key: cord-1029997-khm23xvn
authors: Sithiravel, Cindhya; Røysland, Ragnhild; Alaour, Bashir; Sylte, Marit Sverresdotter; Torsvik, Janniche; Strand, Heidi; Marber, Michael; Omland, Torbjørn; Aakre, Kristin Moberg
title: Biological variation, reference change values and index of individuality of GDF-15
date: 2022-03-28
journal: Clinical chemistry and laboratory medicine
DOI: 10.1515/cclm-2021-0769
sha: 5a26b20fc40f5ec819a3f5fcd8e2f8d1b0e91dbf
doc_id: 1029997
cord_uid: khm23xvn

nan

To the Editor, GDF-15 is an upcoming and promising prognostic inflammatory biomarker found to be elevated in a wide range of diseases, including cancer and cardiovascular diseases [1] [2] [3] . Knowledge about the biological variation of GDF-15 is useful for suggesting analytical performance specifications. Furthermore, it aids in determining the number and frequency of samplings needed, as components with large biological variation may require several measurements to establish an individuals`homeostatic set point, which is necessary to confirm a diagnosis or assess prognosis. Reference change values (RCV) help determine whether the changes in results on serial testing within the same individual is clinically significant or can be explained by analytical or physiological variation. The index of individuality (II) allows us to distinguish between analyses where the population-based reference intervals or absolute cut-offs are relevant (index >1.4) and analyses where the subject-based reference intervals (i.e., delta changes) are more appropriate (index <0.6) [4, 5] . Previous studies have shown conflicting results concerning the biological variation, RCV and II of GDF-15 [6, 7] .

We carried out a multicenter study at Haukeland University Hospital, Akershus University Hospital in Norway, King's College London and Guy's and St Thomas' Hospital, United Kingdom. The study design was planned according to the EFLM checklist for biological variation studies (BIVAC) [8] , and is described earlier [9] . Thirty healthy volunteers were recruited, 16 were women and two were daily smokers. Age range was 21-64 years (median 36 years). Participants had no evident disease, and no previous history of chronic illness. Screening tests including glucose, eGFR, high-sensitivity troponin and NT-proBNP were performed at baseline. Healthy-status was defined biochemically as: -Non fasting glucose <7.8 mmol/L -eGFR (CKD-EPIcreat) >60 mL/min/1.73 m 2 -cTnT <99th percentile for the assay (<14 ng/L) -NT-proBNP <the local reference limit Venous blood sampling was performed weekly, on the same weekday ±1 day, for 10 consecutive weeks, and separated serum was stored at −80°C until analysis. GDF-15 was analyzed in duplicates, with the same reagent and calibrator lot at Akershus University Hospital on the Cobas e801 (Roche Diagnostics, Switzerland). The limit of detection (LOD) and limit of quantification (LOQ) was 400 ng/L (CV A of ≤20%), and the measuring range was 400-20,000 ng/L. The CV A was 0.47% at concentration 1,372-1,386 ng/L and 0.61% at concentration 7,373-7,458 ng/L.

In total 17/30 (57%) individuals had ≥4 duplicate measurements above the LOD, and were eligible for further analysis. Analytical outliers were removed according to Burnett test, and outliers in mean values according to Reeds criteria [9] . The serial values for each individual were checked for linear trend using liner regression; subjects with significant trend (p-value of 0.01) were removed. Cochrane's and Bartlett's test were used to test homogeneity of analytical and within-subject variances, outliers were removed until homogeneity was achieved. Shapiro-Wilk test was used to test normality of the residuals. Thirteen subjects were included in the final analysis. Median concentrations were calculated for the total cohort, genderand age-specific subgroups, differences in individual medians between age and gender groups were tested using Mann Whitney U test. Biological variation, RCV and II were calculated in the total cohort and age-stratified sub-groups (below 45 years and ≥45 years), as GDF-15 is higher in older individuals [3] . Age groups were chosen based on visual inspection of the data which fits well with the age-group stratification Krintus et al. have chosen ( Figure 1 ) [6] .

The distribution of the data was skewed (nonparametric) and calculations of σ A , σ I , σ G were done on ln transformed data, using nested balanced ANOVA. The σ was thereafter retransformed into analytical variation (CV A ), within-subject variation (CV I ), and between-subject variation (CV G ) using:

in which σ is the estimated standard deviation for the lntransformed data and CV ln is the adjoining re-transformed CV.

