Humanity now publishes such a torrent of research that the central problem is not generating knowledge but making sense of it. The author notes that in 2025 academics worldwide published about 7 million scholarly articles, more than 19,000 each day. When individual studies are reported in isolation, health news appears to flip-flop: coffee is good one week, bad the next; red wine extends life until it doesn’t. The real question is how to weigh each new study against everything else we already know.
The stakes can be high. In September 2025 the US Food and Drug Administration announced it would add a warning label to the painkiller acetaminophen, known as paracetamol, claiming that taking it during pregnancy increased a child’s risk of autism. The story made global headlines, but the author says the FDA and its parent health department had cherrypicked a few studies and downplayed stronger findings. A major review published a couple of months later concluded that existing evidence does not clearly link maternal paracetamol use during pregnancy with autism or ADHD in offspring. By then many women had been needlessly scared about a medicine widely considered one of the safest to take during pregnancy.
This pattern has systemic causes. Scientists publish and promote one paper after another because that is how careers advance, while journalists chase the latest, flashiest studies for clicks. Hardly anyone, the author argues, spends time carefully synthesising and taking stock of existing knowledge. Iain Chalmers, a doctor and co-founder of the Cochrane Collaboration, called this the scandalous failure of science to cumulate evidence scientifically.
A key turning point came in 1976, when statistician Gene Glass presented a method at a San Francisco hotel. He had been helped by psychotherapy and was irritated by Hans Eysenck’s influential reviews, which used vote-counting and concluded psychotherapy had little benefit. Glass and psychologist Mary Lee Smith gathered more than 370 studies, converted each study’s outcome into a standard measure of effect size, and averaged them. Glass called the method meta-analysis. The audience was thunderstruck: here was a way to extract meaning from an apparent mess of different results.
Around the same period, systematic reviews emerged as a rigorous way to collect and evaluate evidence. Researchers search databases worldwide, whittle thousands of studies down to the most relevant, assess reliability, extract data and combine results. Many include a meta-analysis. The author compares the approach to seeing the forest rather than the trees. In medicine, Iain Chalmers showed the power of this work. His decade-long project synthesising trials on treatments in pregnancy and childbirth found that practices such as shaving women’s pubic hair during labour and surgical episiotomies were based on little evidence and some were harmful, helping change practice. In 1993 he and colleagues started the Cochrane Collaboration.
Cochrane reviews became a bedrock of medical evidence, shaping clinical guidelines used by doctors. The author says most people who have seen a Western doctor have unknowingly benefited from systematic reviews. A tragic counterexample is infant sleep advice. In 1957 paediatrician Benjamin Spock advised parents to put babies to sleep on their fronts; by the 1980s and ’90s studies showed this contributed to a steep rise in sudden infant death syndrome. A 2005 systematic review and meta-analysis found the link was already clear in 1970, about twenty years before most parents were warned. The authors calculated at least 50,000 infant deaths in Europe, Australasia and the United States might have been prevented.
The model spread beyond medicine. The International Initiative for Impact Evaluation holds more than 1,700 systematic reviews on international development, and the environmental database CEEDER hosts more than 2,100 reviews on protecting species and the planet. The Education Endowment Foundation’s teaching and learning toolkit, built from systematic reviews, is used by nearly 70 per cent of school leaders in England and has been adapted in more than 30 countries. It shows teaching metacognition gives about eight months of additional learning progress on average, while repeating a school year sets students back by about two months.
Climate science has also adopted synthesis. The Intergovernmental Panel on Climate Change’s assessments are among the biggest evidence syntheses ever done, and they have helped inspire global action by showing humans are causing climate change. A group called What Works Climate Solutions, started in 2024, is building an evidence bank of syntheses on the most effective ways to cut emissions or adapt, intended to feed into the next IPCC assessment due by late 2029.
Systematic reviews are now hugely popular but face criticism. Medical reviews rose from around 1,400 in 2000 to more than 29,000 in 2019, and the PRISMA guidelines are among the most cited papers of the 21st century. Critics say conventional reviews privilege quantitative studies and ignore qualitative research, lived experience and Indigenous knowledge. In 2021 Parks Canada and Foundations of Success included Indigenous groups, zoos and reindeer experts from Finland in an analysis that supported a captive breeding programme for endangered caribou in Jasper National Park; calves were born in 2025. Researchers have also developed qualitative, mixed-methods, rapid and living reviews, including living reviews used during the COVID-19 pandemic.
Evidence synthesis has become a profession, but a full systematic review can take nearly three years on average. Artificial intelligence is now being tested to accelerate the work. Start-ups are developing systems that find, sort and summarise research, but most have not been shown to reliably produce high-quality systematic reviews. The author warns that AI can give inaccurate answers and that asking chatbots may bypass rigorous synthesis. The lesson is to resist the glamour of new studies: single studies can be outliers, and the unglamorous work of collating existing knowledge is how we figure out what it all means.
Comments