Van Cauter · JAMA 2000 · Pooled cross-sectional study · n=149

Age-related changes in slow wave sleep and REM sleep and relationship with growth hormone and cortisol levels in healthy men.

Cited 779 times in the scientific literature.

Level 4 - case-series / case-control

Pooled cross-sectional observational study of age cohorts.

PubMed 10938176 · doi:10.1001/jama.284.7.861 · record verified 2026-08-26

What was done

Data were pooled from studies conducted between 1985 and 1999 across 4 laboratories, examining 149 healthy men aged 16 to 83 years (mean BMI 24.1 kg/m²) without sleep complaints or endocrine, psychiatric, or sleep disorders. Subjects underwent 24-hour plasma sampling for growth hormone (GH) and cortisol alongside polygraphic sleep recordings to evaluate age-related alterations in sleep duration, sleep stages, and hormone profiles.

What was found

Deep slow wave sleep (SWS) decreased from 18.9% (SEM 1.3%) in early adulthood (16–25 years) to 3.4% (1.0%) in midlife (36–50 years), replaced by lighter non-REM sleep without immediate loss of REM sleep. From midlife to late life (71–83 years), SWS showed no further significant decline, but awake time increased by 28 minutes per decade, while light non-REM sleep (-24 minutes/decade; P<.001) and REM sleep (-10 minutes/decade; P<.001) declined. Concurrently, GH secretion fell by 372 µg per decade from early adulthood to midlife (P<.001) and slowed to a decrease of 43 µg per decade from midlife to late life (P<.02); GH was significantly associated with SWS independent of age (P<.001). Evening cortisol rose by 19.3 nmol/L per decade (P<.001), becoming significant only after age 50, with a trend linking lower REM sleep to higher evening cortisol (P<.10).

Why it matters

The findings demonstrate that sleep-endocrine aging occurs in distinct chronological phases in men, with SWS and GH loss occurring early in adulthood and REM loss, fragmentation, and cortisol elevation emerging in late life.

Limits

The study is cross-sectional rather than longitudinal, precluding direct within-person trajectory tracking or causal inference. The cohort included exclusively healthy men, limiting generalizability to women and clinical populations. Measurements were also aggregated across four different laboratories over a 14-year period.

Cited by