Measuring Serum Melatonin in Epidemiologic Studies

Circadian rhythm disruption has recently been classified as a probable carcinogen to humans (1) and has been associated with an increased risk for cancers of the prostate (2-7), breast (8-11), and colon (12), and for non–Hodgkin lymphoma (13). Although the underlying mechanism explaining these observations is unknown, it has been suggested that melatonin, a hormone synthesized and secreted by the pineal gland in response to low-light conditions, may mediate the circadian rhythm–cancer relationship (1). Circulating melatonin exhibits a circadian rhythm with highest levels at night, moderate levels in the morning, and substantially lower levels during the afternoon (14). Light exposure can reduce the duration of melatonin secretion and subsequent circulating melatonin levels (15). As such, melatonin may be a key biological intermediary of chronodisruption (16).

Traditionally, in sleep disorder studies, melatonin is measured in urine [as 6-sulfatoxymelatonin (aMT6s)] because overnight or morning first-void urinary aMT6s reflects melatonin levels during the night when melatonin levels are at their highest (17). However, in most large-scale prospective epidemiologic studies, overnight void urine is not routinely collected. In addition, most cohort studies collect blood samples during the day when melatonin levels are quite low. Therefore, reliable assays with high sensitivity are needed to quantify accurately the circulating levels of melatonin. For these reasons, few epidemiologic studies have measured serum melatonin levels and tested the hypothesis that altered melatonin levels are associated with increased cancer risk.

To extend the investigation of the circadian rhythm/melatonin hypothesis to epidemiologic studies, we conducted a methodologic study to measure melatonin levels in serum, plasma, and urine samples, and to assess the reproducibility of a currently available melatonin RIA and the feasibility of applying this assay to epidemiologic studies. Further, we measured serum levels in Caucasian, African-American, African, and Chinese men to assess ethnic differences in melatonin levels.

To choose an acceptable assay for the study, we evaluated several commercial melatonin kits and concluded that the Buhlmann RIA kit (ALPCO) with a preceding extraction step is acceptable for measuring serum melatonin. This method was preferred over the direct assay because the extraction step can minimize interference from melatonin metabolites as well as remove lipids and other components that may be present in serum, which would result in overestimating melatonin levels. The Buhlmann RIA kit also uses the antibody raised by Kennaway and colleagues (18, 19), which has been validated by gas chromatography mass spectrometry (20). We used 60 samples to optimize the assay for both the standard (1.0 mL) and reduced (0.5 mL) serum volumes. Briefly, melatonin is purified from serum samples using reversed phase columns, dried, and reconstituted in assay buffer; melatonin standards for the standard curve were processed similarly. Samples were then incubated with α-melatonin antibody and iodinated melatonin for 20 h, precipitated for the antibody-bound fraction using a solid-phase second antibody for 15 min, and centrifuged. The unbound fraction was aspirated and the antibody-bound fraction of iodinated melatonin was counted using a γ counter.

Urinary aMT6s [expressed as ng/mg Creatinine (Cr)] were measured using the standard (1 mL) protocol by a competitive ELISA kit with a capture antibody technique (Buhlmann) as previously described (21). The assay was done on a microtiter plate coated with polyclonal antibodies specific for rabbit immunoglobulin. Urinary Cr was measured colorimetrically using a standard procedure.

For the study, only blood and urine samples collected from healthy men without a history of cancer were included. All samples were collected for our other epidemiologic studies and processed following standard procedures; serum samples were derived from blood that was processed within 4 h of collection and stored at −70°C, and 24-h urine was collected, processed, and stored similarly.

Initially, reproducibility of the serum melatonin assay using the standard 1.0-mL volume was assessed with triplicate samples from two pooled serum samples; pooled serum samples were created by combining serum collected before 12 a.m. from multiple Chinese men. Subsequently, reproducibility of the refined (0.5 mL) assay was assessed using replicate morning serum samples from 14 Caucasian men (two to four replicates per sample). Reproducibility of urinary aMT6s assays using the standard protocol was assessed with triplicate urine samples from two Chinese men.

