Establishment of biochemistry reference values for healthy Tanzanian infants, children and adolescents in Kilimanjaro Region

Introduction

Few data exist on biochemistry reference intervals for healthy infants, children and adolescents in sub-Saharan Africa. Most reference intervals currently in use are derived from North American or European populations and may not be reflective of population values elsewhere. Interpretation of laboratory results using reference ranges established elsewhere in dissimilar populations is potentially problematic for both clinical care and for clinical trials participation. The proper medical evaluation of children is largely dependent on correct interpretation of laboratory results. Such interpretation requires relevant reference intervals, ideally established in the local setting. Relying on reference intervals from other countries or regions may lead to inappropriate clinical management of children and an underestimation or overestimation of disease states 1, 2. Locally validated reference ranges are also important for clinical trials participation, as the use of inappropriate clinical laboratory reference ranges may result in an unnecessary exclusion of patients from participation in important trials or an inability to reliably assess drug-related toxicities and adverse events.

It is well established that differences occur in haematologic and immunologic parameters among African children compared to their European or North American counterparts. Multiple studies have shown lower haemoglobin values for African children compared to children from the Western Hemisphere 3-5. The 2.5th percentiles for haemoglobin, haematocrit and mean corpuscular volume (MCV), for example, are substantially lower for Tanzanian children compared to North American ranges 6. Whether differences also exist for biochemistry reference intervals is largely unknown.

The aim of this study was to determine the local reference ranges for biochemistry values among a healthy paediatric and adolescent population living in the Kilimanjaro Region of northern Tanzania. Values based on age and gender were also determined, and local results were compared to levels reported in children in North America to the DAIDS toxicity grading tables commonly used in clinical trials.

Methods

Results

A total of 633 children were screened for study participation; 12 were excluded due to the presence of fever; and two were excluded because of a positive HIV-1 antibody test. No females were excluded due to a positive urine pregnancy test. In total, 619 children met all eligibility criteria and were enrolled into the study. Within each age cohort, single values were excluded if they met the criteria for outliers as described in the methods section. The age range was 1 month to <18 years. Table 1 shows the median value and 95% reference interval (RI) for each of five age groups: <12 months, ≥1 to <5 years, ≥5 to <13 years and ≥13 to <18 years. The oldest age cohort is further analysed by gender.

Several chemistry parameters showed notable differences compared to the US-derived reference intervals (Table 1). Alkaline phosphatase among US infants ranges from 150 to 420 U/l 18, and in our infant cohort, the range extended up to 589 U/l. For adolescent males and females, the US reference ranges extend to 390 and 320 U/l, respectively, compared to our cohorts where values approached 500 U/l and higher. Amylase values for our infant, child and adolescent populations were much higher than standard US parameters (Table 1). Our lipase values are also higher than standard US reference values, but are most notable for our cohorts of children ≥1 to <5 years and adolescent girls ≥13 to <18 years of age (Table 1). Finally, phosphorus levels for our Tanzanian cohort tended to be higher across all ages compared to US reference ranges (Table 1).

We also compared our data to the DAIDS grading criteria for classification of adverse events 19, and found that for select parameters, up to 15% of infants or children in certain age categories would be classified as having an adverse event as defined by DAIDS. Among the Tanzanian cohort, the number of children falling outside the reference range based on the DAIDS tables was most notable for low serum glucose across all age ranges, high calcium ranges among infants and young children, and both high and low potassium among two of our child cohorts (Table 2).

Discussion

The availability of accurate and validated reference intervals is fundamental to the practice of medicine, and there is a genuine need to establish paediatric reference intervals for children in resource-poor settings such as in sub-Saharan Africa. The main objective of this study was to establish locally validated reference intervals from a healthy Tanzanian paediatric population that can be applied in the future for similar populations. With such reference data lacking, many clinicians and researchers still use American or European-derived intervals which may not be suitable for the population at hand.

While Tanzanian reference intervals for many of the analytes in our study were comparable to US values, four analytes in particular stood out as being quite different. The ranges for alkaline phosphatase for healthy Tanzanian children were higher across all ages compared to US children. Alkaline phosphatase is present throughout the entire body but is predominantly concentrated in the liver, kidneys, bile duct and bones. It is not uncommon for alkaline phosphatase to be elevated in healthy growing children. While our values do not approach those of transient hyperphosphatasemia (TH) which occurs commonly in infants and young children (>1000 U/l) 20, it exceeds the comparable upper bounds of US ranges by as much as 170 units for some ages. We did not perform isoenzyme analysis, so we were unable to investigate the specific source of elevated alkaline phosphatase.

The phosphorus reference range for our Tanzanian population was also higher than US reference ranges across all age groups. The clinical relevance of this is uncertain, and is likely not meaningful as phosphorus reference ranges increase with decreasing age, with premature infants having the highest reference range 21. The causes of hyperphosphatemia are multiple, and may include endogenous causes of an acute phosphate load such as tumour lysis syndrome or rhabdomyolysis, or exogenous causes such as vitamin D toxicity or phosphate-containing medications such as laxatives 22, 23. Decreased renal clearance is perhaps the most common cause of hyperphosphatemia, which may be related to acute kidney injury, chronic kidney disease, hypoparathyroidism, pseudohypoparathyroidism and acromegaly, among others 23.

