This Week in Family Medicine — Sep 20, 2026

Welcome to This Week in Family Medicine. This week we're covering 10 notable papers spanning cancer detection and surveillance, reproductive and allergy prevention trials, and the broader questions of sleep, global cardiometabolic inequity, and how we explain — and increasingly automate — the primary care consultation. Let's dive in.

We start with cancer, where the headline paper of the week is likely to change a recall letter you have sent hundreds of times. In the New England Journal of Medicine, Jover and colleagues report the interim analysis of a noninferiority trial across eight European countries in which nearly eleven thousand patients with high-risk adenomas — that is, an adenoma of ten millimetres or more, high-grade dysplasia, villous growth, or three to ten adenomas of any kind — were randomised to their first surveillance colonoscopy at five years rather than the currently recommended three [1]. After five and a half years of follow-up, the cumulative incidence of colorectal cancer was around eight in a thousand in both arms, marginally lower in the less intensively surveilled group, and the difference met the prespecified noninferiority criterion. Stage distribution looked similar, and there were five colorectal cancer deaths in total, split almost evenly. The important caveat is that this is an interim read-out of a trial whose primary endpoint is cancer incidence at ten years, so it is reassurance rather than a final verdict. But for the patient in front of you asking whether a three-year scope is truly necessary, there is now randomised evidence that waiting to five years does not appear to cost them anything.

Staying with cancer, two papers approach the harder problem of the undiagnosed patient. In PLOS Medicine, Nicholson and colleagues used English primary care records linked to national cancer registrations to study more than a quarter of a million adults presenting with unexpected weight loss, of whom about one in twenty went on to receive a cancer diagnosis [4]. They asked whether the trend in a patient's blood tests over the preceding years discriminates better than the single most recent abnormal result. The most useful finding is actually the simplest: adjusting blood test results for age and sex improved discrimination substantially, whichever approach was used. Historical trend added further discrimination for some specific test-and-cancer combinations — mean cell volume for bowel cancer and lymphoma, white cell and neutrophil trends for lung cancer — with the best models reaching an area under the curve of around 0.82. The practical message is that a platelet count or an albumin sitting just inside the reference range means something different in a seventy-five-year-old man than in a thirty-year-old woman, and that the direction of travel across old results is information we routinely discard. That sits neatly alongside a framework piece in The Lancet from Rebbeck, who points out that only about thirteen percent of cancers are currently detected through guideline-based screening, while somewhere between a fifth and a half first present in an emergency department [9]. The proposal is precision prevention and early detection — asking who is at elevated risk, with what intervention, at what intensity, and crucially arguing that de-escalating surveillance in genuinely low-risk people matters as much as escalating it in high-risk people. Read alongside the colonoscopy trial, this is the same idea arriving from two directions: less is sometimes the right answer.

Our second theme is reproductive health and allergy prevention, where one trial delivers a clear positive and another a clear negative. In JAMA, Elgemark and colleagues randomised 370 nulliparous individuals aged eighteen to thirty-one across eleven Swedish clinics to intrauterine instillation of ten millilitres of mepivacaine, twenty milligrams per millilitre, or saline placebo through a hydrosonography catheter two minutes before intrauterine device placement [2]. Pain during placement fell by roughly fifteen millimetres on a hundred-millimetre visual analogue scale, from just under sixty down to around forty-four — a genuine and statistically significant reduction, though not abolition of pain. Tolerability was the more striking outcome: nearly every participant in the mepivacaine group described the pain as tolerable, compared with about nine in ten on placebo, which works out to roughly one additional patient spared intolerable pain for every fifteen procedures. Given how often anticipated pain deters people from choosing an intrauterine device at all, this is a low-cost, blinded, multicentre demonstration that we can do better than telling patients to take ibuprofen beforehand. Contrast that with the PrEggNut trial in the New England Journal of Medicine, where Palmer and colleagues randomised over two thousand pregnant women whose unborn child had at least two close family members with allergic disease to either a high egg and peanut diet — at least six eggs and sixty peanuts weekly — or a standard lower-intake diet, from before twenty-three weeks' gestation until four months postpartum [3]. At one year, IgE-mediated egg or peanut allergy occurred in about eight percent of infants in each group, with no significant difference and no safety signal either way. So the answer to the mother who asks whether loading up on peanuts in pregnancy will protect her baby is: it will not, but it will not harm, and the evidence for early introduction in the infant remains where it was.

