Authors’ reply to Yudkin and LipskaBMJ 2012; 344 doi: http://dx.doi.org/10.1136/bmj.e710 (Published 31 January 2012) Cite this as: BMJ 2012;344:e710
- Bianca Hemmingsen, PhD student1,
- Søren S Lund, physician2,
- Christian Gluud, chief physician and head of department1,
- Allan Vaag, professor3,
- Thomas Almdal, chief physician and head of department2,
- Christina Hemmingsen, research assistant1,
- Jørn Wetterslev, chief physician1
- 1Copenhagen Trial Unit, Center for Clinical Intervention Research, Rigshospitalet, Copenhagen University Hospital, Copenhagen, Denmark
- 2Steno Diabetes Center, Gentofte, Denmark
- 3Department of Endocrinology, Rigshospitalet, Copenhagen University Hospital
Yudkin and Lipska note that we did not present the numbers needed to treat for the assessed outcomes in our systematic review.1 2 According to the Cochrane Collaboration, a relative measure—for example, relative risk or odds ratio—is preferred for reporting intervention effects because it remains more stable across different risk groups.3 Therefore, we reported the relative risk ratio for all outcomes.2
Furthermore, we reported the absolute risk difference for the composite microvascular outcome and retinopathy.2 We would have reported the numbers needed to treat on the basis of these association measures if the assessed intervention had shown significant effects. We did not perform this additional analysis because the uncertainty that intensive glycaemic control really reduces these outcomes may render it impossible to calculate 95% confidence intervals on the numbers needed to treat.
As the trial sequential analysis makes evident, the trial sequential monitoring boundaries were not surpassed for the composite microvascular outcome (adjusted 95% confidence interval for the relative risk: 0.76 to 1.01) and retinopathy (0.55 to 1.15), still leaving the possibility for a harmful or neutral effect on these outcomes.2 To calculate numbers needed to treat would therefore not be meaningful.
Yudkin and Lipska highlight the notion that seemingly insufficient evidence exists to show that glucose lowering has a strong beneficial impact on microvascular outcomes in type 2 diabetes. Indeed, one of the key findings from our meta-analysis, using trial sequential analysis, was the apparent lack of firm evidence of benefit when targeting intensive versus standard glucose control on a composite microvascular outcome in patients with type 2 diabetes in usual care settings.2 Retinal photocoagulation and cataract extraction (which we did not assess) are both important to the patient. However, the assessments of these outcomes are highly prone to bias in trials without blinding.4
Cite this as: BMJ 2012;344:e710
Competing interests: SSL, AV, and TA have reported equity in Novo Nordisk A/S; SSL and AV have received fees from Novo Nordisk A/S for speaking; TA is employed at Steno Diabetes Center, Gentofte, Denmark; AV and SSL were employed at Steno Diabetes Center at the time the review was written. Steno Diabetes Center is an academic institution owned by Novo Nordisk A/S. CH has been employed at Novo Nordisk since completion of the data extraction. SSL is employed by Boehringer Ingelheim, Ingelheim, Germany.