The RCV values (with 95% confidence intervals) were calculated according to Fokkema et al. [10] :

in which σ A is the analytic standard deviation and σ I is the within-person standard deviation of the logarithmic data. The II was calculated using the retransformed data as follows: Desirable analytical performance goals was calculated as [5] :

Median GDF-15 concentration for all participants (13 individuals) was 515 (25 and 75 percentile, 463-652) ng/L. No difference in GDF-15 values between men (n=6) and women (n=7) was observed (median, 468 vs. 545 ng/L, for men vs. women, respectively; p-value for difference 0.36). Median concentration was higher in participants ≥45 years (n=8) compared to those <45 years (n=5): 719 vs. 470 ng/L, respectively; p-value for difference 0.03. The smokers were both excluded due to non-measurable (<LOD) results.

Calculated biological variation, RCV and II are presented in Table 1 together with number and reasons for participant exclusion. Results were similar across age groups except that CV G was lower in younger subjects, and consequently the II was higher.

Our study population and analysis method is similar to Krintus et al. We obtain similar estimates for biological variation, RCV, II and CV A [6] , confirming that results are robust. The observation that CV G and II as a trend might be age-dependent is in line with the current interpretation of GDF-15 as a marker of biological age. However, due to overlapping CI of CV G our results cannot be used to draw a firm conclusion of a significant larger inter-individual biological variation in older individuals [3] . It is noteworthy that 1/3 of our study population had GDF-15 levels below the LOD. This was also reported by Krintus, were 10/ 26 participants showed GDF-15 measurements below the LOD [6] . Meijers et al. [7] analyzed GDF-15 by a quantitative sandwich enzyme immunoassay technique (Quantikine ® ; R&D Systems, Inc., Minneapolis, MN, USA). For healthy subjects, they reported a higher CV I (18.9%) and CV A (15.2%), compared to our findings.

Based on our data we suggest the following analytical performance specifications; 3.8% as desirable CV A , and 6.2% as desirable analytical bias. In agreement with our and Krintus' [6] results and the precision data reported by Roche, routine laboratories should be able to fulfill these performance specifications, utilizing the Roche Elecsys immunoassay.

The strength of our study is the standardized protocol and firm data analysis. The limitation is the relative low number of included subjects since many participants were excluded due to measured results below LOD, hence the variability in subjects with GDF-15 below LOD is unknown. The age groups were chosen arbitrary, and small group size resulted in larger uncertainty for the age-specific estimates, and these need confirmation in other studies. Also, our study only included healthy subjects, so the biological variation data are not applicable for interpreting clinical relevant changes in patients with chronic disease. However, Meijers et al. measured CV I and CV G in healthy subjects and heart failure patients, and did not find discrepancy between the two groups. Applicability in diseased populations still needs confirmation in future studies. Our results confirm earlier findings of biological variation, RCV and II of GDF-15. The low biological variation and RCV indicate a biomarker useful for prognostication in individual subjects. The low II signals that subject-based reference intervals could be more appropriate than population based reference intervals or using an absolute cut offs (i.e., 1,800 ng/L) as suggested for prognostication today [3] . Finally, the biological variation is sufficiently large compared to the analytical variation, indicating that desirable analytical performance of GDF-15 measurements should be achievable in routine laboratories. Omland is also listed as an inventor on a patent application entitled: "GDF-15 for predicting the disease severity of a patient with COVID-19" and on the granted patent applications Granin proteins as markers of heart disease. PCT/GB0818650.4 SgII as a prognostic marker in conditions which require critical care. CT/GB0919901.9. Cindhya Sithiravel, Ragnhild Røysland, Bashir Alaour, Marit S. Sylte, Janniche Torsvik, Heidi Strand, Michael Marber report no conflict of interest. Informed consent: Informed consent was obtained from all individuals included in this study. Ethical approval: This study was carried out according to the principles of the Declaration of Helsinki. The protocol was approved by the respective regional ethics committee at each centre: South Central -Berkshire Research Ethics Committee (London), and the Regional Committee for Medical and Health Research Ethics in Bergen (Bergen and Oslo).

Growth differentiation factor 15 (GDF15): a survival protein with therapeutic potential in metabolic diseases

Review article GDF-15 as a target and biomarker for diabetes and cardiovascular diseases: a translational prospective

Growth differentiation factor 15 as a biomarker in cardiovascular disease

Generation and application of data on biological variation in clinical chemistry

Tietz textbook of clinical chemistry and molecular diagnostics

A study of biological and lifestyle factors, including withinsubject variation, affecting concentrations of growth differentiation factor 15 in serum

Variability of biomarkers in patients with chronic heart failure and healthy controls

The biological variation data critical appraisal checklist: a standard for evaluating studies on biological variation

Biological variation of cardiac myosin-binding protein C in healthy individuals

Reference change values for brain natriuretic peptides revisited