S. van Buuren

Multiple imputation of missing blood pressure covariates in survival analysis.

This paper studies a non-response problem in survival analysis where the occurrence of missing data in the risk factor is related to mortality. In a study to determine the influence of blood pressure on survival in the very old (85+ years), blood pressure measurements are missing in about 12.5 per cent of the sample. The available data suggest that the process that created the missing data depends jointly on survival and the unknown blood pressure, thereby distorting the relation of interest. Multiple imputation is used to impute missing blood pressure and then analyse the data under a variety of non-response models. One special modelling problem is treated in detail; the construction of a predictive model for drawing imputations if the number of variables is large. Risk estimates for these data appear robust to even large departures from the simplest non-response model, and are similar to those derived under deletion of the incomplete records.

Continuing positive secular growth change in the Netherlands 1955-1997.

Since 1858, an increase of mean stature has been observed in the Netherlands, reflecting the improving nutritional, hygienic, and health status of the population. In this study, stature, weight, and pubertal development of Dutch youth, derived from four consecutive nationwide cross-sectional growth studies during the past 42 y, are compared to assess the size and rate of the secular growth change. Data on length, height, weight, head circumference, sexual maturation, and demographics of 14 500 boys and girls of Dutch origin in the age range 0–20 y were collected in 1996 and 1997. Growth references for height and weight were constructed with a method that summarizes the distribution by three smooth curves representing skewness (L curve), the median (M curve), and coefficient of variation (S curve). The relationship between height and demographic variables was assessed by multivariate analysis. Reference curves for menarche and secondary sex characteristics were estimated by a generalized additive model using a logit transformation. A positive secular growth change has been present in the past 42 y for children, adolescents, and young adults of Dutch origin, although at a slower rate in the last 17 y. Height differences according to region, educational level of child and parents, and family size have remained. In girls, median age at menarche has decreased by 6 mo during the past four decades to 13.15 y. Environmental conditions have been favorable for many decades in the Netherlands, and the positive secular change in height has not yet come to a halt, in contrast to Scandinavian countries. Main contributors to the increase in height may be improved nutrition, child health, and hygiene, and a reduction of family size.

Body index measurements in 1996–7 compared with 1980

OBJECTIVES To compare the distribution of body mass index (BMI) in a national representative study in The Netherlands in 1996–7 with that from a study in 1980. METHODS Cross sectional data on height, weight, and demographics of 14u2009500 boys and girls of Dutch origin, aged 0–21 years, were collected from 1996 to 1997. BMI references were derived using the LMS method. The 90th, 50th, and 10th BMI centiles of the 1980 study were used as baseline. Association of demographic variables with BMI-SDS was assessed by ANOVA. RESULTS BMI age reference charts were constructed. From 3 years of age onwards 14–22% of the children exceeded the 90th centile of 1980, 52–60% the 50th centile, and 92–95% the 10th centile. BMI was related to region, educational level of parents (negatively) and family size (negatively). The −0.9, +1.1, and +2.3 SD lines in 1996–7 corresponded to the adult cut off points of 20, 25, and 30u2009kg/m2 recommended by the World Health Organisation/European childhood obesity group. CONCLUSION BMI age references have increased in the past 17 years. Therefore, strategies to prevent obesity in childhood should be a priority in child public health.

Nationwide age references for sitting height, leg length, and sitting height/height ratio, and their diagnostic value for disproportionate growth disorders