Most interesting from our study are amylase ranges across all age groups, which were higher than the reference ranges listed for healthy US counterparts. The upper limit of normal beyond 1 year of life is 100 U/l for US children, while our upper limit ranges extended as high as 246 U/l in some age groups. Similarly, for our infant cohort, the range extended to 183 U/l, while for US ranges, it extends only to 50 U/l up to the age of 6 months and 80 U/l up to the age of 12 months. While the lipase reference ranges for our Tanzanian paediatric cohorts are also higher than US counterparts of the same age, the magnitude is not as great as that seen for amylase. There are several possibilities that could explain hyperamylasaemia in our paediatric population. Elevated amylase along with a parallel increase in lipase would point to an inflammatory state of the pancreas, such as pancreatitis. As all of our children were required to be physically well for study entry, such a diagnosis in any child in our cohort is unlikely. Secondly, we did not perform amylase isoenzyme testing, and thus cannot differentiate between salivary and pancreatic amylase, the two major sites of amylase production. As the children in our study were clinically well, a possible explanation for their elevated amylase levels could be what is now termed chronic asymptomatic hyperamylasaemia unrelated to pancreatic diseases (CHUPD) 24. CHUPD can typically be classified further into macroamylasaemia, benign pancreatic hyperamylasaemia and salivary hyperamylasaemia. Again, without being able to perform isoenzyme testing and urinary amylase levels, we were unable to make these diagnoses. Mortelé et al. assessed the magnetic resonance cholangiopancreatographic findings of patients with asymptomatic hyperamylasaemia and hyperlipasaemia and found that more than 50% had pancreatic abnormalities, with pancreas divisum in particular being found more often than in the general population 25. Coeliac disease and cystic fibrosis are two diseases that would typically be entertained in a paediatric patient with persistent pancreatic hyperenzymemia 26, but again, we would expect our patients to show some minimal symptoms or signs if this were the case. Finally, familial hyperamylasaemia is also described in case reports and points to an autosomal dominant pattern of inheritance 27, 28. Regardless of the aetiology of the hyperenzymemia, especially the hyperamylasaemia these values represent the reference ranges for our Tanzanian paediatric population and may represent no pathology whatsoever.

We specifically chose to compare data from this cohort to the DAIDS grading criteria for adverse events (AE) as this has important implications for clinical trials participation, particularly in sub-Saharan Africa, where recruitment for paediatric HIV studies is often sought. These criteria are a set of descriptive terminology that can be utilised for AE reporting, with a severity scale provided for each AE term. The tables include parameters for both adult and paediatric AEs, and are typically graded from grade 1 (mild) to grade 4 (potentially life-threatening) 19. We found that more than 35% of infants in the present cohort would have been classified as having a grade 1 hypercalcaemia, for example. While grade 1 laboratory abnormalities in and of themselves may not cause clinical symptoms, the classification of a participant's laboratory result as abnormal becomes the problem. If this same participant developed a more marked hypercalcaemia, he or she may be classified as having a grade 3 or 4 laboratory abnormality that may have more serious consequences in terms of study participation. Similarly, almost 7% of our school-age children had a low serum glucose and almost 6% of children ages 1–5 years had a grade 1 hyperkalaemia. Thus, using the DAIDS adverse event grading scale for our population would lead to an over-reporting of adverse events in our paediatric population or to unnecessary exclusion of healthy children from participation because their baseline test results, although normal for the population, may be classified as in the toxicity range.

Our study had several limitations. First, the Clinical and Laboratory Standards Institute (CLSI) guidelines for establishing reference intervals recommends a minimum of 120 observations for each analyte and age range under study, which has the advantage of allowing for the nonparametric computation of 90% confidence limits for each reference limit (Clinical and Laboratory Standards Institute 17 2008). While we originally recruited a minimum of 120 participants for each age range, we were able to use <120 participants for some analytes due to the exclusion of outliers. Second, our infant values for potassium were quite high, likely in part due to haemolysis and the inherent difficulty in obtaining blood draws from our youngest cohort. Third, we did not record anthropometric data on study participants, which might have enabled us to detect some cases of malnutrition. This may have been important as the nutritional status of a child can affect certain biochemistry parameters. However, children meeting criteria for severe malnutrition (such as marasmus or kwashiorkor) would not have been eligible for study participation, simply because of ill appearance. The study nurses were quite experienced in identifying such children, and it would be unlikely that mild or even moderate malnutrition, which could be overlooked without anthropometric data, would considerably alter plasma chemistries. Finally, the children in our cohort were not required to fast before laboratory testing, and therefore, all cholesterol and triglyceride levels are non-fasting. Whether children truly need to be in a fasting state prior to lipid screening is somewhat debatable. A review of over 12 000 children with lipid levels (total cholesterol, triglycerides, LDL and HDL) collected via the National Health and Nutrition Examination Survey (NHANES), compared non-fasting children with fasting children. The results yielded only small differences in lipid levels that the authors conclude are not clinically meaningful 29. Still, our values are higher than for US reference ranges, particularly for total triglycerides, and ideally, these should be repeated in a fasting state for those children with abnormal levels.

In conclusion, to our knowledge, we have established the first set of locally validated biochemistry reference intervals for children in Tanzania. These reference intervals demonstrate that healthy African children do not always have the same reference values for some parameters as their US or European counterparts of the same age. This information will provide physicians in Tanzania with the appropriate ranges they need to correctly interpret laboratory data and provide the highest quality of care to patients, and will assist researchers and scientists in the correct management of clinical trials participants to ensure proper management of study participants and the highest quality research data.