Pediatric reference intervals for alkaline phosphatase

Introduction

Alkaline phosphatase is a ubiquitous enzyme in humans and an important biomarker for skeletal and hepatobiliary diseases [1], [2], [3]. Clinical interpretation is enabled with reference intervals, which allow classification of samples in the context of intra- and inter-individual variation. However, growth, changes in bone metabolism, and changes in the hepatobiliary system result in a pronounced age- and sex-dependent variability of alkaline phosphatase activity during childhood, especially in infancy and puberty. These dynamics must be considered when defining and using reference intervals for alkaline phosphatase activity to distinguish between change due to physiological development and change due to disease.

An overview of published reference intervals for alkaline phosphatase established using current analytical methods is shown in Table 1. Age-dependent changes are represented either with separate reference intervals for different age groups or using reference intervals which change continuously with age. While separation into age groups is well-established in clinical practice and allows easy integration of reference intervals into current laboratory information systems, this approach can lead to misclassification of samples, especially at the margins of age groups (i.e. when upper and lower reference limits change abruptly due to changes in age). Continuous reference intervals reflect the continuous change of reference intervals with age and therefore allow a more accurate representation of the physiological dynamics of alkaline phosphatase activity. The most sophisticated approach to alkaline phosphatase activity interpretation are percentile charts analogous to other developmental quantities (e.g. weight- and height-for-age charts), which enable a more precise consideration of intra- and inter-individual differences than a categorization relative only to the reference interval boundaries. However, large numbers of samples are needed to create continuous representations of reference intervals and percentile charts. As access to blood samples from healthy children is restricted by ethical and practical objections, availability of percentile charts and continuous reference intervals for alkaline phosphatase from birth to adulthood is limited, leading to limitations when interpreting alkaline phosphatase activity in childhood [24].

In the studies listed in Table 1, upper and lower reference intervals for alkaline phosphatase activity are provided. Less recent studies, including surveys referred to by assay manufacturers [25], [26], only report upper reference limits because of a focus on diseases characterized by elevated alkaline phosphatase activity. However, this approach results in misclassification of abnormally low values and may result in missing medical conditions associated with reduced alkaline phosphatase activity. The introduction of therapeutic agents for hypophosphatasia and the accompanying marketing have emphasized the importance of lower reference limits for alkaline phosphatase activity as a screening marker for this congenital condition. Continuous classification of test results with percentile charts from birth to adulthood can improve the usefulness of alkaline phosphatase activity as a screening marker for hypophosphatasia and related conditions.

In a recent study – based on a data-mining approach using a single center’s laboratory data – we demonstrated the complex pattern of change in alkaline phosphatase activity over time, with a high accuracy of age- and sex-dependent changes in infancy and puberty [6]. However, representation of the extensive dynamics after birth in neonates (0–28 days) with a high age-specific accuracy, and the creation of percentile charts require analysis of a more comprehensive data set.

The aim of this report is the creation of continuous reference intervals and percentile charts for alkaline phosphatase activity from birth to adulthood. To overcome restrictions due to sample size limitations, we employ a data-mining approach using multi-center laboratory data. This allows special consideration of age-specific accuracy in the first days of life and the creation of percentile charts to enable precise interpretation of patient test results over time.

Materials and methods

An established indirect method was used to generate continuous percentile charts [6].

Results

We analyzed 361,405 samples from 124,440 different patients to create percentile charts for alkaline phosphatase activity from birth to 18 years. 192,972 samples from 64,670 boys and 168,433 samples from 59,770 girls were examined. Table 2 shows the age- and sex-specific distribution of samples with a substantial proportion of samples in the newborn period and infancy. Graphical representations of the percentile charts created are available in Figure 1 (2.5th, 10th, 25th, 50th, 75th, 90th, and 97.5th percentiles) and Supplemental Figure 3 (1st, 2.5th, 10th, 25th, 50th, 75th, 90th, 97.5th, and 99th percentiles). Data tables with age-specific percentile values are available in the online Supplemental Database and allow incorporation of our results into laboratory information systems.

