Five-cup coffee warning for older women's bones rests on a faint curve

In about 8,000 US women over 65, followed for roughly eight years, coffee drinkers and non-drinkers showed no clear difference in hip bone density. Above five cups a day a fitted curve drifted lower, but not clearly beyond chance, and the paper doesn't say how many women drank that much. Tea drinkers' hips were 0.003 g/cm² denser, less than a year's worth of normal age-related loss. Fractures weren't studied.

5 October 2026 · How we rate evidence

Between their early 70s and about 80, the women in this study lost roughly 0.004 grams of mineral per square centimetre of hip bone each year. Keep that number in mind, because the tea result is smaller than one year of it, and the coffee result that made headlines isn't clearly there at all.

The claim going around. "Too much coffee may weaken bones, study finds," ScienceDaily wrote on 2 September, reprinting a Flinders University press release. The release says that "women who drank more than five cups a day had lower BMD," or bone mineral density, and that tea "could offer a simple way to support bone health." A second wave followed in September. Morning Overview wrote that heavy drinkers had "measurably lower bone mineral density," and The Economic Times and Chelsea News NY repeated the five-cup line. The paper itself came out in November 2025.

What they did. Ryan Yan Liu and Enwu Liu at Flinders University in Adelaide reused data from the Study of Osteoporotic Fractures, a US study that enrolled 9,704 women aged 65 and over in the late 1980s. At four visits over about a decade, the women said whether they "currently" drank regular coffee or tea and how many cups a day, and had their hips scanned with DXA, the low-dose X-ray used in osteoporosis clinics. Their answers didn't include cup size or strength.

The team compared drinkers with non-drinkers at each visit, adjusting for age, body size, smoking, illnesses, lifetime alcohol and exercise, diet, hormone use and starting bone density. The last of the four visits was named the main endpoint. Separately, they fitted a smooth curve of bone density against cups per day.

What they found. Coffee first, because that is the headline. Across the four main visits, women who drank coffee had essentially the same hip bone density as women who didn't. At the last of them the difference at the whole hip was 0.000 g/cm², with a range from 0.003 lower to 0.002 higher. That rules out a difference of 0.01 g/cm², which we take as the smallest that would matter, about two and a half years of normal loss; the paper sets no such figure. A range that narrow is this study's strength: thousands of women, scanned the same way at repeated visits, with many other habits allowed for. One result points the other way. When the authors added two later visits, which many women missed, coffee drinkers' femoral-neck density at the last of them was 0.011 g/cm² lower (range 0.003 to 0.019 lower). The authors say to read it cautiously because so much data was missing.

The five-cup warning comes from the curve. It stays flat to about two or three cups, then slopes gently downward; by ten cups it sits roughly 0.005 g/cm² below non-drinkers, by our reading of the chart. But the shaded band covers the no-difference line all the way along, and the paper's own tests for a bend in the curve were far from significant (0.60 and 0.24, where 0.05 is the usual cut-off). The authors write that more than five cups was "visually suggested to be potentially associated with lower BMD." The paper is cautious here; the Flinders press release was not. The paper also never says how many women drank more than five cups, so we don't know how many people that end of the curve rests on.

Tea was the one main result clear of chance. At the last main visit, tea drinkers averaged 0.718 g/cm² at the whole hip against 0.715 for non-drinkers. That gap was not clear at the three earlier visits, and at the femoral neck, the other hip site measured, tea drinkers were slightly lower, though not clearly.

The release also says women with higher lifetime alcohol intake "experienced stronger negative effects" from coffee. In the paper, that was one of 26 comparisons across subgroups. Among women with more than 25 drink-years of lifetime alcohol (25 is, for example, half a drink a week for 50 years), coffee went with 0.002 g/cm² lower density at the femoral neck, with a range reaching zero, and not at the whole hip. The authors call their subgroup results "likely underpowered" and "primarily hypothesis-generating."

What it doesn't show. The study didn't measure fractures, only bone density, which predicts fractures but isn't the same thing. A 0.003 g/cm² gap is about 2% of the normal spread between women. Taking at face value a 1996 pooled analysis of how hip density predicts hip fracture, that would mean around 2% fewer hip fractures in relative terms, if the gap were caused by tea at all (our rough arithmetic). The authors say the tea difference is "below the threshold typically considered clinically meaningful."

Coffee and tea were self-reported, with no cup sizes. Alcohol and exercise were recorded once, at the start. And by the last main visit 42% fewer women were being scanned, because they had died or missed scans; if heavy drinkers with fragile bones were more likely to die first, the survivors' comparison would be skewed. Differences this small can come from any of these.

Chelsea News NY described the cohort as a "Swedish and Dutch" study; it is American. It also said adding milk to coffee "seemed to mitigate" its effects. Milk wasn't studied here; the authors mention it from an older review of how caffeine affects calcium.

What it means beyond the study. For people, the question is whether what's in an older woman's cup changes her chances of breaking a hip. This study can't answer that directly. What it does show is that coffee drinking made no clear difference to these women's hip bone density at the four main visits, though one late visit with much missing data showed lower femoral-neck density in coffee drinkers. For five cups and up, the honest answer is that nobody has shown a difference yet.

Why it's still interesting. Few studies have tracked the same women's drinking habits and bone scans repeatedly over this long, which makes the coffee comparison more informative than a single snapshot would be. And a tea signal, however small, lines up with earlier studies in British and Australian women, though other studies and a pooled analysis found no link. Because the cohort also recorded fractures, the question that matters could be tested in the same data.

Open questions.

Do the tea and coffee differences show up in broken bones? The study recorded fractures for about 20 years, so the same women could be compared on hip fractures by what they drank. If fractures don't differ, nobody needs to change their drink for their bones; if they do, bone density was hiding something.

Is there a heavy-coffee threshold at all? Report how many women drank five or more cups and compare them directly with moderate drinkers, ideally pooling cohorts with more heavy drinkers. That would tell heavy drinkers whether cutting back is worth it for their bones.

Does alcohol change coffee's link with bone? Test the interaction, decided in advance, in another cohort that records current drinking, not a lifetime estimate. If it holds, advice could be aimed at a specific group; if not, it was one of 26 chances.

A note on timing. The paper was published online on 23 November 2025, and first coverage followed in December. ScienceDaily's reprint of the release arrived nine months later, in September 2026, and set off a second wave. The study hadn't changed.