After publication of this work , it came to our attention that the laboratory we use changed the method of analysis for insulin and C-peptide during our data collection. Before March 2007 the method used for insulin was competitive radioimmunoassay (Linco Research Inc, St Charles, MO, USA) and for C-peptide competitive luminoimmunoassay (Immulite 2000, Diagnostic Products Corporation, CA, USA). After March 2007 noncompetitive immunofluorometric assays (DELFIA kit form Wallac OY, Turku, Finland) were used.
We have therefore recalculated the insulin and C-peptide values analyzed after March 2007, and repeated the statistical analyses involving insulin, C-peptide and HOMA-IR.
Corrections in the Abstract
In Results the fifth sentence should have read:
“Even a very small reduction in BMI z-score (group 3) was associated with significantly lower total cholesterol, LDL cholesterol and total/HDL cholesterol ratio.”
After recalculation the significant reduction in insulin in the group with a very small reduction in BMI z-score disappeared. None of the groups had significant improvements in insulin concentration (the improvements in group 1 and 2 were of borderline significance).
The seventh sentence should have read:
“An increase in BMI z-score was associated with a worsening of HOMA-IR, insulin, C-peptide and total/HDL cholesterol ratio.”
The group with an increase in BMI z-score had a worsening in HOMA-IR and insulin in addition to C-peptide and total/HDL cholesterol ratio as previously reported.
Corrections in the Main manuscript