Figure 1:

Age- and sex-dependent percentile charts for alkaline phosphatase activity, showing the 50th percentile (solid lines, blue), 25th and 75th percentiles (dashed lines, green), 10th and 90th percentiles (dashdotted lines, orange), and 2.5th and 97.5th percentiles (dotted lines, red); the 1st and 99th percentiles are shown in Supplemental Figure 3.

The x-axes are scaled differently on the left- and right-hand sides of the figure to account for the dynamics in young infants; the x-axis on the right-hand side starts at 61 days.

Our results demonstrate the complex age- and sex-related dynamics in alkaline phosphatase activity: Levels surge after birth and peak at approximately 20 days of life. After this, activity decreases and reaches a local minimum at about 4 years of age, succeeded by a sex-specific rise and subsequent decline. Activity in girls peaks at 10–12 years with a median and maximum activity of 240 and 400 U/L (50th and 97.5th percentile, respectively). Males’ activity peaks at 13–15 years with similar median activity (50th percentile, 250 U/L) and higher maximum activity (97.5th percentile, 450 U/L).

Comparison of our 2.5th and 97.5th percentile results with findings from the CALIPER study [11] demonstrates similar reference limits and age-dependent dynamics (Figure 2). In the first 6 months of life, the upper reference limits of the two studies intersect, but those provided by us are higher than those from CALIPER for a substantial period of time. However, direct comparison of results from these studies is complicated by the different strategies for age-stratification and different analytical devices (Roche Cobas and Abbott Architect). Comparison of the different approaches to represent age-dependent dynamics demonstrated in Figure 2 illustrates the limitations of discrete age groups to represent the dynamics of an analyte changing continuously with age.

Figure 2:

Comparison of 2.5th and 97.5th percentiles for alkaline phosphatase activity (solid lines) to reference intervals from the CALIPER study (dotted lines) [11].

The x-axes are scaled differently on the left- and right-hand side of the figure to account for the dynamics in young infants; the x-axis on the right-hand side starts at 61 days.

Discussion

We report percentile charts for alkaline phosphatase activity from birth to adulthood generated with a multi-center data-mining strategy (Figure 1). This approach has provided a large number of samples for evaluation (n=361,405), even for the challenging subgroups of neonates (birth to 28 days, n=11,513) and infants (29 days to 1 year, n=30,780). Reference intervals for these age groups are often restricted by limited access to samples due to ethical and practical objections, although the rapid age-dependent changes in alkaline phosphatase activity after birth require a disproportionate number of samples for the construction of accurate reference intervals. Consequently, our results show a steady rise in alkaline phosphatase activity immediately after birth – a finding not demonstrated in other reference intervals due to partitioning into large-scale age groups or reduced sample availability.

The percentile charts provided allow a continuous representation of change in alkaline phosphatase activity with age. This distinguishes them from the majority of current reference intervals, which use discrete age groups to represent age-dependent dynamics (Table 1). Partitioning into discrete age groups with significant differences, however, can only approximate the physiological dynamics leading to change in alkaline phosphatase activity. This results in difficulties in interpretation of patient test results at age group margins, as the classification of identical measurements can change from pathological to normal or vice-versa due to a biologically negligible change in patient age. Continuous implementations of reference intervals as percentile charts, analogous to other quantities typically specified in relation to age (e.g. weight- and height-for-age charts), eliminate this problem. Similarly, classification of test results according to percentiles allows a more precise consideration of intra- and inter-individual differences than a categorization relative only to reference interval boundaries (e.g. “inside” the reference interval or multiples of the “upper limit of normal”). Use of percentiles also reflects the fact that healthy children may have test results outside the reference interval (i.e. below the 2.5th percentile or above the 97.5th percentile) and vice-versa (i.e. not all values between the 2.5th and 97.5th percentiles reflect physiological alkaline phosphatase metabolism). Interpretation of test results with percentile charts therefore seems especially appropriate for an analyte as closely linked to growth as alkaline phosphatase.

We selected a data-mining method (indirect method) for the determination of reference intervals. Most available reference intervals were created using population-based approaches (direct methods), in which blood samples are drawn from healthy children. Population-based methods are generally regarded as the best method available for the determination of reference intervals [24] because they allow a precise control of the reference population and preanalytical factors, and are not subject to the uncertainties intrinsic to data-mining approaches. The use of an indirect method might therefore be regarded as a severe limitation of our study and might oppose application of the published reference intervals.

