Epidemiology and determinants of non-diabetic hyperglycaemia and its conversion to type 2 diabetes mellitus, 2000–2015: cohort population study using UK electronic health records

Objectives To study the characteristics of UK individuals identified with non-diabetic hyperglycaemia (NDH) and their conversion rates to type 2 diabetes mellitus (T2DM) from 2000 to 2015, using the Clinical Practice Research Datalink. Design Cohort study. Settings UK primary Care Practices. Participants Electronic health records identified 14 272 participants with NDH, from 2000 to 2015. Primary and secondary outcome measures Baseline characteristics and conversion trends from NDH to T2DM were explored. Cox proportional hazards models evaluated predictors of conversion. Results Crude conversion was 4% within 6 months of NDH diagnosis, 7% annually, 13% within 2 years, 17% within 3 years and 23% within 5 years. However, 1-year conversion fell from 8% in 2000 to 4% in 2014. Individuals aged 45–54 were at the highest risk of developing T2DM (HR 1.20, 95% CI 1.15 to 1.25— compared with those aged 18–44), and the risk reduced with older age. A body mass index (BMI) above 30 kg/m2 was strongly associated with conversion (HR 2.02, 95% CI 1.92 to 2.13—compared with those with a normal BMI). Depression (HR 1.10, 95% CI 1.07 to 1.13), smoking (HR 1.07, 95% CI 1.03 to 1.11—compared with non-smokers) or residing in the most deprived areas (HR 1.17, 95% CI 1.11 to 1.24—compared with residents of the most affluent areas) was modestly associated with conversion. Conclusion Although the rate of conversion from NDH to T2DM fell between 2010 and 2015, this is likely due to changes over time in the cut-off points for defining NDH, and more people of lower diabetes risk being diagnosed with NDH over time. People aged 45–54, smokers, depressed, with high BMI and more deprived are at increased risk of conversion to T2DM.


GENERAL COMMENTS
Please maintain consistency in the abbreviation of type 2 diabetes mellitus, choose either T2DM or T2D or T2diabetes (all 3 have been used in the manuscript) I would recommend that the authors comment on which Read code was most commonly used to detect NDH and whether NDH defined from either a fasting glucose or an abnormal glucose tolerance test was more predictive of converting to T2DM page 2, line 27 and page 7, line 50 -please add units after BMI value of 30 page 4, line 4 -please review the concept that pancreatic dysfunction causes insulin resistance in patients with T2DM. Pancreatic dysfunction and insulin resistance are 2 of at least 8 pathophysiological processes increasing risk of T2DM. There are multiple potential causes of insulin resistance.
page 4, line 26 -is the the range 5.5-6.9 mmol/L correct? Most international societies define abnormal glucose as a fasting level of 5.6 mmol/L or greater. A glucose value of 5.5 mmol/L would generally be considered normal page 6, line 8 -why was alcohol intake, previous pancreatitis and current medication not included as relevant for NDH and conversion of NDH to T2DM? If this was not available then it should be added as a limitation of the study page 6, line 55 -why were those that converted to T2DM after 5 years excluded from the analysis?
Page 17 -please add BMI units to Table 3 Page 19 -please add units to Table 5 page 23, line 10 -was ethics approval given for this study? If it wasn't obtained then can the authors please provide an explanation. At the end of the study under "Ethical Approval" it states "Not applicable". For all studies it is applicable and should be obtained, especially if the study it is to be published.

nih.gov/pmc/articles/PMC4769302/). Although this change occurred in 2004, quality was already high from 2000 onwards, in anticipation for the scheme and other smaller-scale frameworks. The only potential issue with the QOF was the non-distinction in coding between Type-1 and Type-2, until explicitly requested in 2006. This may have led to us missing a few cases that exited the database before 2006 (at which point it time they would have to be given a specific code to be included in the QOF returns), if they had type-2 diabetes but were only given a generic diabetes code. In our experience this is very rare, however and it would not affect our finding that conversion rates for NDH have dropped over time. In terms of false-positive rates, in the past we have experimented with defining cohorts of chronic conditions differently (i.e. medications and two or more relevant Read codes), but we found that resulting changes were negligible. As previously mentioned, the quality of recording is very high and people associated with a Read code for T2DM, have the conditionthere is no provisional coding and GPs are encouraged to add to records only if certain since they know retracting such a diagnosis is very complicated. If someone is suspected of having the condition they will be not be given a Read code, but information will be added in notes (or with a "suspected diabetes" code). Remission is possible of course, although rare, but it is not relevant for this study (where T2DM is the outcome of interest in a time to event analysis). Regarding NDH coding, the situation is more complicated because of the absence of financial incentives through the QOF, hence practice variability is greater. In addition, the definition of NDH has changed over time, as we explain in the paper, making it difficult to operationalise through biological measurements, which are very often missing. As for T2DM, however, we would expect many cases of false positives, as the reviewers suggests, because of the coding practices previously explained. We would expect many cases to be missed, something that is well known and acknowledged in the paper, but something that should have no significant bearing on our findings and their implications (unless there is something fundamentally different about the "missed" NDH cases, which we do not think is the case).
It seems almost all factors identified to be related to the conversion from NDH to T2DM are already known. It would be better if authors can list factors already known and those which are novel in a table. Otherwise, it is hard to determine what is the contribution of the work. I would recommend that the authors comment on which Read code was most commonly used to detect NDH and whether NDH defined from either a fasting glucose or an abnormal glucose tolerance test was more predictive of converting to T2DM

