Statis tical Analyse s
For patient description, median and IQRs were provided for
continuous variables and percentage for binary variables. For
group comparison (patients included in the study vs patients
lost to follow-up) Wilcoxon rank test (continuous variables),
and c2 test (binary variables) were applied. P values were
adjusted for multiple comparisons according to Holm (Bonferroni step-down). For intra-individual comparisons of
metabolic control during the prepubertal, pubertal, and postpubertal/young adult period, respective differences were
analyzed by Wilcoxon signed-rank test.
Spearman correlation with Fisher z-transformation to
calculate 95% CI was used to reflect tracking of metabolic
control, relating median HbA1c-values in each patient during the prepubertal period with the respective values during
puberty, as well as in adulthood. The c2test was used to
analyze the relationship between tertiles of metabolic control
during prepuberty and adulthood, as well as the achievement
of adequate control as recommended in current guidelines.
To analyze the contribution of prepubertal metabolic control on adult HbA1c, a mixed hierarchic regression model
was used (dependent variable: HbA1c in adulthood, independent variable: HbA1c during prepuberty). Center was entered
as a random effect in the model (covariance structure: Cholesky), the intraclass correlation (between center variation)
was 21%. Estimation was based on residual pseudolikelihood, denominator degrees of freedom were calculated
according to Kenward-Roger, and iterations were optimized
according to Newton-Raphson. In addition to this simple
model, a fully adjusted model including sex, migration background, and year of manifestation, diabetes duration, insulin
therapy, and adult BMI together with type of treatment center was implemented.
Logistic regression models, with identical covariates as in
the fully adjusted model, were used to calculate odds for
good metabolic control in adulthood based on recommended HbA1c levels adopting either American Diabetes Association (ADA) or International Society for Pediatric and
Adolescent Diabetes guidelines respectively.
For all analyses, a 2-sided P value of <.05 was considered
statistically significant. The statistical software package SAS
9.3 was used for analysis (SAS Institute Inc, Cary, North
Carolina).
Results
By March 2013, the DPV database included 15 162 patients
with type 1 diabetes manifestation younger than 11 years of
age, born prior to 1993, with complete baseline documentation. Among these, 1146 patients were followed continuously
from prepuberty through puberty to adulthood. The large
number of patients lost to follow-up is due to change of providers of diabetes care, especially during transition from pediatric centers to adult internal medicine. We consequently
compared the study group (n = 1146) with 14 016 patients
Tab le I. Patient demographics
Study cohort,
complete follow-up
until the age of
20 y (n = 1146)
Patient group,
loss-of
follow-up
(n = 14 016) P
Male (%) (SD) 49.4 ( 0.5) 49.0 ( 0.5) n.s.
Mean age at onset (y) (SD) 6.9 ( 2.8) 6.9 ( 2.8) n.s.
Mean age (y) (SD) 21.6 ( 2.8) 17.8 ( 2.8) <.0001
Migration background (%) 8.0 ( 0.3) 6.5 ( 0.3) n.s.
Pediatric center (%) 77.3 ( 0.4) 83.4 ( 0.4) <.0001
BMI (kg/m2) (SD) 24.5 (3.7) 23.1 ( 4.1) <.0001
Mean HbA1c (%) (SD) 8.3 (1.8) 8.7 ( 2.0) <.0001
n.s., not significant.
Comparison of patient demographics of the study cohort with complete documentation until the
age of $20 years and the cohort with incomplete documentation and loss of follow-up.
Vol. 165, No. 5 November 2014
95
การแปล กรุณารอสักครู่..
![](//thimg.ilovetranslation.com/pic/loading_3.gif?v=b9814dd30c1d7c59_8619)