Introduction

Over the last decade an endocrine function of the adipose tissue has emerged. Among adipose tissue-derived signaling factors, the novel adipokines omentin-1 and retinol-binding protein 4 (RBP4) are of special interest for their possible role in systemic metabolic regulation. Omentin-1 is a 35 kDa glycoprotein [1] which is mainly expressed in visceral adipose tissue [2]. Inverse associations of omentin-1 with body fatness, metabolic syndrome, type 2 diabetes and CVD were reported from cross-sectional human studies [3–6]. RBP4 is a 21 kDa protein with the highest expression rate in the liver, and the second highest rate in adipose tissue [7]. Its main function is the extracellular transport of retinol. More recently, RBP4 was further characterized as novel adipokine [8] which was directly associated with components of the metabolic syndrome [9–12]. Furthermore, RBP4 levels were also elevated in patients with chronic kidney diseases [13], raising the question to which extent the direct association of RBP4 levels with prevalent type 2 diabetes might be explainable by early renal complications [14]. Still, prospective studies have reported longitudinal associations of RBP4 with the incidence of type 2 diabetes [15, 16] and cardiovascular disease (CVD) [17] in humans. Thus, associations of omentin-1 and RBP4 with cardio-metabolic diseases are suggested but incompletely understood and active research regarding these two adipokines can be awaited.

In observational studies, however, the assessment of biomarker concentrations often relies on blood samples from a single time point. Hence, the interpretability of such measurements with regard to biological differences between study participants depends on the variability of the biomarker concentrations within individuals. Intraindividual variation might depend on environmental factors such as season, daytime, and dietary exposures or endogenous factors such as health status, menstrual cycle, and stress level among others, which are typically not controlled in observational studies. Therefore, the STROBE-ME guidelines explicitly recommended reporting on the long term variability of repeated measurements within persons for the biomarkers applied in epidemiological studies [18]. The biological variation of circulating concentrations of RBP4 and omentin-1 within individuals has not been systematically investigated in large samples yet. Hence, our study aimed to evaluate the within person reproducibility of measurements of circulating RBP4 and omentin-1. This aim was achieved by analyzing measurements of these two adipokines in two blood samples per individual taken several months apart in a group of 207 middle-aged white participants.

Methods

Results

From the initial sample of 207 participants one individual had missing values for plasma omentin-1 concentrations and was therefore excluded from all subsequent analyses leaving an analytical sample of 206 participants. Table 1 summarizes basic characteristics of the study population including adipokine measurements. Briefly, the average age of participants was 55 years (SD 4.2), 59% were women, 60% were overweight or obese, and 92% had not eaten for at least 8 hours at the time of blood draw. Men were older and had a higher BMI than women. The average concentrations of RBP4 were 50.8 μg/mL (95% CI 48.8, 52.9) in samples from the first blood draw and 50.0 μg/mL (95% CI 48.0, 52.0) in samples from the second blood draw (p for difference 0.21); the average concentrations of omentin-1 were 401.0 ng/mL (95% CI 383.6, 418.4) in samples from the first blood draw and 396.8 ng/mL (95% CI 379.4, 414.2) in samples from the second blood draw (p for difference 0.41), respectively. At both time points the average RBP4 concentrations were higher while the average omentin-1 concentrations were lower among men compared to women (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of the study population.

https://doi.org/10.1371/journal.pone.0138480.t001

The agreement between repeated adipokine measurements within individuals in relation to individual means was visualized in Bland-Altman-Plots (Fig 1). The Bland-Altman-Plots do not show clustering of differences at certain ranges of the adipokine concentrations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Bland-Altman-Plot.

Agreement of repeated biomarker measurements in individuals in relation to the overall variability of measurements; agreement was calculated as difference in plasma-concentrations between the two blood-sampling occasions (t2-t1) within individuals; range of agreement was defined as mean ± 1.96 SD and is marked by the dashed lines. A: retinol-binding protein 4 (RBP4); B: omentin-1.

https://doi.org/10.1371/journal.pone.0138480.g001

Table 2 shows reliability parameters of the repeated adipokine measurements. The ICCs for repeated RBP4 and omentin-1 measurements were 0.77 (95% CI 0.71, 0.82) and 0.83 (95% CI 0.78, 0.87), respectively. The CVs for intraindividual measurement variation between repeated measurements were 14% for RBP4 and 13% for omentin-1. These results were robust across strata according to sex, age, BMI, blood pressure (diastolic and systolic, respectively), and time-interval between measurement time points (Table 2); i.e. the ICCs within strata were all within the 95% CI of the full cohort ICC. Neither excluding the 5% most extreme observations (2.5% highest and 2.5% lowest values) of all RBP4 and omentin-1 measurements, respectively, nor excluding non-fasted participants from the analyses substantially affected the results.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Reliability parameters of repeated measurements of RBP4 and omentin-1.

https://doi.org/10.1371/journal.pone.0138480.t002

We also calculated ICC-corrected (“true”) RRs from hypothetically observed RRs of 1.5, 2.5, and 3.5, respectively. Correcting these observed RRs for the RBP4-related ICC of 0.77 resulted in RRs of 1.69, 3.29, and 5.09, respectively, corresponding to a change in estimate of 12.7%, 31.6%, and 45.4%. Correcting the hypothetically observed RRs for the omentin-1-related ICC of 0.80 resulted in “true” RRs of 1.66, 3.14, and 4.79, respectively, corresponding to a change in estimate of 10.7%, 25.6%, and 36.4%.