Response: Thank you for this suggestion. This is something we considered as well but we decided against reporting for a couple of reasons. First, it is secondary to the aims of the paper and the paper is already quite complicated and long. Second, Read code usage changes over time and is often computer system specific (so may not be generalisable to England/the UK), hence this is as simple a task as it may originally seem, while its usefulness is perhaps questionable. In terms of how NDH is defined, we did conduct secondary analyses for 5 different cohorts, according to the different definitions of the NDH in the literature exploring the Fasting Plasma glucose tolerance tests (FPG) as well as abnormal glucose tolerance test ( HBA1C categories). The categories of NDA definitions we explored were 1) American Diabetes Association (ADA) FPG (5.6-6.9)mmol/mol 2) ADA HBA1C (39-46)mmol/mol (This was also explored further by categorising HBA1C levels to quartiles) 3) WHO FPG (6.1-6.9)mmol/L 4) International Expert Committee [IEC] HBA1C (42-46)mmol/mol 5) Diabetes UK {FBG(5.5-6.9) orHBA1C (42-47)} We explored whether conversion to T2DM varied across these definitions. Although some variability was observed it did not explain the drop in conversion rates over time, which one of the key findings of the study. Thus, we concluded that this analysis would not add much to the paper (considering the significant expansion needed to explain the cohorts and the methods), and was not included.
Below are the plotted patterns for these cohorts

Response: This has been reviewed and a sentence has been added on Page 4, line 7-9. "There are other key pathophysiological processes which increase the risk of T2DM, which involve organs including pancreas, liver, skeletal muscle, kidneys, brain, small intestine and adipose tissue 3 ".
page 4, line 26 -is the the range 5.5-6.9 mmol/L correct? Most international societies define abnormal glucose as a fasting level of 5.6 mmol/L or greater. A glucose value of 5.5 mmol/L would generally be considered normal

patient (and the relevant volumes), which is a tremendous amount of work, with no clear link to conversion as far as we know; secondly, and more importantly, including treatment in our model would probably introduce unmeasured confounding, with treatments being associated to conversion when the underlying conditions and the health of the patient are the driving causes. Regarding pancreatitis, this was an omission and we thank again the reviewer for highlighting this.
Our risk prediction model did not attempt to include and reaffirm all known drivers of diabetes, but we primarily aimed to examine the role of socio-economic drivers and lifestyle factors, along with "overall health" (using the Charlson Comorbidity Index as a proxy), and depression which has been found to be particularly important (and potentially actionable) in the context of T2DM. We have now expanded the limitations section to explain our modelling choices. Page 9 , lines 9-18 "Our risk prediction model did not attempt to include and reaffirm all known drivers of diabetes, but we primarily aimed to examine the role of socio-economic drivers and lifestyle factors, along with depression (potentially actionable and important comorbidity for T2DM 19 ), and a proxy for "overall health". Alcohol intake was not included in the model, since the quality of recording such information in UK primary care is rather poor 20 . We also decided not to use medication for two reasons: first, we would need to capture and organise everything to a patient (and the relevant volumes), which is a tremendous amount of work, with no clear link to conversion as far as we know; secondly, and more importantly, including treatment in our model would probably introduce unmeasured confounding, with treatments being associated to conversion when the underlying conditions and the health of the patient are the driving causes. " page 6, line 55 -why were those that converted to T2DM after 5 years excluded from the analysis? Response: As the number of years taken to convert from NDH to T2DM ranged up to 22 years where the numbers converted at the extreme years were quite low, for the consistency in our analysis we decided to restrict our analysis to those who converted from NDH to T2DM within a 5 year period.
Page 17 -please add BMI units to Table 3 Page 19 -please add units to Table 5 Response: Amended, thanks page 23, line 10 -was ethics approval given for this study? If it wasn't obtained then can the authors please provide an explanation. At the end of the study under "Ethical Approval" it states "Not applicable". For all studies it is applicable and should be obtained, especially if the study it is to be published.