Within- and between-subject biological variation of hemostasis parameters in a study of 26 healthy individuals

Introduction

Various coagulation tests, such as prothrombin time (PT) and activated partial thromboplastin time (aPTT), are required in numerous clinical scenarios. Routine coagulation tests are frequently utilized in the preliminary assessment of bleeding: PT, aPTT, and fibrinogen assays. Coagulation tests are often used to monitor patients taking anticoagulant medications, such as warfarin, to ensure that the drug prevents blood clots from forming. They are also used to diagnose and monitor bleeding disorders, such as hemophilia, and evaluate the effectiveness of treatment for these disorders [1].

Biological variation (BV) refers to the variation observed in the concentration or activity of various constituents in a person, which reflects the regulation by homeostatic processes within the body [2]. BV has two main components: within-subject BV (also known as CV_I) and between-subject BV (also known as CV_G). Within-subject BV refers to the random fluctuation around a homeostatic set point of an analyte within a single individual. Between-subject BV refers to the variation in homeostatic set points of an analyte among different individuals [3].

In laboratory medicine, BV data can be used for various purposes. Setting analytical performance specifications (APS) for imprecision, bias, total error, and measurement uncertainty is a common application for BV data. The use of empiric analytical performance standards and the absence of other BV-related implications are caused by the fact that there is still a lack of data on the BV of coagulation tests [4]. To precisely assess longitudinal changes in individuals, it is necessary to account for the biological variability of the measurement [5]. However, there is currently a scarcity of data on biological variation in hemostasis factors [6]. It is crucial for diagnostic tests to yield results that aid clinicians in making diagnoses and selecting the best treatment plan for their patients. The best way to define analytical performance criteria is still under development. Three different models are proposed to define the parameters of the analytical performance. One of the models, Model 1, prioritizes analytical performance for clinical outcomes. However, it is also the most challenging to accomplish, and only a small number of analytes strive for it. Model 2 focuses on the components of biological variation of the measurand, while Model 3 relies on the state-of-the-art methods available [7]. To determine the appropriate performance specification for a measurement, it is necessary to consider the requirements and constraints of the measurement system and its intended use. Therefore, accurate measurements of coagulation tests are required for precise hemostasis evaluations and patient therapy. Reliable BV data are necessary to assure the clinical safety of hemostasis testing [8]. Also, reference change values (RCV) can be computed using estimated CV_I and analytical imprecision to determine the extent to which a difference between two consecutive results in an individual can be attributed to such sources of variation as analytical and within-subject biological variation.

In this study, it was aimed to calculate the biological variation components of a total of 26 healthy individuals and to compare them with the biological variation database. In addition, our study will calculate analytical performance specifications based on BV data and the number of samples required to estimate the homeostatic set points for hemostasis parameters (PT, aPTT, and fibrinogen).

Materials and methods

Results

Our study included 26 healthy participants (14 females and 12 males) over a 10-week period, with samples collected weekly. The median (minimum-maximum) age of all subjects as well as males and females, were 30 (23–45), 33 (25–45), and 27 (23–45) years. There was no significant difference between the ages of males and females (p=0.076). Each parameter’s obtained data was examined for outliers using the Dixon test. The datasets for PT, aPTT, and Fibrinogen were found to contain outliers. Upon closer examination, it was discovered that these outliers were attributable to errors in the sample collection process. As a result, these outliers were excluded from subsequent analyses.

Table 1 presents data for all parameters, including CV_I, CV_G, and analytical CV (CV_A).

Table 2 shows an overview of a calculation of reference change values. The RCV is used to determine if there is a statistically significant change (increase or decrease) between two subsequent test results of an individual and the Individuality index (II). According to Roraas et al. [11], the asymmetrical RCV should be calculated when the data does not follow a normal, symmetrical distribution. In other words, if the data are asymmetrical, using asymmetrical RCV is indicated to provide a more accurate representation of variability.

All parameters had CV_A values below 2 %. The CV_I for PT was calculated to be 3.2 %, whereas the CV_G was 5.8 %. II of 0.73. The CV_I and CV_G for aPTT were 5.8 and 8.4 %, respectively, and II was 0.91. Fibrinogen exhibited greater BV than the other two tests. It was determined that the CV_I was 11 %, the CVG was 17.4 %, and II was 1.31. Our asymmetrically calculated RCVs for PT, aPTT, and fibrinogen were −7.5 to +8.1, −12.7 to 14.6, and −22.7 to 29.4, respectively (Figure 1A–C).

Figure 1:

Biological variation of prothrombin time (PT), activated partial thromboplastin time (aPTT), and fibrinogen tests. The triangular and circular figures show measurement points individually and median values with the range (minimum-maximum) of the tests. The dashed lines show the 5th and 95th percentile. The continuous thin lines indicate 95 % CI, and the continuous thick lines indicate the median.