However, the ethical and practical challenges associated with blood sampling in healthy children required for population-based methods cause restrictions in the number of samples available for analysis, which reduces the age-specific precision in these reference intervals and often restricts analysis to certain age groups. The calculation of continuous reference intervals from birth to adulthood for alkaline phosphatase activity using population-based methods has therefore been prevented by the excessive requirement of samples for such an approach. Meanwhile, data-mining approaches give access to a large number of measurements from clinical laboratories and have therefore fewer restrictions regarding age-specific dynamics. To ensure valid reference intervals when using indirect procedures, great care must be taken in the selection of samples and statistical algorithms. Previous evaluations of the applied algorithm [6], [27] and comparison of our results to reference intervals from CALIPER, created using a direct method (Figure 2), demonstrate the selected approach’s ability to generate correct reference intervals. In the context of an analyte subject to such extensive age-dependent changes as alkaline phosphatase, the benefits of an increase in sample numbers make data-mining approaches especially appropriate.

Although comparison of our results to reference intervals from CALIPER shows consistent upper and lower reference limits, there are noteworthy differences in the first months of life. The 97.5th percentiles provided by us exceed the upper reference limits reported by CALIPER for a substantial period of time (Figure 2), which may raise doubts about the validity of our approach. However, interpretation of these differences requires inspection of the distribution in CALIPER’s age groups: Reference intervals for children from birth to 14 days are taken from 46% samples of children ≤4 days old and 96% samples from children ≤7 days. Similarly, in the CALIPER data set for children aged 15 days to 1 year only 7% of samples were from children ≤30 days. In total, 16 samples from children aged 8–30 days are included in the CALIPER database, precluding a comparison of reference limits in this age group. As available reference intervals – including those from CALIPER – show a significantly lower alkaline phosphatase activity immediately after birth than in the following months, there is a strong indication of a rapid rise in alkaline phosphatase activity after birth. This surge is presumably masked by the low number of children aged 8–30 days (the age group showing most disagreement) and the selection of age groups (0–14 days, and 15 days to 1 year) in CALIPER’s and other’s reference intervals.

We established percentile charts for alkaline phosphatase activity in a German population of mainly Caucasian origin on the Roche Cobas platform. The reported reference intervals are therefore directly applicable only for this population and analytical platform and require transference according to guidelines before usage outside this context [28], [29]. However, the physiological dynamics in alkaline phosphatase activity with age are independent from analytical procedures; our results can therefore be used to extrapolate existing reference intervals, even if our results cannot be transferred directly.

A variety of reference intervals and percentile charts for alkaline phosphatase established with current analytical methods have been published (Table 1). While these studies provide important insights into physiological dynamics and support clinical decision making, only one investigation by Loh and Metz reports continuous reference intervals from birth to adulthood [5]. Their results close an important gap in pediatric reference intervals using an indirect data-mining approach – laboratory data from primary care providers were analyzed and all test results from children tested only once were considered to be normal and used to construct percentile charts, allowing the accurate representation of the age-dependent dynamics for the majority of children. However, an unknown number of children will have visited their doctor due to an underlying disease affecting alkaline phosphatase activity and not undergone repeated testing, resulting in uncertainty in the 2.5th and 97.5th percentiles. On the other hand, this approach to sample selection may lead to a bias towards the existing reference intervals, as physicians might tend to retest children with results “outside the reference range”. To address these issues, we performed an independent analysis of an increased number of children from another population using a different indirect approach.

Our multi-center analysis of laboratory data from six pediatric tertiary centers and one laboratory service provider has resulted in the most comprehensive data set of alkaline phosphatase activity available, containing 361,405 samples from 124,440 different children. Furthermore, the employment of identical analytical methods at distributed locations has allowed us to create reference intervals for different laboratories using the Roche Cobas platform. Considering the lack of continuous reference intervals from birth to adulthood, our results are an important addition to our understanding of the changes in alkaline phosphatase activity with age and allow a more precise consideration of these dynamics when guiding clinical decisions.

Conclusions

We provide sex- and age-dependent percentile charts for alkaline phosphatase activity from birth to adulthood. These charts allow an accurate appreciation of the changes in analyte activity with age and can therefore improve clinical decision making.