For comparison of urinary aMT6s and serum melatonin levels, urine-serum paired samples collected from 49 Chinese men were used for this purpose, applying the standard protocols (1-mL sample volumes). All serum samples were derived from blood collected before noon. Urine samples were collected for the same participants during a 24-h period after blood collection.

For comparison of melatonin levels in serum and plasma, serum-plasma paired samples from the same blood draw from 20 Caucasian men (13 morning blood draws and 6 afternoon blood draws) were measured using the refined (0.5 mL) assay.

To determine the serum melatonin profile by blood collection time, we measured serum melatonin levels in 43 Caucasian American men with blood drawn from one of four time blocks: (a) before 9:00 a.m. (n = 18), (b) between 9:00 a.m. and 10:59 a.m. (n = 15), (c) between 11:00 a.m. and 12:59 p.m. (n = 5), and (d) 1:00 p.m. and later (n = 5). To evaluate melatonin levels in various racial/ethnic groups, we measured morning serum melatonin in 50 Chinese, 35 Caucasian American, 11 African-American, and 30 Ghanaian men. For subjects with replicate measurements, the mean of the replicate measurement was used for analysis; all replicate samples were derived from the same blood draw.

We used two measures of reproducibility to determine the precision of the melatonin assays: the coefficient of variation (CV) and the intraclass correlation coefficient (ICC). The CV, expressed as a percent, is calculated as 100 times the ratio of the SD to the mean: CV = 100 × σ/μ. Small CVs mean less dispersion from the mean and thus better reproducibility. The ICC, expressed as a percent, is calculated as 100 times the ratio of the variance associated with subjects (σα²) to the sum of all variances [associated with subjects and other factors such as measurement error (σ_α²+σ_ϵ²): ICC = 100 × σ_α²/(σ_α²+σ_ϵ²)]. In our experience, the useful assays have CVs that are <20% and ICCs of >80%; better assays have ICCs of >90%. Pearson's correlation was used to measure the correlation between the paired serum and urine samples as well as between the paired serum and plasma samples.

Table 1 shows the means, SEMs, and medians of urine aMT6s, serum melatonin, and plasma melatonin levels in assays using either 1.0 or 0.5 mL of serum. The sensitivity of the melatonin RIA using both 1.0 and 0.5 mL serum volumes was 0.5 pg/mL. Urinary aMT6s levels among 49 men ranged from 0.49 to 12.36 ng/mg Cr, with a mean of 3.89 ng/mg Cr (SEM, 2.27 ng/mg Cr). In the same 49 men, melatonin concentrations measured by the assay using 1.0 mL of serum ranged from 0.50 to 10.44 pg/mL, with a mean of 3.43 pg/mL (SD, 2.41 pg/mL). Reproducibility for both urinary aMT6s and serum melatonin assays seems to be good, with mean CVs of 3.5% for the urine assays and 6.9% for serum assays. Reproducibility for serum melatonin assays using 0.5 mL of serum was similar to that using 1.0 mL of serum (mean CV, 6.6%) and ICCs were >97% for all assays. Serum and plasma levels were similar (plasma: mean, 3.44 pg/mL; range, 1.10-13.22 pg/mL; serum: mean, 3.46 pg/mL; range, 1.25-12.42 pg/mL).

Although direct comparisons between absolute concentrations of urinary aMT6s and serum melatonin are difficult, due to differences in units of measure (ng/mg Cr versus pg/mL, respectively), there was a positive and significant correlation between urine aMT6s and serum melatonin levels (r = 0.46; P = 0.008; Fig. 1A). There was also a strong correlation between plasma and serum melatonin levels (r = 0.97; P < 1 × 10⁻⁴; Fig. 1B).