Aims: To obtain age references for sitting height (SH), leg length (LL), and SH/H ratio in the Netherlands; to evaluate how SH standard deviation score (SDS), LL SDS, SH/H SDS, and SH/LL SDS are related to height SDS; and to study the usefulness of height corrected SH/H cut-off lines to detect Marfan syndrome and hypochondroplasia. Methods: Cross-sectional data on height and sitting height were collected from 14 500 children of Dutch origin in the age range 0–21 years. Reference SD charts were constructed by the LMS method. Correlations were analysed in three age groups. SH/H data from patients with Marfan syndrome and genetically confirmed hypochondroplasia were compared with height corrected SH/H references. Results: A positive association was observed between H SDS, SH SDS, and LL SDS in all age groups. There was a negative correlation between SH/H SDS and height SDS. In short children with a height SDS <−2 SDS, a cut-off limit of +2.5 SD leads to a more acceptable percentage of false positive results. In exceptionally tall children, a cut-off limit of −2.2 SDS can be used. Alternatively, a nomogram of SH/H SDS versus H SDS can be helpful. The sensitivity of the height corrected cut-off lines for hypochondroplasia was 80% and for Marfan syndrome only 30%. Conclusions: In exceptionally short or tall children, the dependency of the SH/H ratio (SDS) on height SDS has to be taken into consideration in the evaluation of body proportions. The sensitivity of the cut-off lines for hypochondroplasia is fair.

The Terneuzen Birth Cohort: BMI changes between 2 and 6 years correlate strongest with adult overweight

Background Complications of overweight amplify with age, and irreversible damage already exists in young persons. Identifying the most sensitive age interval(s) for adult overweight is relevant for primary prevention. The aim of the study was to assess the relative contribution of body mass index (BMI) changes between 0 and 18 years to adult overweight, and to identify the earliest critical growth period. Methods and Findings Data from 762 subjects in the Terneuzen Birth Cohort with an average of 21 growth measurements per subject from birth until 18 years were used. The main outcome measure was the BMI standard deviation score (SDS) at young adulthood. For each subject BMI SDS was fitted by a piecewise linear model at eight different ages and correlated to adult BMI SDS. The age intervals in between are considered critical according to three criteria, tested by respectively Students t-tests, multiple linear regression analyses and Pearsons correlation tests. In the age intervals 4 months(m) -1 year(y), 2–6 y, 6–10 y and 10–18 y the BMI SDS change differs between adults with and without overweight (P≤0.001). The age intervals 2–6 y and 10–18 y also meet the second criterion, implying that the BMI change during this period has a predictive value for adult BMI SDS in addition to BMI SDS at the end of the period. The largest rise in correlation between estimated BMI SDS and measured adult BMI SDS occurs during the period 2–6 y (from 0.36 to 0.63), which results in a high sensitivity (0.6) and specificity (0.8) by the age of 6 y. Conclusions/Significance The age interval from 2 y to 6 y is the earliest and most critical growth period for adult overweight. Therefore, primary prevention of adult overweight seems most likely to be successful if targeted at this specific age interval. By identifying those with an upwards centile crossing between 2 and 6 years, the development towards adult overweight might be reversed.

Recursive partitioning for missing data imputation in the presence of interaction effects

Standard approaches to implement multiple imputation do not automatically incorporate nonlinear relations like interaction effects. This leads to biased parameter estimates when interactions are present in a dataset. With the aim of providing an imputation method which preserves interactions in the data automatically, the use of recursive partitioning as imputation method is examined. Three recursive partitioning techniques are implemented in the multiple imputation by chained equations framework. It is investigated, using simulated data, whether recursive partitioning creates appropriate variability between imputations and unbiased parameter estimates with appropriate confidence intervals. It is concluded that, when interaction effects are present in a dataset, substantial gains are possible by using recursive partitioning for imputation compared to standard applications. In addition, it is shown that the potential of recursive partitioning imputation approaches depends on the relevance of a possible interaction effect, the correlation structure of the data, and the type of possible interaction effect present in the data.

Towards evidence based referral criteria for growth monitoring

Aims: To evaluate the performance of growth monitoring in detecting diseases. Turner’s syndrome (TS) is taken as the target disease. Methods: Case-control simulation study. Three archetypal screening rules are applied to longitudinal growth data comparing a group with TS versus a reference group from birth to the age of 10 years. Main outcome measures were sensitivity, specificity, and median referral age. Results: Clear differences in performance of the rules were found. The best rule takes parental height into account. Combining rules could improve diagnostic accuracy. Conclusion: Growth monitoring is useful to screen for TS. A combined rule that takes absolute height SDS, parental height, and deflection in height velocity into account is the best way to do this. Similar research is needed for other diseases, populations, and ages, and the results should be synthesised into evidence based referral criteria.

