Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Cytokine. Author manuscript; available in PMC 2012 November 1.
Published in final edited form as:
PMCID: PMC3185107

Intra-individual variability over time in serum cytokine levels among participants in the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial



Serum measurements of cytokines, mediators of various B cell and T cell activities, are important markers of inflammation and immune dysregulation. We assessed the reproducibility of serum cytokine measurements over a five-year period among participants in the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO).


Levels of 13 cytokines [interleukin (IL) 1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12p70, IL-13, interferon-gamma (IFNγ), granulocyte macrophage colony-stimulating factor (GM-CSF), and tumor necrosis factor-α (TNFα)] in stored sera from three collections (study baseline, +1 yr, and +5 yr) among 28 randomly selected PLCO participants were measured using a high-sensitivity Luminex xMap-based multiplex panel. Within- and between-subject components of variance were estimated from random effects models and were used to calculate the coefficient of variation (CV) and intraclass correlation coefficient (ICC) for analytes with <30% of samples below the limit of detection (LOD). Spearman correlation coefficients between measurements of the same analyte over time and between analytes were also calculated.


Among the six cytokines with <30% of samples below the LOD, we observed excellent reproducibility for IL-6, IL-7, IL-13, and TNFα (ICC ≥ 0.73), and fair to good reproducibility for IL-8 (ICC = 0.55) and IL-10 (ICC = 0.60). Spearman correlation coefficients comparing paired measurements of each cytokine at baseline and at +5 yr were high (ρ ≥ 0.74) with the exception of IL-10 (ρ = 0.44).


These results suggest that measurements of most of the cytokines evaluated in this study were highly reproducible over a five-year period.

Keywords: cytokines, inflammation, variability, serum, cancer

1. Introduction

Cytokines are secreted polypeptides or glycoproteins that mediate various B and T lymphocyte responses such as proliferation, antibody production, and interaction with immunoglobulins. Circulating cytokine concentrations are considered to be important biological markers in the mechanistic pathway between inflammation and immune dysregulation and development of cancer or other chronic diseases [1,2]. Although many molecular epidemiologic studies have used serum or plasma cytokine measurements from a single blood collection, few studies have assessed within-subject variability of these measurements over relatively long (≥2 yr) time periods [35]. Understanding the degree to which cytokine measurements from a single collection are representative of levels over a long time period for a given subject is critical to assessing the utility of these measurements for etiologic research.

In this study, we evaluated temporal variability of selected cytokines measurements over a five-year period using serial serum samples obtained from participants in the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO).

2. Materials and Methods

2.1 Study design

The study design and sample collection procedures in PLCO have been described [6,7]. Briefly, participants in the screening arm of PLCO provided nonfasting blood samples at six annual medical examinations. Samples were processed and frozen within 2 hours of collection and stored at −70ºC. A total of 28 cancer-free PLCO screening arm participants were included in this study; this subsample included participants from each of the 10 PLCO screening centers. These subjects were selected at random from among PLCO participants meeting the inclusion criteria, as described previously [8]. We analyzed stored serum samples that were collected from each subject at study baseline, the first annual follow-up visit (+1 y), and the fifth annual follow-up visit (+5 y). Of the 28 subjects, 27 had serum from baseline, 28 had serum from +1 y, and 25 had serum from +5 y. All subjects had serum from at least two collection times, and 24 subjects had serum from all three collections. Stored specimens from these study subjects have been used previously to evaluate temporal variability of other analytes (e.g., serum 25-hydroxyvitamin D concentration; ref. [9]). The serum used in the present study went through two freeze-thaw cycles (thaw of the parent vial and thaw of the aliquot at the lab).

2.2 Serum cytokine measurements

The following cytokines were measured using a high-sensitivity multiplex cytokine panel: interleukin (IL) 1β, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12p70, IL-13, interferon-gamma (IFNγ), granulocyte macrophage colony-stimulating factor (GM-CSF), and tumor necrosis factor-α (TNFα). Assays were performed using the MILLIPLEX xMAP 13-plex Cytokine Kit (Millipore, Billerica, MA) on the BioRad BioPlex200 (Luminex) instrument at the Protein Expression Laboratory at SAIC-Frederick. The lower limits of detection (LOD) for specific analytes ranged from 0.01–0.48 pg/mL based on manufacturer specifications. Samples were analyzed in three batches; all of the samples for a given participant were included in the same batch, and three blinded replicate samples from a common QC pool were also included in each batch. Measurements for each sample were performed in duplicate and the average of the two measurements was used.

