Predicting the Need for Insulin Treatment: A Risk-Based Approach to the Management of Women with Gestational Diabetes Mellitus

Round 1

Reviewer 1 Report

In this paper, another attempt was made to "predict" insulin treatment in GDM women.  The authors end up wit pretty much the same old variables that have been around for 50 years.

- The authors used their Astraia database (line 66) which contains ultrasound data.  One would hope that the latter would have been included, but this is not the case.  What we need is better predictions of fat and muscle tissue in fetuses, particularly the latter.  Hyperinsulinemia does not only promote adipose tissue (line 32) but also muscle - hence the shoulder dystocia and the often faulty birth weight predictions made with the traditional parameters.  If the authors would include new ultrasound parameters, then finally we would be able to move forward.

- The existing analysis ends up with the same risk factors we know for ages.  The authors even discard their own analysis: lines 176-8 "Parity and family history ... after adjustment the significance level shifted.  However, they were retained in the model".  Why do statistical modeling if you reject the result??  The final AUC (Figure 2) is a very modest 0.75, with no clear threshold, a simple bulging line.

- The Danish guidelines apparently use less strict cut-off points to start insulin (6.0 before and 8.0 mmol/l after 90 min).  Usually, 5.3 and 7.2 mmol/l are recommended for these time-points.  This may also have an impact on the decision to start insulin.

- Line 105: small for gestational age was Z-score <1,3, which is probably incorrect.

- The Discussion is too long in view of the newly presented data.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

I have question regarding allocating score points to each predictor, does each predictor has the same score point (1)?

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

This is a well presented and comprehensive study with a good discussion. The investigators have used a large well characterised data set of women with GDM to come up with a prediction model for the requirement of insulin treatment in women with GDM.

Four of the seven variables used in the eventual point scoring system are evident at first presentation for antenatal care and the other three after the HbA1c is measured and the OGTT is carried out.

The 7 point score has a high negative predictive value of 88% so as the authors this is useful in practical application of the model

The four variables evident at the outset are somewhat different from those used in the studied cohort to select women for early screening at 10-20 weeks gestation. This could be discussed.

A related issue is how and whether the model is applicable in in the early screening process/group and how soon insulin is commenced - again for discussion.

In table 1 check if you mean HbA1c at diagnosis

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

See previous comments.