Discussion

In our study in a middle aged white population, measurements of circulating concentrations of RBP4 and omentin-1 were highly reproducible in individuals over a period of several months, indicated by low CV and excellent ICC for both signaling factors. These results were robust across strata according to age, sex, BMI, blood pressure, and time-interval between repeated blood-sampling. Thus, the study is one of the first that provided detailed data on the reliability of measurements of RBP4 and omentin-1 applicable in observational study settings with a single blood draw.

In our study, the CVs for repeated measurements of RBP4 and omentin-1 several months apart were consistently below 20% and ICCs were over 0.75 corresponding to an excellent reliability [20]. Our results indicate that the biological variance of RBP4 and omentin-1 within individuals is sufficiently low compared to the difference of the circulating concentrations between individuals.

Associations of RBP4 and omentin-1 with age, sex, BMI, and systolic as well as diastolic blood pressure were reported [6, 22–25]. An important question is if the reliability might depend on these factors. The stability of reliability-related parameters (ICC and CV) across strata according to these factors in our study did not indicate such dependencies. Furthermore, our sensitivity analysis over strata according to time-interval between the two blood-sampling occasions did not indicate that reliability substantially depended on this timespan. Still, the majority of blood samples in our study were taken not more than 5 five months apart. Therefore, our results do not provide information with regard to larger timespans.

Assumed that false positive case classifications are rare, the biomarker-associated risk estimates from observational studies will be biased towards the null by imprecision in the measurements [26]. The exemplary correction of hypothetically observed risk estimates by the ICCs assessed in the present study demonstrated two insights. First, qualitative inference about the risks related to observed levels of circulating RBP4 and omentin-1 seems not to be severely biased by biological variation of these markers within individuals. Second, even for biomarkers with excellent reliability the underestimation of the “truly” associated risk induced by the random measurement error related to single time point biomarker measurements seems relevant in some regards. For a risk estimate of 1.5 correcting for the omentin-1-related ICC of 0.80 would induce a change in estimate over 10%, while correcting a risk estimate of 3.5 by the RBP4-related ICC of 0.77 would lead to a 45% higher estimated risk. In the future, in epidemiological studies, reporting ICC-corrected risk estimates in sensitivity analyses might be a convenient possibility to quantify the bias induced by imprecision of the measurement due to variation of RBP4 and omentin-1 within individuals.

Strengths of our study include the good characterization of participants that allowed us testing robustness of our results according to several strata and performing sensitivity analyses. Furthermore, regarding age and BMI distributions our population is likely to resemble conditions among prospective cohorts in which the relations of RBP4 and omentin-1 with the incidence of metabolic and cardiovascular diseases are likely to be investigated in future. As our population was predominantly from European descent, caution needs to be taken in generalizing our results to other ethnicities. Our study population represented a normal population and it is unclear in how far our results can be generalized to cohorts with diseased subjects. Of note, we most likely did not include participants with severe renal diseases and markers of kidney function were not available in this validation study. Glomerular filtration rate, however, is a main determinant of RBP4 levels, with impaired kidney function being related to unphysiologically high circulating concentrations [13, 27]. Importantly, we did not evaluate stability of RBP4 levels in patients with impaired renal function and our results might not be generalizable to study populations enriched with such patients. Generalization of our study results for use in error correction models might also be hampered by sampling and measurement procedures. We used EDTA plasma and the ELISAs were validated for the use of EDTA plasma. Care should be taken if other anticoagulants or other assays are used. Regarding RBP4, the quantification in ELISAs was generally criticized and new mass spectrometry-based approaches were proposed [28]. Such approaches are of interest as they provide additional information on the posttranslational modifications of RBP4. Still, ELISAs provide the most cost effective and widespread method to quantify RBP4 in the blood in large cohorts and our results encourage their use in such studies.

In conclusion, our results indicate that single measurements of RBP4 and omentin-1 provide highly reliable estimates for the concentration of these adipokines in individuals. Thus, it seems reasonable to rely for risk estimates on concentrations of RBP4 and omentin-1 from single baseline samples in observational studies.

Acknowledgments

We thank all study participants. Furthermore, we thank Ellen Kohlsdorf, Dr. Manuela Bergmann, Wolfgang Fleischhauer, Herbert Piechot and Andrea Teichmann for excellent technical assistance. The study was supported in part by a grant from the German Federal Ministry of Education and Research (BMBF) to the German Center for Diabetes Research (DZD, http://www.dzd-ev.de). The authors declare that they have no conflict of interest.