Furthermore, Table 2 outlines the necessary number of samples required to achieve a 95 % statistical confidence level in determining the homeostatic set point for routine coagulation tests within a variation of ±5 %, ±10 %, and ±20 %. The number of samples for the homeostatic set point, as per our calculations, fluctuated between 2 and 19 at the 5 % level, between 1 and 5 at the 10 % level, and consistently remained at 1 for the 20 % level of variation.

Discussion

The present study aimed to determine the biological variation in routine hemostasis parameters, specifically PT, aPTT, and fibrinogen, in a healthy population. The significance of this study lies in its potential to contribute to a more accurate interpretation of laboratory results, enabling physicians to make informed diagnostic and treatment decisions. The high relevance of these parameters in a wide range of clinical scenarios, such as bleeding disorders, thromboembolic events, and monitoring of anticoagulant therapy, makes our findings critical for improving patient outcomes. There is still a sizable gap in the literature regarding the estimation of biological variability in common coagulation parameters. The CV_I, CV_G values for the PT, aPTT, and fibrinogen tests were recently updated in the EFLM BV database in accordance with a systematic review and meta-analysis, as shown in Table 1 [6].

It is expected to observe varying BV data across different populations for the parameters under investigation. Influences such as dietary patterns, specific genetic factors (for example, fibrinogen gene polymorphism), and smoking habits can significantly impact coagulation results. The II values procured further substantiate that these parameters hold utility for both monitoring and diagnostic applications. Consequently, both RCV and population-based reference intervals should be employed for this purpose [6].

Our study demonstrated a CV_I for PT of 3.2 % compared to the database CV_I of 2.6 %. The slightly higher CV_I value in our study may be due to differences in the sample handling and measurement techniques. This could also be attributed to the larger diversity of the study population with respect to age, diet, lifestyle, or other undetermined variables. However, despite these factors, the CV_I obtained was reasonably close to the BV database, suggesting a consistent degree of variation within individuals over time. On the other hand, the CV_G in our study was 5.8 %, which was slightly higher than the reported CV_G of 5.1 % in the database. Falay et al. reported the CV_I and CV_G for PT to be 2.78 and 5.07 %, respectively. Notably, their study utilized a similar methodology and subject group to our own. Falay et al. reported an RCV of 9.28 % for PT. In contrast, our study determined the asymmetric RCV values for PT to be within the range of −7.5 to 8.1 %. While the slightly lower RCV values observed in our study compared to those reported by Falay et al. warrant consideration, they do not contradict the larger body of evidence on biological variation in PT. In fact, they contribute to a richer understanding of potential variations, providing a valuable context for clinicians to interpret PT results. However, this finding has some implications that are worth exploring. For example, a low within-subject variation suggests that PT may not be sensitive enough to detect minor changes in hemostasis parameters. Physicians may need to consider using other tests in conjunction with PT to ensure a comprehensive evaluation of hemostasis parameters [12].

In our study, the Index of Individuality (II) values we found were higher than 0.6, differing from what is reported in the existing database. When II values fall below 0.6, it can limit the effectiveness of using population-based reference intervals. However, the II values we estimated did not exceed the recommended 1.4 thresholds, a criterion necessary for the optimal use of population-based reference intervals [13].

An observed II of 0.73 suggests that population-based reference intervals may not be ideal for monitoring PT changes within an individual, and personalized reference ranges may be more suitable. This highlights the importance of understanding the individual variations in laboratory parameters and the potential benefits of personalized medicine in laboratory diagnostics. In summary, while the II value observed in our study was higher than that reported by Falay et al., both studies highlight the significance of individual-specific factors in determining PT values. This underscores the need for further research aimed at improving our understanding of the individual variations in coagulation profiles, which could ultimately enable more personalized approaches to coagulation management [12].

Estimating the BV of the aPTT test, it is evident that the findings can differ significantly across various studies. The BV database reports CV_I and CV_G values at 2.8 and 7.2 %, respectively, whereas Falay et al. noted slightly lower values, with CV_I at 2.26 % and CV_G at 4.9 %. In contrast, our study found higher values for both CV_I and CV_G, at 5.8 and 8.4 %, respectively. Furthermore, a study conducted by Chen et al. reported values of 3.48 % for CV_I and 10.39 % for CV_G, which fall between our results and those reported in the BV database and by Falay et al. Similarly, Moniek et al. found a CV_I of 6.4 % and CV_G of 7.1 % [5, 14].

Such discrepancies in results could be attributed to multiple factors, including differing storage conditions and sample stability under various conditions, which are known to affect aPTT values. These elements might have played a significant role in our study, potentially leading to the higher CV_I and CV_G values observed [15].