2.3 Statistical analysis

We restricted our analyses to the six cytokines with <30% of samples below the LOD (IL-6, IL-7, IL-8, IL-10, IL-13, and TNFα). The percentage of measurements below the LOD for the other seven cytokines were as follows: IL- 1β, 79%; IL-2, 78%; IL-4, 50%; IL-5, 50%; IL-12p70, 71%; IFNγ, 73%; and GM-CSF, 36%. For the six selected cytokines, data were natural log-transformed prior to analysis to achieve more normal distributions. Data for some cytokines were left-truncated due to measurements falling below the LOD. We used the multiple imputation method described by Lubin et al. [10] to account for measurements below the LOD. To evaluate assay precision, we estimated the within-batch coefficient of variation (CV) for each analyte using the blinded QC replicate data. For the main reproducibility analyses, we used mixed effects models with batch and subject included as random effects (subject was nested within batch). The within- and between-subject variance components were used to calculate the within-subject CVs and the intraclass correlation coefficients (ICC). The following formula was used to calculate the ICC: σ2B / (σ2W + σ2B) [between-subject variance / (within-subject variance + between-subject variance)]. As suggested by Rosner [11], an ICC ≥0.75 indicates excellent reproducibility, 0.4 ≥ ρ < 0.75 indicates fair to good reproducibility, and <0.4 indicates poor reproducibility. Analyses were repeated after stratifying by batch. Because sex hormones may influence serum levels of several cytokines [12], analyses stratified by sex were also performed. Sensitivity analyses were performed after excluding subjects with high outlying values (defined as >3 standard deviations above the mean value) and measurements below the LOD. Spearman rank correlation coefficients were also calculated to evaluate agreement between measurements of the same analyte at different collection times and to assess the relations between different analytes. Statistical analyses were performed using Stata 10.1 (StataCorp LP, College Station, TX) and R statistical software (Free Software Foundation, Boston, MA).

3. Results

The mean age at baseline among subjects included in this study was 61 years (range, 55–70 years). Most participants were non-Hispanic Caucasian (N=22), and over half were male (N=17). Among all the analytes assayed, the proportion of measurements that were below the LOD ranged from 0% to 79%.

The median, 5th–95th percentile range, and percentage of measurements below the LOD at each collection time for the six analytes evaluated in this study are reported in Table 1. No statistically significant differences in cytokine levels by collection time were observed (P ≥ 0.09, Kruskal-Wallis test). Measures of variability in serum cytokine levels are reported in Table 2. The CVs estimating assay precision for the pooled QC replicate samples were fairly high for most analytes (4.6% to 29.0%); TNFα was the only analyte with a CV < 10%. However, after excluding a single pooled QC sample for which discrepancies between duplicate well measurements were observed for several analytes (duplicate well CVs ≥77%), the corresponding assay CVs were much lower for most analytes: IL-6, CV=7.1%; IL-7, CV=17.2%; IL-8, CV=6.0%; IL-10, CV=4.7%; IL-13, CV=4.3%; and TNFα, CV=5.1%. The within-subject CVs, which account not only for assay precision but also temporal variation within subjects over the five-year study period, were considerably higher, ranging from 36% to 137%. Exclusion of subjects with high outlying values resulted in only modest improvements in the within-subject CVs for most analytes (data not shown).

Table 1
Descriptive statistics for serum cytokine levels (pg/mL) at study baseline, +1yr, and +5yr*
Table 2
Measures of agreement for serum cytokine levels in serial samples collected over a 5-year period

Based on the estimated ICCs, we observed high reproducibility for IL-6, IL-7, IL-13, and TNFα (ρ ≥0.73), and fair to good reproducibility for IL-8 (ρ = 0.55) and IL-10 (ρ = 0.60). Analyses were repeated after restricting to paired samples from baseline/+5yr. The resulting ICCs were similar for IL-6, IL-7, IL-13, and TNFα. The ICC for IL-8 was higher in this analysis (ρ = 0.81), and the ICC for IL-10 was lower (ρ = 0.46). For each cytokine, sensitivity analyses were performed after excluding one subject with high outlying values (not the same subject for each analyte). After these exclusions, the ICC for IL-8 was higher (ρ = 0.84) but otherwise results were essentially unchanged. For the three analytes with measurements below the LOD (IL-6, IL-10, IL-13), the ICCs were generally similar after restricting to observations with detectable levels, except that for IL-10 which was somewhat lower (ρ = 0.48). After stratifying by sex, we observed similar ICCs among men and women for most analytes, with the exception of IL-7, for which the ICC was considerably higher among men than among women (ρ = 0.88 and 0.27, respectively). After stratifying by batch, for IL-8 we observed a lower ICC in the first batch relative to the second and third batches (ICCs of 0.32, 0.66, and 0.90, respectively); otherwise, no consistent patterns of differences in the ICCs across batches were observed (Table 3).

