Risk factors relate to the variability of health outcomes as well as the mean: A GAMLSS tutorial

Establishment of gender- and age-related reference intervals for serum uric acid in adults based on big data from Zhejiang Province in China

OBJECTIVES

This study utilized large-scale health examination data to explore gender- and age-specific reference intervals (RIs) for serum uric acid (UA) using indirect methods and assessed the consistency of different approaches.

METHODS

UA data were collected from a hospital in Zhejiang Province, China. The test set covered January 2019 to December 2023, with a validation set from January to June 2024. Various methods - EP28 nonparametric (EP28-NP), parametric (EP28-P), TMC, refineR, and Kosmic - were used to establish gender- and age-specific RIs. Continuous age-based RIs were derived using the Generalized Additive Model for Location Scale and Shape (GAMLSS). Validation rates were calculated for each method using the validation set.

RESULTS

Using EP28-NP as the benchmark, other methods showed similar UA RIs (bias ratios ≤0.375, except for one group), with Kosmic, refineR, and TMC yielding slightly higher values than EP28-NP and EP28-P. For males, UA RIs varied by age: 19-42 years (256-537 μmol/L), 43-66 years (235-513 μmol/L) and ≥67 years (214-515 μmol/L), with validation rates ranging from 94.05 to 96.50 %. Male continuous RIs declined from ages 20-79 and then gradually increased after age 80. For females, UA RIs were age-dependent: 19-48 years (169-374 μmol/L), 49-74 years (178-405 μmol/L), and ≥75 years (186-470 μmol/L), with validation rates ranging from 92.70 to 96.80 %. Female continuous RIs decreased from ages 20-48, then increased significantly from age 49 onward.

CONCLUSIONS

Three indirect methods and two EP28 methods demonstrated good consistency in establishing UA RIs. Males had higher RIs than females, and RIs showed a non-linear correlation with age.

Day-to-day affect fluctuations in adults with childhood trauma history: a two-week ecological momentary assessment study

BACKGROUND

Childhood trauma (CT) may increase vulnerability to psychopathology through affective dysregulation (greater variability, autocorrelation, and instability of emotional symptoms). However, CT associations with dynamic affect fluctuations while considering differences in mean affect levels across CT status have been understudied.

METHODS

346 adults (age = 49.25 ± 12.55, 67.0% female) from the Netherlands Study of Depression and Anxiety participated in ecological momentary assessment. Positive and negative affect (PA, NA) were measured five times per day for two weeks by electronic diaries. Retrospectively-reported CT included emotional neglect and emotional/physical/sexual abuse. Linear regressions determined associations between CT and affect fluctuations, controlling for age, sex, education, and mean affect levels.

RESULTS

Compared to those without CT, individuals with CT reported significantly lower mean PA levels (Cohen's d = -0.620) and higher mean NA levels (d = 0.556) throughout the two weeks. CT was linked to significantly greater PA variability (d = 0.336), NA variability (d = 0.353), and NA autocorrelation (d = 0.308), with strongest effects for individuals reporting higher CT scores. However, these effects were entirely explained by differences in mean affect levels between the CT groups. Findings suggested consistency of results in adults with and without lifetime depressive/anxiety disorders and across CT types, with sexual abuse showing the smallest effects.

CONCLUSIONS

Individuals with CT show greater affective dysregulation during the two-week monitoring of emotional symptoms, likely due to their consistently lower PA and higher NA levels. It is essential to consider mean affect level when interpreting the impact of CT on affect dynamics.

Exploring the association of parity and its interaction with history of preterm delivery on gestational duration

PURPOSE

Preterm delivery is a major cause of child mortality. While the relationship between parity and preterm delivery is known, its association with gestational duration and variability remains underexplored. Differences in variance may suggest interaction with other well-established risk factors.

METHODS

With 1.1 million spontaneous deliveries (1990-2012) from the Swedish Medical Birth Register, we assessed while accounting for potential confounders the effects of parity on the mean and variance of gestational duration, and its possible interactions with history of preterm delivery. Pedigrees allowed to account for nonobserved, shared confounders using linear mixed models.

RESULTS

Parity has a modest association with mean gestational duration, but a large effect on its variance. For example, the first pregnancy had the shortest mean gestational duration, 0.29 days shorter (95% CI: -0.33, -0.25) than the second, and the largest variance (σ2 = 135 days2). Accounting for shared unobserved confounders highlighted a group effect bias, likely linked to the mothers' total number of offspring. Parity interacts with other risk factors, including previous preterm delivery where the magnitude of its effect increases with parity (up to 4.6 days effect difference).

CONCLUSIONS

Nonshared factors across a mother's pregnancies highlight parity's importance to gain insight into the mechanisms governing the timing of delivery.

Gender differences in the distribution of children's physical activity: evidence from nine countries

BACKGROUND

Physical activity in childhood is thought to influences health and development. Previous studies have found that boys are typically more active than girls, yet the focus has largely been on differences in average levels or proportions above a threshold rather than the full distribution of activity across all intensities. We thus examined differences in the distribution of physical activity between girls and boys in a multi-national sample of children.

