% were critically ill, 10% were not critically ill, and 5% were in a trial that included patients in both groups. Of the included patients, 69% had an infectious fever, 20% had a non-infectious fever and the rest were not classified. The authors included all types of therapy including antipyretics (ibuprofen or acetaminophen), IV catheter-based cooling, external cooling just to name a few. What this meta-analysis doesn’t tell us is if any of these treatment options are better than another. Obviously that wasn’t the intended research question, but if it was, the most appropriate methodology would have been to conduct a network meta-analysis.

What’s the difference between a traditional meta-analysis and a network meta-analysis?

Traditional pairwise meta-analysis is a statistical technique for combining the results of multiple studies measuring the same outcome into a single pooled estimate. One limitation of traditional meta-analyses is they are restricted to head-to-head comparisons and cannot evaluate treatments that have not been directly compared.

With a network meta-analysis, we can compare all possible drug treatments, even if they have not been studied in head-to-head trials.

We can estimate the relative effect of treatment B to C as long as each treatment has been directly compared to a common intervention, in this case treatment A. If you know the direct treatment effect from trials comparing A to B, and A to C, you can get an indirect estimate of the treatment effect of B to C. A network meta-analysis combines both direct and indirect evidence over the entire network of interventions, allowing us to compare multiple treatment effects.

To sign up for an email blast with the EMC Journal Club of the month delivered to your inbox go to our subscribe page

Drs. Mohindra, McLeod and Helman are supported by the Schwartz-Reisman Emergency Medicine Institute.

References

  1. Holgersson J, Ceric A, Sethi N, Nielsen N, Jakobsen JC. Fever therapy in febrile adults: systematic review with meta-analyses and trial sequential analyses. BMJ . 2022 Jul 12;378:e069620. doi: 10.1136/bmj-2021-069620.
  2. Rouse B, Chaimani A, Li T. Intern Emerg Med. 2017 Feb; 12(1): 103–111. doi: 10.1007/s11739-016-1583-7.