Table 3
ICCs (95% CIs) for repeated measures of serum cytokine levels in samples collected over a 5-year period, stratified by batch

Spearman correlation coefficients comparing paired measurements of each cytokine at baseline vs. +5yr were also generally high (ρ ≥0.74) for most analytes, with the exception of the modest correlation for IL-10 (ρ = 0.44). For comparisons between analytes (based on measurements from baseline samples), correlations were highest between IL-13 and most other analytes (ρ = 0.31–0.85; Supplementary Table 1). A fairly strong positive correlation (ρ = 0.61) between IL-7 and IL-8 was also observed.

4. Discussion

There is considerable interest in measuring circulating levels of cytokines in molecular epidemiologic studies to better understand the mechanistic pathways by which infections and inflammation may influence the development of cancer and other chronic disease. The results from this study suggest that, for several cytokines, measurement in a single serum sample provides a good surrogate for levels over a five-year period and should be useful for ranking an individual’s levels relative to other subjects. The high ICCs for most analytes demonstrated that there was considerably more variation between subjects than within subjects, which supports the use of these markers in nested case-control studies with serum samples from matched sets of cases and controls analyzed in the same batch.

Based on the high ICCs observed in this study, we would expect a relatively modest attenuation of risk estimates for a given disease in relation to most cytokines. We used the following formula derived from Rosner et al. [13] to estimate the expected difference between true and observed risk estimates in a nested case-control study with matched sets analyzed in the same batch: RRobs = exp [ICC × ln (RRtrue)]. For this exercise we assumed a true relative risk of 2.0 comparing subjects with high and low levels of a given cytokine, which is consistent with risk estimates observed in previous studies of non-Hodgkin lymphoma risk in relation to circulating cytokine levels in pre-diagnostic specimens [14,15]. Assuming a true relative risk of 2.0, we would expect observed relative risks between 1.66 and 1.82 for IL-6, IL-7, IL-13, and TNFα (ICCs ranging from 0.73 to 0.86). However, the attenuation might be somewhat greater for IL-8 (ICC = 0.55, RRobs = 1.46) and IL-10 (ICC = 0.60, RRobs = 1.52), which could potentially obscure true associations for these markers.

The ICCs observed in this study are consistent with those reported in a previous study of reproducibility of multiplex-based cytokine measurements in serum samples obtained annually over 2–3 years from pre- and post-menopausal women [5]. However, somewhat lower ICCs were observed for IL-6 and TNFα in previous studies among young women using ELISA measurements in samples collected over a two-year period (IL-6, ICC = 0.48; TNFα, ICC = 0.73; ref. [3]) and multiplex measurements in samples collected over a three-year period (IL-6, ICC = 0.47; TNFα, ICC = 0.39; ref. [4]). Differences in the age and sex distributions and assay methods may explain the discrepancy between our findings and the findings of these studies. Several other studies have evaluated reproducibility of cytokine measurements [1618], but samples were collected over relatively short time intervals (six months to one year), thus limiting the generalizability of their findings regarding temporal variation.

The use of serial samples collected from PLCO participants is an important strength of this study. Through the use of PLCO specimens collected five years apart, we were able to assess temporal variability across a longer time period than in previous studies. In addition, the standardized blood collection and processing protocol used in PLCO, which specified that blood specimens be centrifuged, processed and frozen with two hours of blood collection, minimizes blood collection and processing procedures as a source of variation in serum analyte levels. This is particularly important for the measurement of cytokines, given concerns regarding the potential of ex vivo expression from leukocytes and enzymatic digestion of cytokines with prolonged time to processing [19,20].

This study also had limitations. The sample size was relatively small, although the 95% CIs for the estimated ICCs still support the inference of high reproducibility over the five-year period. Because samples were obtained a maximum of five years apart, we are unable to generalize these findings to agreement beyond a five year window, which may not reflect the most etiologically relevant period for cancer or other diseases with long latency intervals. Finally, we were unable to assess the reproducibility of some analytes for which a high proportion (≥30%) of the measurements were below the LOD.

5. Conclusions

The high ICCs observed for most, but not all, of the cytokines evaluated in this study suggest that serum cytokine measurements (particularly IL-6, IL-7, IL-13, and TNFα) from a single sample are likely to be useful for characterizing immune and inflammation response over a five-year period in molecular epidemiologic studies.


  • We evaluated variability in serum cytokine levels in samples collected up to 5y apart
  • Levels of IL6, IL7, IL8, IL10, IL13, and TNFα were consistent over time (ICC≥0.55)
  • A single measurement of these cytokines is a good surrogate for levels over a 5y period

Supplementary Material



This research was supported by the Intramural Research Program of the Division of Cancer Epidemiology and Genetics and by contracts from the Division of Cancer Prevention, National Cancer Institute, NIH, DHHS. The authors thank Drs. Christine Berg and Philip Prorok, Division of Cancer Prevention, National Cancer Institute, the Screening Center investigators and staff of the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial, Mr. Tom Riley and staff, Information Management Services, Inc., Ms. Barbara O’Brien and staff, Westat, Inc., Mr. Tim Sheehy and staff, DNA Extraction and Staging Laboratory, SAIC-Frederick, Inc, and Ms. Jackie King and staff, BioReliance, Inc. Most importantly, we acknowledge the study participants for their contributions to making this study possible.


Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.


1. Shacter E, Weitzman SA. Chronic inflammation and cancer. Oncology (Williston Park) 2002;16:217–26. [PubMed]
2. Lin WW, Karin M. A cytokine-mediated link between innate immunity, inflammation, and cancer. J Clin Invest. 2007;117:1175–83. [PMC free article] [PubMed]
3. Ho GY, Xue XN, Burk RD, et al. Variability of serum levels of tumor necrosis factor-alpha, interleukin 6, and soluble interleukin 6 receptor over 2 years in young women. Cytokine. 2005;30:1–6. [PubMed]
4. Kaplan RC, Ho GY, Xue X, et al. Within-individual stability of obesity-related biomarkers among women. Cancer Epidemiol Biomarkers Prev. 2007;16:1291–3. [PubMed]
5. Gu Y, Zeleniuch-Jacquotte A, Linkov F, et al. Reproducibility of serum cytokines and growth factors. Cytokine. 2009;45:44–9. [PMC free article] [PubMed]
6. Gohagan J, Prorok P, Levin D, et al. The Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial. Control Clin Trials. 2000;21:249S–406S.
7. Hayes RB, Reding D, Kopp W, et al. Etiologic and early marker studies in the prostate, lung, colorectal and ovarian (PLCO) cancer screening trial. Control Clin Trials. 2000;21:349S–55S. [PubMed]
8. Gunter MJ, Cross AJ, Huang WY, et al. A prospective evaluation of C-reactive protein levels and colorectal adenoma development. Cancer Epidemiol Biomarkers Prev. doi: 10.1158/1055-9965.EPI-10-1099. Published online 06 Jan 2011. [PMC free article] [PubMed] [Cross Ref]
9. Hofmann JN, Yu K, Horst RL, et al. Long-term variation in serum 25-hydroxyvitamin D concentration among participants in the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial. Cancer Epidemiol Biomarkers Prev. 2010;19:927–31. [PMC free article] [PubMed]
10. Lubin JH, Colt JS, Camann D, et al. Epidemiologic evaluation of measurement data in the presence of detection limits. Environ Health Perspect. 2004;112:1691–6. [PMC free article] [PubMed]
11. Rosner B. Fundamentals of Biostatistics. 5. Pacific Grove, CA: Duxbury; 2000.
12. Sites CK, Toth MJ, Cushman M, et al. Menopause-related differences in inflammation markers and their relationship to body fat distribution and insulin-stimulated glucose disposal. Fertil Steril. 2002;77:128–35. [PubMed]
13. Rosner B, Spiegelman D, Willett WC. Correction of logistic regression relative risk estimates and confidence intervals for random within-person measurement error. Am J Epidemiol. 1992;136:1400–13. [PubMed]
14. Saberi Hosnijeh F, Krop EJ, Scoccianti C, et al. Plasma cytokines and future risk of non-Hodgkin lymphoma (NHL): a case-control study nested in the Italian European Prospective Investigation into Cancer and Nutrition. Cancer Epidemiol Biomarkers Prev. 2010;19:1577–84. [PubMed]
15. Gu Y, Shore RE, Arslan AA, et al. Circulating cytokines and risk of B-cell non-Hodgkin lymphoma: a prospective study. Cancer Causes Control. 2010;21:1323–33. [PMC free article] [PubMed]
16. Cava F, Gonzalez C, Pascual MJ, et al. Biological variation of interleukin 6 (IL-6) and soluble interleukin 2 receptor (sIL2R) in serum of healthy individuals. Cytokine. 2000;12:1423–5. [PubMed]
17. Gonzalez C, Cava F, Ayllon A, et al. Biological variation of interleukin-1beta, interleukin-8 and tumor necrosis factor-alpha in serum of healthy individuals. Clin Chem Lab Med. 2001;39:836–41. [PubMed]
18. Lee SA, Kallianpur A, Xiang YB, et al. Intra-individual variation of plasma adipokine levels and utility of single measurement of these biomarkers in population-based studies. Cancer Epidemiol Biomarkers Prev. 2007;16:2464–70. [PubMed]
19. Thavasu PW, Longhurst S, Joel SP, et al. Measuring cytokine levels in blood. Importance of anticoagulants, processing, and storage conditions. J Immunol Methods. 1992;153:115–24. [PubMed]
20. House RV. Cytokine measurement techniques for assessing hypersensitivity. Toxicology. 2001;158:51–8. [PubMed]