The J-curve is a pattern that appears when you plot a health behavior or measurement against its associated risk. Instead of risk climbing steadily as the behavior increases (a straight line), the curve dips down first, forming a shape like the letter J. A moderate amount of the behavior is linked to the lowest risk, while both too little and too much are linked to higher risk. This pattern shows up across blood pressure, sleep, exercise, body weight, and alcohol research, though its validity is debated in some of those areas.
How the J-Curve Works
In a simple linear relationship, more of something always means more risk (or less risk). The J-curve breaks that pattern. Imagine graphing risk on the vertical axis and dose on the horizontal axis. Starting from zero, risk initially drops as the dose increases, hits a low point (called the nadir), then climbs sharply as the dose keeps rising. The result is a line shaped like the letter J, with the bottom of the curve representing the sweet spot.
The biological concept behind this is called hormesis, from the Greek word for “rapid motion.” At low doses, a stressor triggers the body’s repair and defense systems, producing a net benefit. At high doses, the harmful effects overwhelm those defenses. When you overlay these two opposing processes, activation and inhibition, the combined curve is J-shaped. This framework helps explain why moderate stress on the body (from exercise, for example) is beneficial while extreme stress is harmful.
A closely related shape is the U-curve, where both extremes carry similarly elevated risk. In practice, researchers sometimes use the terms interchangeably, though a true J-curve has one tail higher than the other.
Blood Pressure: The Most Studied Example
The J-curve debate in medicine started with blood pressure treatment. In the early days of blood pressure medication, doctors assumed that lowering blood pressure as far as possible would always reduce heart attack risk. But in the late 1970s, researcher Stewart reported that patients whose diastolic pressure (the bottom number) dropped below 90 mmHg actually had more heart attacks than those whose readings stayed between 100 and 109. Cruickshank then described the pattern formally, finding the lowest heart attack rate at diastolic levels of 85 to 90 mmHg.
More recent evidence from Johns Hopkins confirms there is increased risk of coronary events when diastolic blood pressure falls below 60 to 70 mmHg. This makes intuitive sense: blood pressure exists to push blood through your organs, so a pressure that’s too low starves the heart, brain, and kidneys of oxygen. The evidence for a J-curve in systolic pressure (the top number) is less clear, and some high-risk patients may benefit from systolic levels as low as 120 mmHg. For kidney health, the lowest risk of protein leaking into the urine occurs at a systolic reading around 110 and a diastolic reading around 70.
Sleep Duration and Mortality Risk
Sleep follows one of the clearest J-curves in health research. A large meta-analysis of prospective studies found that people who consistently sleep less than 7 hours per night have a 12% greater risk of dying from any cause compared to those sleeping 7 to 8 hours. People sleeping more than 8 or 9 hours per night face a 30% greater risk, making the long-sleep side of the curve steeper than the short-sleep side, a classic J-shape rather than a symmetrical U.
People who consistently sleep 5 hours or fewer are considered a higher-risk group. On the other end, routinely sleeping 9 or more hours is associated with even greater risk than short sleep, though this may partly reflect underlying illness driving longer time in bed. The current evidence suggests that habitually sleeping 6 to 8 hours carries no measurable harm for adults.
Exercise and Immune Function
In the 1990s, exercise physiologist David Nieman proposed a J-curve for the relationship between exercise intensity and susceptibility to upper respiratory infections like colds. Moderate exercise improves immune function above sedentary levels, but extreme training suppresses it.
Nieman’s survey of Los Angeles Marathon participants found that runners training more than 97 kilometers per week (about 60 miles) had roughly double the risk of catching an infection compared to those training 32 kilometers or fewer per week. Runners who actually competed in the marathon experienced more respiratory illness afterward than trained runners who sat the race out. Intense exercise temporarily drops the count and activity of natural killer cells, a key part of the immune system. This suppression normally resolves within 24 hours, but without adequate recovery between sessions, the immune dip can become chronic.
Other factors layer on top: age, nutrition, psychological stress, and baseline fitness all influence where you fall on the curve. The J-shape here doesn’t mean intense exercise is always harmful, but it does mean the relationship between training volume and health isn’t a simple “more is better” line.
Body Weight and Mortality
BMI and death risk also follow a J-shaped or U-shaped pattern, though the lowest-risk zone may surprise you. A 2024 meta-analysis in the Journal of Clinical Medicine found the lowest mortality in the BMI range of 25 to 30, which is technically classified as “overweight.” Risk climbed sharply above a BMI of 35.
The curve shifts with age and health status. Among elderly populations, a BMI below 20 carried the highest risk, with no data available above 35. Among people with diabetes, the greatest mortality risk was also at the low end (below 20), though risk increased again above 35. This suggests that carrying a small amount of extra weight may be protective in older age and chronic illness, a finding sometimes called the “obesity paradox.”
Alcohol: A J-Curve Under Fire
For decades, the alcohol J-curve was perhaps the most famous version of this concept. Older meta-analyses suggested that 1 to 2 drinks per day was associated with lower mortality than abstaining entirely, with risk rising steeply above 4 drinks per day for men and 2 for women. The protective effect appeared to peak at about 2 non-heavy drinking occasions per week.
This finding is now heavily contested. The central problem is the “sick quitter” effect: many studies grouped former drinkers, people who may have quit because of illness, together with lifelong non-drinkers in the “abstainer” category. This made abstainers look sicker than they actually were, artificially inflating the apparent benefit of moderate drinking. Over 70% of systematic reviews on alcohol and mortality published through March 2022 failed to exclude former drinkers from their reference group.
When researchers used genetic analysis (Mendelian randomization) to test whether moderate alcohol truly causes lower heart disease risk, the U-shaped association disappeared. The one review rated at low risk of bias found a straightforward dose-response: every step up in consumption, from low to medium to high, increased mortality risk, with no protective dip at all. Large-scale studies have led to a blunt public health conclusion from the Global Burden of Disease project: “the safest level of drinking is none.” The alcohol J-curve, long used to justify a nightly glass of wine, increasingly looks like a statistical artifact rather than a biological reality.
Why the J-Curve Matters
The J-curve is important because it challenges the assumption that health relationships are linear. If you think lower blood pressure is always better, you might over-treat someone into a dangerous range. If you think more exercise always helps, you might ignore warning signs of overtraining. The curve forces a more nuanced question: not just “is this good or bad?” but “how much is optimal?”
It also carries a built-in caution. Not every observed J-curve reflects true biology. As the alcohol debate demonstrates, study design flaws can create the illusion of a sweet spot where none exists. The strongest J-curves, like those for blood pressure and sleep, are backed by clear physiological mechanisms. The body needs a minimum blood pressure to perfuse organs, and it needs a minimum amount of sleep to repair itself. When a J-curve lacks that kind of mechanistic explanation, it deserves extra scrutiny.