A site willing to tell you that one of its numbers is weaker than it looks is a site you can believe about the numbers that are strong.
You notice it getting out of the car. A hesitation in the knee that was not there before. A shoulder that complains for days after a session that used to cost you nothing. In the same stretch of years your energy dropped, your waist grew, and someone told you to check your testosterone. So the obvious question arrives. Are the joints part of the same story?
A 2023 analysis of 10,439 American adults produces an answer that looks spectacular until you read the rest of the table.
Where that number comes from
The researchers used NHANES, the US national health and nutrition survey, covering 2013 to 2016. They started with 20,146 eligible people, removed 5,380 with no testosterone measurement and 4,327 with no arthritis data, and analyzed the 10,439 who were left. Mean age was 47.25 years. Just under half, 48.11 percent, were men. And 26.97 percent reported arthritis.
The raw contrast is wide. Adults without arthritis averaged 233.91 ng/dL of total testosterone. Adults with arthritis averaged 163.67. Sort the whole sample into quartiles and the top quartile carried an odds ratio of 0.49, confidence interval 0.31 to 0.76, against the bottom quartile. That is where 51 percent comes from.
Now watch it fall apart
The same paper also ran testosterone as a continuous variable through three models, one built on top of the next. This is the part worth your attention.
- Model 1, no adjustment at all: odds ratio 0.87, interval 0.85 to 0.89. Strongly protective.
- Model 2, adjusted for age, sex and race: 0.93, interval 0.90 to 0.97. Weaker, still there.
- Model 3, adjusted for age, sex, race, education, marital status, BMI, alcohol, smoking, hypertension, diabetes, cardiovascular disease, estradiol, SHBG and income to poverty ratio: 1.02, interval 0.97 to 1.07.
Read that last line slowly. The association did not merely shrink. It vanished.
What the adjustment took away
Look at who the arthritis group actually was. Mean age 59.90 years, against 43.36 in the group without arthritis. A BMI of 30 or above in 55.24 percent of them, against 34.59 percent. Hypertension in 55.57 percent, against 25.58 percent. Cardiovascular disease in 18.56 percent, against 5.01 percent. Women made up 61.69 percent of them, against 48.27 percent.
Every one of those things lowers testosterone, or travels with something that does. Older, heavier, sicker people have more arthritis and less testosterone. Model 3 asks a harder question. Once you already know a person's age, sex, weight and blood pressure, does testosterone tell you anything extra about their joints? In this dataset, no.
Then why did the quartile result survive?
A fair objection, because in that same fully adjusted model the quartiles still moved. Q2 came in at 0.85, interval 0.72 to 1.00. Q3 at 0.53, interval 0.35 to 0.79. Q4 at 0.49, interval 0.31 to 0.76, with a p for trend of 0.0327. It looks like a contradiction. It is really a warning about what quartiles do.
Look at the boundaries. The bottom quartile runs from 0.52 to 18.77 ng/dL. The top quartile runs from 378 to 2000. Men and women were pooled into the same four brackets. Those are not two doses of the same hormone in the same kind of person. They are largely two different populations. The continuous analysis, which uses each person's real value instead of a bracket, is the stricter test, and it showed nothing.
And what counted as arthritis
Arthritis here was self-reported. Participants were asked whether a doctor had ever told them they had it. In a validation study the authors cite, self-report agreed with clinically confirmed osteoarthritis 81 percent of the time. Among those who said yes, 52.00 percent called it osteoarthritis, 14.34 percent rheumatoid, 1.17 percent psoriatic, 9.98 percent other, and 22.51 percent did not know or would not say.
Those are different diseases. One is mechanical wear on cartilage. Others are the immune system attacking a joint. Asking one hormone to explain the whole pile is a great deal to ask of a single blood test.
Worth saying plainly. This was a cross-sectional survey. Testosterone and arthritis were captured at the same moment, so nothing in it can tell you which came first. Pain reduces activity, lost activity adds fat, and added fat lowers testosterone. That chain runs in the opposite direction to the headline, and this design cannot rule it out. The authors state the limitation themselves.
What is left standing
Something real, just smaller than advertised. People with arthritis in this sample genuinely had lower testosterone. The signal was strongest in women and in people with a BMI of 30 or above, and the interaction by sex and by BMI reached statistical significance. That is a lead worth chasing with a study that follows people over time. It is not evidence that moving a number repairs a joint. This paper did not test that question at all.
If your own reading is low and your knees hurt, both facts deserve an explanation. But they have to be read against your age, your weight, your blood pressure and your history, not against a quartile in a national survey. That comparison is the work a consultation exists to do.
Why you are reading this on a site that sells nothing
A marketing page would have stopped at 51 percent. It is a good number. It comes from a real journal. Quoting it is not lying.
We took it apart because the honest version is more useful to you. You are not really deciding about testosterone and cartilage today. You are deciding whether the person explaining hormones to you is describing the evidence or selling from it. The test is simple and it travels. Watch what they do with a finding that does not hold up.