Instead, decreases in the male sex hormone may simply be a sign of overall health status, which also dips with age, researchers say.

The findings come in the wake of surging interest in testosterone's role in men's sex drive, their mood and thinking, and even early death.

While drug companies refer to "low T" as a treatable medical problem affecting millions of men, critics say they are trying to make a buck by turning normal aging into a disease.

There is no doubt that very low testosterone levels affect the body and dampen sex drive.

But where to set the threshold between normal and abnormal levels is still an open question, whose resolution is not made easier by natural variation in the male sex hormone.

It dwindles with age, for instance, and fluctuates between countries, races and individuals -- even from one hour to the next.

The new study, published in the Journal of Clinical Endocrinology & Metabolism, takes an in-depth look at earlier research into the link between testosterone and death.

"It's the ultimate endpoint," said Andre Araujo of the New England Research Institutes, who worked on the study.

So far, he said, most but not all research shows men with low testosterone die before their peers with higher hormone levels.

So he and his colleagues gathered all the medical literature they could find on the topic, including a report of their own that had found no link. They only chose studies of men living in the wider community, as opposed to hospitals or long-term care, to try to avoid the muddying influence of severe illnesses on the data.

CORRELATION, NOT CAUSATION

Based on 11 studies, which tracked more than 16,000 middle-aged men for between five and 17 years, the researchers did find signs of trouble: it appeared that men in the lowest third of testosterone levels had a 35 per cent higher chance of dying than men in the highest third.

"There is certainly a link," said Araujo. "But the important thing is that the estimates are a little bit off from one another."

That means they aren't really comparable, statistically speaking. Digging deeper, Araujo found that shorter studies were more likely to find a connection, as were those including mainly older men.

On the other hand, low testosterone levels did not seem to be associated with early deaths in younger men.