Is your coffee too hot? Plus, AI might make your doctor better. And what your fertility test can and cannot tell you

Welcome to Health(y) Discourse.

This week, we’re looking at whether drinking your coffee or tea too hot could raise your cancer risk, whether AI can actually help doctors make better clinical decisions and what a new study on acetaminophen use during pregnancy does — and doesn’t — tell us about potential risks to children. 

Last week, we looked at “eating jet lag,” why the FDA is inspecting fewer foreign food producers and what you can do when a hospital asks you to pay before receiving care. 

Other stories that caught my eye this week:

• Federal health officials declared that the cyclosporiasis outbreak that has sickened thousands of people has ended.
• A new survey found that all 34 cancer centers included in the poll are facing a shortage of at least one medication, with 20% reporting shortages of more than five medications. More than half of the centers said they can still treat patients despite shortages. 
• The CDC merged its Influenza Division with the Coronavirus and Other Respiratory Diseases Division to create the Influenza and Respiratory Viruses Division. The reorganization has been in the works since the Biden administration. 
• A Canadian study of nearly 9,000 adults found that people who reported having COVID-19 at least twice were more likely to report being diagnosed with a new medical condition afterward than those who had never been infected. Researchers examined 21 conditions, ranging from cardiovascular and respiratory diseases to neurologic disorders, diabetes, mental illness and chronic fatigue syndrome, adding to evidence that the relationship between COVID and chronic illness may run in both directions. 

Peer Review

Rehashing the week’s health headlines

☕Under the Microscope

Should you let your coffee cool down? 

If you like your coffee or tea piping hot, a new study suggests it might be worth waiting a few minutes before taking that first sip.

Researchers followed nearly one million U.K. adults for more than a decade. They found that people who preferred their drinks “very hot” had about three times the risk of developing a type of throat cancer called esophageal squamous cell carcinoma compared with people who preferred warm drinks. Those who preferred hot drinks had nearly twice the risk.

Coffee and tea themselves weren’t the problem. Researchers found the association with temperature. One theory is that repeatedly exposing the esophagus to very hot liquid damages its lining, potentially contributing to cancer development over time. This isn’t an entirely new concern: the International Agency for Research on Cancer has classified drinking beverages above 65°C, or 149°F, as “probably carcinogenic” since 2016.

This was an observational study, so it can’t prove that hot drinks caused the cancers, and researchers asked people whether they drank beverages warm, hot or very hot rather than actually measuring the temperature. Esophageal squamous cell carcinoma is also relatively uncommon; the lifetime risk in the UK is about 1%. Smoking and alcohol remain much more important and established risk factors.

Bottom line: No need to give up your morning coffee or tea. But if it’s hot enough to burn or feel uncomfortable, letting it cool before drinking is a simple precaution with essentially no downside.

👀 Worth Watching

Does AI make doctors better at diagnosing patients? 

We’ve heard plenty about AI’s potential to transform medicine. A new randomized trial offers some interesting evidence that it could actually help doctors make better clinical decisions.

Researchers recruited 249 physicians in Kenya, Indonesia and the Netherlands and randomly gave some access to OpenAI’s GPT-4o AI model while they worked through clinical scenarios. Their answers were scored on how well they identified the likely diagnosis, recommended appropriate tests and treatment and explained their reasoning. Doctors using AI performed better in all three countries. The biggest improvement occurred in Kenya, where scores were 18 percentage points higher with AI assistance, compared with improvements of about 11 points in Indonesia and 7 points in the Netherlands.

The researchers found AI appeared to help with several parts of clinical decision-making, rather than simply helping doctors name the right diagnosis. That’s encouraging, particularly because one potential benefit of these tools is giving clinicians rapid access to medical knowledge and decision support. And the larger improvement among Kenyan physicians raises an intriguing possibility that AI could be especially useful in settings where clinicians have less access to specialists, diagnostic resources or other forms of clinical support — although this study wasn’t designed to prove why the benefit differed between countries.

But there’s a big catch: This wasn’t real patient care. Doctors were completing controlled clinical scenarios, and those in the comparison group weren’t allowed to use resources they might ordinarily consult, such as clinical guidelines or the internet. The researchers also didn’t evaluate whether AI introduced harms. 

There is also some evidence of a potential downside of relying on AI. A 2025 study found that after physicians routinely used AI to help spot potentially precancerous growths during colonoscopies, their detection rate during procedures performed without AI fell from 28% to 22%. That study was observational and did not prove AI caused the decline, but it raised an important question: Could relying on AI weaken some of the skills doctors need when technology isn’t available?

Bottom line: This is stronger evidence than simply showing that an AI model can pass a medical exam. It suggests AI can improve doctors’ performance when used as a tool. But the much bigger question remains unanswered: Does giving doctors AI actually lead to better — and safer — care for real patients?

💊Consumer Watch

What does the new Tylenol-in-pregnancy study actually tell us?

Another study is raising questions about acetaminophen — the active ingredient in Tylenol — during pregnancy. Danish researchers studied 302 three-month-old girls whose mothers tracked acetaminophen use during pregnancy. On average, girls exposed in the womb had smaller ovaries and uteruses, fewer ovarian follicles and lower levels of some reproductive hormones. Among babies exposed before 17 weeks, ovarian volume was about 40% smaller. Researchers also examined a separate group of 1,210 girls followed from infancy through adolescence and found some differences in reproductive-organ size associated with prenatal exposure.

Those findings sound alarming, but they do not show that acetaminophen damaged the girls’ reproductive systems or will affect their future fertility. This was observational research; many exposed girls had measurements similar to unexposed girls, and researchers don’t yet know whether the differences they observed will matter later in life.

That’s particularly important given the ongoing debate over acetaminophen during pregnancy. Current guidelines continue to consider it an option for treating pain and fever during pregnancy, and untreated high fever or severe pain can carry risks of their own. The researchers themselves said women who have taken acetaminophen during pregnancy shouldn’t be alarmed by these findings.

Bottom line: This is a finding worth following, not evidence that Tylenol causes infertility. Researchers will need to follow these girls for years to determine whether the differences they observed actually affect fertility or reproductive health.

🫄 News You Can Use

What can an “egg count” fertility test actually tell you? 

Blood tests measuring anti-Müllerian hormone, or AMH, are increasingly marketed to women as a way to check their fertility or find out how many reproductive years they have left. In a new analysis, researchers reviewed records from nearly 1.4 million Australian women and found AMH testing increased more than 13-fold between 2011 and 2021. The increase was especially pronounced among women ages 30 to 34, whose testing rate rose nearly 20-fold.

AMH can help estimate ovarian reserve, essentially the remaining supply of eggs, and can be valuable in fertility treatment. But it cannot reliably tell an otherwise healthy woman whether she’ll become pregnant naturally, how quickly she’ll conceive or precisely how much reproductive time she has left. Age remains a much stronger predictor. 

That’s where researchers see a potential problem. A low result could cause someone to unnecessarily panic or change plans about when to have children, while a normal result could create false reassurance about fertility.

Bottom line: AMH tests are useful but should not be marketed as “fertility crystal balls.” If you’re considering the test, the most important question may be what you’re hoping the result will tell you — and whether AMH can actually answer it.

Have your say

To view this week’s poll, read this newsletter in a web browser.

Last week, we asked if you’ve ever been asked to pay for medical care before receiving it. Here’s what you said:

What else I’m watching

Get in Touch

That’s all for this week. If you have questions about a health-related story or study you saw in the wild, send me an email at jcraig@san.com or share them below.
— Jess