Response to letter to editor: ‘Comment on Arch et al., Trials. 2016;17:517’

Abstract In October 2015 we published the paper ‘Measurement of HbA1c in multicentre diabetes trials – should blood samples be tested locally or sent to a central laboratory: an agreement analysis’. Chatterjee and Pradhan have submitted a letter to the editor asking critical questions regarding the methods we used. We offer this letter in response. Trial registration Eudract No. 2010-023792-25. Registered on 4 November 2010. ISRCTN No. ISRCTN29255275. Registered on 12 November 2010


Main text
We thank Chatterjee and Pradhan for their letter regarding our paper in Trials. 2016; 17-517. We agree with their sentiment that local HbA1c measurement cannot be implemented at the expense of clinically unacceptable disparities between centralised and localised measurements despite its greater cost efficiency. We hope that the following provides the additional information that will aid their assessment of our results.

Time-lag
In their letter, Chatterjee and Pradhan draw attention to Fig. 3 of our paper, and notice that within centres the distribution of differences is centred on 0 (suggesting that there are no centre-specific systematic biases present). We would argue that this does not imply that the same relationship would necessarily be true for time-lag. Figure 3 displayed results by site; however, within a site the time lag may vary. We have produced boxplots to show the distribution of differences by time-lag and a scatterplot as requested. This demonstrates an absence of a linear relationship between time-lag and discrepancy. They also indicate that in practice time-lag can be an important factor for high glucose values. As part of the underlying assumptions of the Bland-Altman method we investigated heteroscedasticity (see the Verification of assumptions section within the 'Results' section of our paper), i.e. we did not observe any increase in discrepancy with higher glycosylated haemoglobin (HbA1c).
Of the 590 measurements analysed for agreement, for 79 (13.4%) the date of measurement at the central laboratory was not recorded. These 79 are indicated with 'M' on

HbA1c measurement methodology
We specify in our paper that in almost all cases, both local and central, HbA1c was measured via immunoassay using portable machines. Local measurements were normally taken at outpatient clinicsbut the technical method of measurement employed was not recorded. At the time of analysis, we contacted all sites to establish what the local methodology was. Table 1 gives details of what the sites' responses were.
We can give further detail here to say that the same portable machine was used at Alder Hey for outpatient clinics as at the central laboratory (based at Alder Hey). At this centre therefore, the methodology was identical.
Differences were still incurred despite using the same machine with a short time-lag and removing the courier and post-transfer issues (see Fig. 3).
Whether central laboratory results should be used in preference to local results is an issue that needs to be considered at the design stage of any study. We hope that the information presented will enable greater clarity  in decisions made. However, any decision needs to be born against the size of the effect that is to be detected and the potential size of discrepancies. This study demonstrates that despite quality control placed on local machines such discrepancies do occur. It should also be emphasised that this study took place in the UK and the climate and transport conditions elsewhere may determine whether local measurements are preferable.
Acknowledgements None to add.
Funding SCIPI was funded by the Health Technology Assessment (HTA) programme of the National Institute for Health Research (NIHR). Project reference: 08/14/39.

Availability of data and materials
The data for these analyses are available on request from the lead author.

Authors' contributions
All authors read and approved the final manuscript. BA carried out the analysis and wrote the statistical methods, results and much of the discussion section, and drafted the manuscript. JB wrote part of the background section and reviewed the manuscript. AM helped with data issues and reviewed the manuscript. JWG reviewed the manuscript. PN contributed the biochemistry methodology and reviewed the manuscript. CG raised the original question and led the statistical team contributing to the statistical analysis plan, analyses, and manuscript.
Authors' information BA is a statistician at the CTRC 2 ; JB is a consultant endocrinologist, and the SCIPI 1 CI; AM is the SCIPI 1 trial statistician at the CTRC 2 ; JWG is a professor in paediatric endocrinology at the University of Cardiff, and a co-investigator and PI for SCIPI 1 ; PN is a consultant biochemist at Alder Hey Children's Hospital; CG is deputy-HoD of Biostatistics at the University of Liverpool and deputy director of the CTRC 2 . 1 SCIPI: Randomised controlled trial of continuous subcutaneous insulin infusion compared to multiple daily injection regimens in children and young people at diagnosis of type I diabetes mellitus.

Competing interests
The authors declare that they have no competing interests.

Consent for publication
Not applicable.