Melatonin levels were detectable in serum throughout the day (Fig. 2A), although samples collected before 9 a.m. had levels that are five times higher than those collected after 11 am (mean, 10.98 pg/mL; range, 1.61-26.67 pg/mL; and mean, 1.96 pg/mL; range, 1.25-4.08 pg/mL, respectively); these differences remained significant after adjusting for age (data not shown). In addition, serum melatonin was detectable in all four racial groups, with Chinese men (mean, 3.42 pg/mL) having lower serum levels, whereas Caucasian American (8.26 pg/mL), African-American (6.67 pg/mL), and Ghanaian (8.61) men had similar levels (Fig. 2B).

In this methodologic study, we showed that melatonin can be detected reliably in serum, plasma, and urine samples collected in the morning from healthy men by RIAs with preceding extraction. In addition, there is good correlation between urine and serum levels as well as plasma and serum melatonin levels, suggesting that these biological specimens may be used for epidemiologic studies to quantify melatonin levels. We further showed that serum melatonin is detectable throughout the day, although at much lower levels in the afternoon, and in men of different ethnic backgrounds. These data suggest that it is possible to measure melatonin in biological samples, in particular in serum samples, to test the emerging hypothesis that circadian rhythm may play a role in cancer etiology.

In our study, the reproducibility of the serum assay was good but is considerably lower than that of the urine assay, due largely to the lower concentration of melatonin in serum. Although the absolute melatonin levels measured in urine and serum differ (ng/mg Cr versus pg/mL, respectively), the strong correlation between these two specimen types and between serum and plasma (r > 0.46) suggests that when ranks or quartiles are used to categorize melatonin levels, measurements from each of these sample types are likely to distribute subjects into virtually identical classes, thereby yielding similar risk estimates for assessing the relationship between melatonin and cancer. Although we showed that the assay using 0.5 mL of serum has similar reproducibility as the 1.0-mL assay, for large-scale prospective studies, an even smaller amount is desirable.

Similar to data from previous studies that measured melatonin in the same subjects throughout the day (22-26), in our study, higher melatonin levels were found in subjects with blood collected in the morning hours compared with those collected in the afternoon (up to a 5-fold difference). Although larger studies are needed to confirm these findings and to characterize further the determinants of circulating melatonin (e.g., age, gender, body size, life-style factors, metabolic factors, and hormones, etc.), these data underscore the importance of using morning samples for the measurement of melatonin and the need to control for differences in time of blood collection in epidemiologic studies.

Emerging data suggest that circadian rhythm disruption may play a role in the etiology of several cancers. Recently, the IARC concluded that “shift-work that involves circadian disruption is probably carcinogenic to humans” (1). For example, data from observational studies suggest that increased risks for prostate cancer are seen for men with exposure to light at night, occupations that can disrupt their circadian rhythms, such work on a rotating shift or as an airline pilot, and shorter sleep duration (2-7). Similar findings are also reported for breast cancer (8-11), colorectal cancer (12), and non–Hodgkin lymphoma (13). Genetic studies have shown that variants of circadian genes, which dictate the endogenous clock, may be associated with increased risk for prostate cancer (27, 28), breast cancer (29, 30), and non–Hodgkin lymphoma (31, 32). Together, these data suggest a potential role of circadian rhythms in several cancers, warranting further investigation.

It has been proposed that melatonin may be a key biological intermediary of chronodisruption (16). It is possible that cancer risk associated with chronodisruption may be due to low serum melatonin levels. Currently no epidemiologic study has addressed this hypothesis, due largely to the lack of reliable assays with high sensitivity to detect relatively low melatonin levels in serum collected during daylight. Although future studies are needed to determine if our findings are generalizable to women, our data suggest that with improved assays, it may be possible to investigate the role of serum melatonin and the emerging circadian rhythm hypothesis in cancer etiology in epidemiologic studies.

No potential conflicts of interest were disclosed.

We thank Dr. Frank Z. Stanczyk and Lilly Cheng of the University of Southern California for melatonin assays, Dr. Karen Stewart and Ann Truelove of Westat for the help with study management, Dr. B.J. Stone the for expert editorial assistance, and the study participants for their contributions to making this study possible.

Grant Support: Intramural Research Program of the Division of Cancer Epidemiology and Genetics, National Cancer Institute, NIH, Department of Health and Human Services.