METHODS

We used the harmonised International Children Accelerometry Database (ICAD), including waist-worn accelerometry data from 15,461 individuals (Boys: 48.3%) from 9 countries. Employing Generalised Additive Models of Location, Shape, and Scale (GAMLSS) we investigated gender differences in the distribution of individuals, including comparisons of variability (SD) and average physical activity levels (mean and median) and skewness. We conducted this analysis for each activity intensity (Sedentary, Light, and Moderate-to-Vigorous (MVPA)) and a summary measure (counts per minute (CPM)).

RESULTS

Sizable gender differences in the distribution of activity were found for moderate to vigorous activity and counts per minute, with boys having higher average levels (38% higher mean volumes of MVPA, 20% higher CPM), yet substantially more between-person variability (30% higher standard deviation (SD) for MVPA, 17% higher SD for CPM); boys' distributions were less positively skewed than girls. Conversely, there was little to no difference between girls and boys in the distribution of sedentary or light-intensity activity.

CONCLUSIONS

Inequality in activity between girls and boys was driven by MVPA. The higher mean volumes of MVPA in boys occurred alongside greater variability. This suggests a need to consider the underlying distribution of activity in future research; for example, interventions which target gender inequality in MVPA may inadvertently lead to increased inequality within girls.

Developing Fractional Exhaled Nitric Oxide Predicted and Upper Limit of Normal Values for a Disadvantaged Population

BACKGROUND

Fractional exhaled nitric oxide (Feno), used as a biomarker, is influenced by several factors including ethnicity. Normative data are essential for interpretation, and currently single cutoff values are used in children and adults.

RESEARCH QUESTION

Accounting for factors that influence Feno, (1) what are appropriate predicted and upper limit of normal (ULN) Feno values in an underserved population (First Nations Australians), (2) how do these values compare with age-based interpretive guidelines, and (3) what factors influence Feno and what is the size of the effect?

STUDY DESIGN AND METHODS

Feno data of First Nations Australians (age < 16 years, n = 862; age ≥ 16 years, n = 348) were obtained. Medical history using participant questionnaires and medical records were used to define healthy participants. Flexible regression using spline functions, as used by the Global Lung Function Initiative, were used to generate predicted and ULN values.

RESULTS

Look-up tables for predicted and ULN values using age (4-76 years) and height (100-200 cm) were generated and are supplied with a calculator for clinician use. In healthy First Nations children (age < 18 years), ULN values ranged between 25 and 60 parts per billion (ppb) when considering only biologically plausible age and height combinations. For healthy adults, ULN values ranged between 39 and 88 ppb. Neither the current Feno interpretation guidelines, nor the currently recommended cutoff of 50 ppb for First Nations children 16 years of age or younger were appropriate for use in this cohort. Our modelling revealed that predicted and ULN values of healthy participants varied nonlinearly with age and height.

INTERPRETATION

Because single pediatric, adult, or all-age Feno cutoff values used by current interpretive guidelines to define abnormality fail to account for factors that modify Feno values, we propose predicted and ULN values for First Nations Australians 4 to 76 years of age. Creating age- and height-adjusted predicted and ULN values could be considered for other ethnicities.

Polygenic and socioeconomic risk for high body mass index: 69 years of follow-up across life

Genetic influences on body mass index (BMI) appear to markedly differ across life, yet existing research is equivocal and limited by a paucity of life course data. We thus used a birth cohort study to investigate differences in association and explained variance in polygenic risk for high BMI across infancy to old age (2-69 years). A secondary aim was to investigate how the association between BMI and a key purported environmental determinant (childhood socioeconomic position) differed across life, and whether this operated independently and/or multiplicatively of genetic influences. Data were from up to 2677 participants in the MRC National Survey of Health and Development, with measured BMI at 12 timepoints from 2-69 years. We used multiple polygenic indices from GWAS of adult and childhood BMI, and investigated their associations with BMI at each age. For polygenic liability to higher adult BMI, the trajectories of effect size (β) and explained variance (R2) diverged: explained variance peaked in early adulthood and plateaued thereafter, while absolute effect sizes increased throughout adulthood. For polygenic liability to higher childhood BMI, explained variance was largest in adolescence and early adulthood; effect sizes were marginally smaller in absolute terms from adolescence to adulthood. All polygenic indices were related to higher variation in BMI; quantile regression analyses showed that effect sizes were sizably larger at the upper end of the BMI distribution. Socioeconomic and polygenic risk for higher BMI across life appear to operate additively; we found little evidence of interaction. Our findings highlight the likely independent influences of polygenic and socioeconomic factors on BMI across life. Despite sizable associations, the BMI variance explained by each plateaued or declined across adulthood while BMI variance itself increased. This is suggestive of the increasing importance of chance ('non-shared') environmental influences on BMI across life.