Immune biomarker reference range estimation for healthy paediatric patients in South Africa

Date
2017-03
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT : Understanding and quantifying human peripheral blood T-lymphocyte immunophenotypes is necessary for the diagnosis and treatment of immune and haematological disorders. This is only possible if comparison to control-data from healthy subjects is available for all the biomarkers of interest. Historical empirical studies in industrialized countries have described normal reference ranges for such biomarkers in children by grouping particular age ranges into ‘age-blocks’. Since such markers change with age this has resulted in a loss of precision in determining whether patients that lie close to the limits of age-ranges are normal or not. Previous studies have relied on fitting single exponential models to such data, which makes the simple assumption of an exponential decline in cell markers with age. However, the counts of such markers have been observed to increase from birth to between 6 months to 12 months from birth and then decrease continuously with age. There is a dearth of reference range estimation methods which are age-continuous and incorporate biologically mechanistic models. A more ideal solution would be the development of appropriate mathematical models and modelbased, age-continuous reference range estimation methods that describe such changes in a continuous manner. Such models may then be used to investigate the influence of population covariates on age-related changes in the biomarkers of interest. In this study, we employ paediatric data from a cohort of 381 healthy South African children. This is cross-sectional in design and the biomarkers described include: CD3+, CD19+, CD8+, CD4+, ratio of CD4+ naive/memory, CD18+CD56+ and CD3-CD56+. Using weighted generalized nonlinear least squares, we fit and compare single and double exponential semi-mechanistic models. An ideal model is selected based on the Akaike’s Information Criterion (AIC). The double exponential model is found to be the best fit for age-related changes in such biomarkers. This predicts that cell counts rise after birth to a maximum at approximately 12 months of age and decline in an exponential manner towards an asymptote in adulthood. This is in agreement with prior empirical and mechanistic studies. We extend the double exponential model to investigate the influence of particular covariates. The type of feeding in the first 6 months following birth is found to be the covariate with the greatest influence on age-related changes in the majority of the biomarkers investigated. A model-based method to estimate age-continuous reference ranges is then proposed. This assumes that particular reference ranges are a specified shift of the ‘running’ standard deviations of residuals away from a fitted central model function by a Z-score. We compare this method to reference ranges calculated using traditional centile curves. Centile curves demonstrate a simpler negative singleexponential decline as age advances and enable no mechanistic interpretation for this pattern. The models employed in this study may lead to the development of a laboratory tool by which individual cell-marker values may be compared to healthy agecontinuous reference ranges.
AFRIKAANSE OPSOMMING : Die diagnose en behandeling van immuun en hematologiese afwykings benodig die begrip van en kwantifisering van menslike perifere bloed T-limfosiet immuun-fenotipes. Dit is slegs moontlik indien vergelyking met data van gesonde pasiente beskikbaar is vir al die biomerkers van belang. Historiese empiriese studies in geïndustrialiseerde lande het al sulke normale verwysings waardes vir kinders beskryf, maar deur groepering van die data in spesifieke ‘ouderdoms-blokke’. Aangesien sulke biomerkers aanhoudend verander met ouderdom, het dit gelei tot ‘n verlies aan akkuraatheid in die bepaling van normaliteit van pasiënte wat naby aan die grense van sulke ouderdoms-groepe lê. Ander studies het staatgemaak op die pas van enkel-eksponensiële modelle op sodanige data, wat die eenvoudige aanname maak van ‘n eksponensiële afname in sel merkers met ouderdom. Dit was wel al waargeneem dat sulke merkers verhoog vanaf geboorte tot tussen 6 en 12 maande na geboorte en dan voortdurend daal met vergrotende ouderdom. Daar is ‘n gebrek aan verwysings waarde skattings metodes wat ouderdomaaneenlopend is en biologies meganistiese modelle inkorporeer. ‘N Meer ideale oplossing sou die ontwikkeling wees van toepaslike wiskundige verwysings waarde skattings metodes wat model-gebaseer en ouderdoms-aaneenlopend is. Sulke modelle kan dan gebruik word omdie invloed van koveranderlikes wat ouderdoms-verwante veranderinge in die biomerkers van belang veroorsaak te ondersoek. In hierdie studie gebruik ons die ouderdoms-deursnee data van ‘n groep van 381 gesonde Suid-Afrikaanse kinders. Wat die volgende biomerkers insluit: CD3+, CD19+, CD8+, CD4+, die verhouding van CD4+ naïef/geheue selle, CD18+, CD56+ en CD3-CD56+. Ons gebruik dan nielineêre kleinste kwadrate metodes om enkel en dubbel eksponensiële semi-meganistiese modelle te pas en vergelyk. Die ideale model is gekies op grond van die Akaike se inligtings Maatstaf. Sodoende bepaal ons dat die dubbel eksponensiële model die mees geskik is om ouderdoms-verwante veranderinge in sulke biomerkers aan te dui. Hierdie model voorspel dat seltellings styg na geboorte tot ‘n maksimum by ongeveer 12 maande en daarna eksponensiël daal tot ‘n asimptoot in volwassenheid. Hierdie resultaat stem ooreen met vorige empiriese en meganistiese studies. Daarna gebruik ons die dubbele eksponensiële model om die invloed van koveranderlikes te ondersoek. Die koveranderlike met die meeste invloed was die tipe voeding in die eerste 6 maande na geboorte. Model gebaseerde metode wat ouderdoms-aaneenlopende verwysings waardes skat is daarna voorgestel. Hierdie metode veronderstel dat die verwysings waardes ‘n gespesifiseerde Z-verskuiwing van die ‘lopende’ standaardafwykings van die residue vanaf die sentraal gepaste model funksie is. Ons vergelyk hierdie metode met tradisionele persentiel kurwes. Persentiel kurwes toon ‘n eenvoudiger negatiewe enkeleksponensiële daling met ouderdom, en is nie in staat om meganistiese interpretasies te maak vir hierdie patroon nie. Die modelle wat in hierdie studie voorgestel is mag lei tot die ontwikkeling van ‘n laboratorium instrument waarmee individuele sel-merker waardes kan vergelyk word met gesonde ouderdoms aaneenlopende verwysings waardes.
Description
Thesis (MSc)--Stellenbosch University, 2017
Keywords
Reference ranges (Medicine), UCTD, Laboratories -- Tools, Immunology -- Research, Immune system biomarkers -- Mathematical models, Pediatric patients -- South Africa
Citation