In relation to the aPTT test, the BV database has published an II value of 0.388, whereas Falay et al. reported a slightly higher value of 0.461. Chen et al., on the other hand, documented a lower II value of 0.33. Notably, Moniek et al. observed a significantly higher II value (0.90). Our study found a similarly high II value of 0.91, aligning closely with the findings of Moniek et al. However, because our II value is still less than 1.0, it underscores that there is a meaningful degree of variation between individuals. Consequently, individual-specific reference intervals may be more appropriate for interpreting the aPTT test results [5, 12, 14].

The RCV is a crucial statistical measure that represents the smallest significant change in a test result for a particular individual. Historically, the RCV has been computed symmetrically, assuming an equal probability of a test value increasing or decreasing. However, this approach does not always account for the inherent skewness often present in biological data. In response, it has been proposed that RCV should be calculated asymmetrically using a natural logarithmic transformation (lnRCV) to better represent the real-world distribution of biological data. The EFLM BV database utilizes this asymmetrical approach for computing RCVs. Therefore, the RCVs listed in this database more accurately reflect the inherent asymmetry of biological data and enhance the clinical relevance and utility of these RCVs [6].

In our study, we adopted this asymmetrical approach to compute RCVs for all analytes. Our asymmetrically calculated RCV value for the aPTT test was −12.7 to 14.6. This range indicates that for an individual in our study, a change in the aPTT test result would need to decrease by more than 12.7 % or increase by more than 14.6 % to be considered significant [2].

Our study’s estimation of the biological variation of aPTT presented significant discrepancies compared to previous investigations, largely due to an unusually high number of outliers in our data. These outliers necessitated rigorous data processing and their subsequent removal to maintain data integrity. Despite a reduced sample size, our results, achieved through stringent testing procedures, provide an insightful understanding of BV in aPTT. The prevalence of outliers in our study underscores the complexities of BV studies and the importance of rigorous data handling and exclusion criteria in ensuring reliable and valid outcomes.

In our study, fibrinogen was found to have the highest degree of biological variation among the analyzed parameters, with a CV_I of 11 % and CV_G of 17.4 %. These findings suggest that fibrinogen levels fluctuate considerably both within individuals over time and between different individuals and that the BV database reports very similar values, with a CV_I of 10.2 % and a CV_G of 17.1 % for fibrinogen. This close alignment between our results and those of the BV database reinforces the validity of our findings and provides further evidence of the significant biological variation inherent in fibrinogen measurements; the range of values reported in previous studies for fibrinogen, with CV_I values ranging from 5.6 to 11.9 % and CV_G values ranging from 8.53 to 17 %, further corroborates our results and the BV database values. In our study, the II value for fibrinogen was calculated to be 1.31. This value, greater than 1, indicates that there is more variation in fibrinogen concentrations within individuals over time than between different individuals. In practical terms, this suggests that population-based reference intervals are more effective and appropriate for interpreting fibrinogen test results in our study population. In a study by Falay et al., the II value for fibrinogen was similarly greater than 1, reported as 1.22. This aligns closely with our findings and suggests a higher degree of intra-individual variation in fibrinogen levels. In contrast, Chen et al. reported an II value of 0.31 for fibrinogen, which is less than 1. In addition, the BV database reported an II value of 0.596 for fibrinogen. For fibrinogen in our study, the calculated asymmetric RCV values ranged from −22.7 % to 29.4, and Falay et al. reported a symmetrical RCV of 29.06 % [5, 12].

Differences in the observed CV_I and CV_G, II, and RCV across studies can be attributed to several factors. These include the state of the plasma samples (fresh or frozen), storage conditions, various methodologies used, calculation methods, timeframe of the study, and timing of sample collection. Each of these factors introduced significant variations.

These discrepancies underline the complexity of accurately determining biological variations. They stress the importance of considering these variables when comparing the results of different studies or planning new research. However, our study had some limitations: only three hemostasis parameters were assessed, and the inclusion of additional parameters, such as D-dimer and platelet function tests, could provide a more comprehensive overview of biological variations in hemostasis testing.

The APS values we calculated matched those on the EFLM database, except for aPTT. The differences can be accounted for by factors such as the sample size, whether the samples were fresh or frozen, or the methodology used.

Conclusions

It is evident that the biological variation in routine hemostasis parameters, specifically PT, aPTT, and fibrinogen, is significant and has implications for clinical diagnoses and treatment decisions. The study highlighted the complexities of accurately determining biological variations and the importance of rigorous data handling and exclusion criteria in ensuring reliable and valid outcomes. The study also underscored the need for considering these variables when comparing the results of different studies or planning new research. Despite some limitations, the findings contribute to a richer understanding of potential variations, providing a valuable context for clinicians to interpret results from hemostasis tests. Future research should consider including additional parameters for a more comprehensive overview of biological variations in hemostasis testing.