In the December 2007 edition of Clinical Endocrinology, Brambilla et al. state that greater than one but not necessarily greater than three blood samples are required to characterize an individuals hormone levels.

They have calculated intra-individual variation for serum testosterone (total, free and bioavailable), dihydrotestosterone, sex hormone-binding globulin, luteinizing hormone, dehydroepiandrosterone, dehydroepiandrosterone sulphate, oestrone, oestradiol and cortisol in a bid to calculate the number of blood samples needed to characterize an individuals hormone levels. Previous published data were collated on small samples so Brambilla et al. performed a prospective study on 132 men aged 30–79 years, of varying ethnicity, all community dwelling, generally healthy, randomly chosen respondents to the Boston Area Community Health Survey (n = 2301). Exclusion criteria included hypogonadism, drug treatment known to alter hormone levels, liver/kidney disease or a compromised immune system.

Paired blood samples were taken (20 minutes apart) after one to three days, three months and six months, all within four hours post awakening. Samples were transported on ice and stored at −70°C post centrifugation. Hormones were measured by immunoassay. Intra subject standard deviations were calculated under four sampling schemes, after transforming measures to eliminate the contribution of the hormone level to the intra-individual variation. An average of the paired samples 20 minutes apart was used to minimize the effect of pulsatile release. Contributions of biological variation and assay variation to total intra-individual variation are calculated with biological variation generally contributing more to the total. They find that differences of 18–28% between two hormone measurements on a subject can be expected around half of the time. The difference will exceed 27–54% about a quarter of the time.

The main uses of these data are for ambiguous clinical situations such as:

Determining how far a measured testosterone level is above/below the threshold for hypogonadism, to conclude that an individual is reliably above/below the threshold.

Determining if the difference between two measurements is a change in hormone concentration compared with fluctuation around a steady state mean.

Clinical monitoring in trials of hormone replacement therapy.

They point out the need to use their protocols for blood drawing and time of sampling as well as analytical methods for their figures to be valid. This is important, as the diurnal variation may be greater than the intra-individual standard deviations calculated.

It would be interesting to see these data produced using the more accurate methodologies of LC/MS-MS or GC/MS, particularly as these technologies become more widespread.