Third theme: the population-level determinants of chronic disease. In PLOS Medicine, Luo and colleagues analysed wrist-worn accelerometer data from more than ninety-five thousand UK Biobank participants, using an algorithm to derive sleep stages, duration, irregularity, and wakefulness after sleep onset, then mapped these against over a thousand health outcomes over a median of about nine years [8]. Variation in sleep patterns was associated with 156 incident diseases; more REM sleep was associated with lower risk across eighty-three of them, while greater irregularity and more wakefulness after sleep onset tracked with higher risk for a smaller set. Most importantly for counselling, the minimum-risk zone clustered at six to eight hours, and those sleeping under five hours accounted for the overwhelming majority of the harmful associations identified. This is observational, susceptible to reverse causation and residual confounding, so it cannot tell you that fixing sleep fixes disease — but it does strengthen the case for treating habitual very short sleep as a modifiable risk marker worth asking about. Meanwhile, in The Lancet Global Health, Gaye and colleagues pooled individual data from more than three hundred thousand adults across 109 surveys in 76 countries to examine the care cascade for hypertension, diabetes, and hypercholesterolaemia by household wealth [5]. Inequalities widened at every step of the cascade and were widest at the final step, disease control. But the regional patterns diverged sharply: in the Americas, control consistently favoured the wealthy, whereas in the African region coverage was uniformly low and the largest absolute gaps actually favoured the least wealthy, particularly for lipid treatment. And the modelling delivered an uncomfortable finding — the achievable absolute cardiovascular risk reduction tracked baseline risk rather than the size of the treatment gap, being highest in Europe and lowest in Africa, and consistently greater in men than in women. Populations with the biggest treatment gaps are not necessarily those who stand to gain most, which complicates how we target global cardiometabolic programmes. Also in The Lancet, the LATA trial from Bwakura-Dangarembizi and colleagues randomised 476 virologically suppressed adolescents aged twelve to under twenty across Kenya, South Africa, Uganda, and Zimbabwe to eight-weekly injectable cabotegravir-rilpivirine or daily oral triple therapy [10]. By ninety-six weeks, confirmed viral rebound occurred in about one percent of the injectable group versus just over six percent on daily tablets — not merely noninferior but superior, with only seven permanent discontinuations and comparable serious adverse events. For an age group with persistently poor adherence, long-acting injectable therapy looks like a genuine advance in a real-world African setting.

Finally, two conceptual pieces, both in the Scandinavian Journal of Primary Health Care, ask how we think and who thinks with us. Van Boven and colleagues propose predictive processing as a working framework for the consultation — the idea that symptoms emerge from the interaction between bodily signals and the brain's prior expectations, which lets you tell a patient that their symptoms are entirely real and potentially disabling even when the pathology is minimal or stable [6]. They offer a worked consultation example and are honest that direct trial evidence is limited. Stummer and colleagues take on generative artificial intelligence, arguing that the problem is not raw accuracy but misalignment: benchmarks were built in hospitals, not in general practice, and they note that GPT-4 scored below general practitioners on the Swedish family medicine specialist examination even as other models outperform physicians on hospital-style cases [7]. They set out six conditions — clinical, relational, organisational, cultural, linguistic, and epistemic — that evaluation must respect before these tools earn a place in our consulting rooms.

If you only have time for one paper this week, make it the colonoscopy surveillance trial in the New England Journal of Medicine [1]. It is the paper most likely to alter guidelines and the recall intervals you are already managing, and it lets you reassure high-risk adenoma patients about a longer wait with randomised evidence behind you.

Here are the key takeaways from this week in Family Medicine. First, beginning surveillance colonoscopy at five years rather than three after high-risk adenoma removal appears safe on interim data, though the definitive ten-year result is still pending. Second, when you interpret blood tests in a patient with unexpected weight loss, adjust mentally for age and sex, and look at the trend across old results rather than only the latest value. Third, intrauterine local anaesthetic meaningfully reduces pain and improves tolerability during intrauterine device placement in nulliparous patients — worth raising with whoever does your device fittings. Fourth, a high egg and peanut maternal diet does not prevent infant food allergy, so do not recommend it. And fifth, habitual sleep under five hours carries the broadest associations with incident disease, with six to eight hours the sweet spot worth counselling toward.

That's your roundup for This Week in Family Medicine. The full transcript and references are available on the episode page in your AudioScholar library. This is an AI-curated summary — for clinical decisions, always consult primary sources and current guidelines. See you next week.

If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And think of one colleague — in any specialty — who never has time to keep up with the literature. Tell them about AudioScholar: a free ten-minute weekly for every specialty, to listen to in any podcast app, or to read at audioscholar dot C C.

This is an automated summary generated by artificial intelligence, which can make mistakes. Always review the original source materials.