Body size and growth in 0- to 4-year-old children and the relation to body size in primary school age

Excess weight in early life is believed to increase susceptibility to obesity, and in support of such theory, excess weight and fast weight gain in early childhood have been related to overweight later in life. The aim of this study was to review the literature on body size and growth in 0‐ to 4‐year‐old children and the association with body size at age 5–13 years. In total, 43 observational studies on body size and/or growth were included, of which 24 studies had been published in 2005 or later. Twenty‐one studies considered body size at baseline, and 31 studies considered growth which all included assessment of weight gain. Eight (38%) studies on body size, and 15 (48%) on weight gain were evaluated as high‐quality studies. Our results support conclusions in previous reviews of a positive association between body size and weight gain in early childhood, and subsequent body size. Body size at 5–6 months of age and later and weight gain at 0–2 years of age were consistently positively associated with high subsequent body size. Results in this review were mainly based on studies from developed Western countries, but seven studies from developing countries showed similar results to those from developed countries.

Estimating regional centile curves from mixed data sources and countries

Regional or national growth distributions can provide vital information on the health status of populations. In most resource poor countries, however, the required anthropometric data from purpose-designed growth surveys are not readily available. We propose a practical method for estimating regional (multi-country) age-conditional weight distributions based on existing survey data from different countries. We developed a two-step method by which one is able to model data with widely different age ranges and sample sizes. The method produces references both at the country level and at the regional (multi-country) level. The first step models country-specific centile curves by Box-Cox t and Box-Cox power exponential distributions implemented in generalized additive model for location, scale and shape through a common model. Individual countries may vary in location and spread. The second step defines the regional reference from a finite mixture of the country distributions, weighted by population size. To demonstrate the method we fitted the weight-for-age distribution of 12 countries in South East Asia and the Western Pacific, based on 273 270 observations. We modeled both the raw body weight and the corresponding Z score, and obtained a good fit between the final models and the original data for both solutions. We briefly discuss an application of the generated regional references to obtain appropriate, region specific, age-based dosing regimens of drugs used in the tropics. The method is an affordable and efficient strategy to estimate regional growth distributions where the standard costly alternatives are not an option.

Trends in body mass index distribution and prevalence of thinness, overweight and obesity in two cohorts of Surinamese South Asian children in The Netherlands

Objectives Asians have a smaller muscle mass and a larger fat mass at the same body mass index (BMI) than most other ethnic groups. Due to a resulting higher cardiometabolic risk, the BMI cut-offs for overweight and obesity were lowered for adults. For Asian children universal criteria apply. The objectives of this study were to determine the normal BMI distribution and assess the BMI class distribution in a reference cohort of affluent South Asian children born before the obesity epidemic and to assess the influence of the obesity epidemic on the distributions. Methods Historical cohort study with 4350 measurements of height and weight of two cohorts (born 1974–1976 and 1991–1993) of Surinamese South Asian children living in The Netherlands, analysed with WHO Child Growth References and International Obesity Task Force (IOTF) BMI cut-offs. Results The reference cohort 1974–1976 was significantly lighter (BMI Z-score=−0.63; 95% CI −0.69 to −0.58) and more variable (SD=1.19) than WHO reference. Total thinness prevalence was exceptionally high, both in cohort 1974–1976 (WHO 38.3%; IOTF 36.4%) and 1991–1993 (WHO 23.6%; IOTF 23.9%). Overweight and obesity prevalences were low in the reference cohort (WHO respectively 6.0% and 2.1%; IOTF 5.3%, 0.9%), but much higher in cohort 1991–1993 (WHO 13.6%, 9.1%; IOTF 11.7%, 6.0%). Conclusions The low mean BMI Z-score and high prevalence of thinness are likely expressions of the characteristic body composition of South Asians. Universal BMI cut-offs should be applied carefully in South Asian populations as thinness prevalence is likely to be overestimated and obesity underestimated. The development of ethnic specific cut-offs is recommended.