Discussion
We investigated the associations between attainment of QOF clinical indicators, completion of NDA care processes, and non-traumatic lower limb amputations among those with type 2 diabetes. We observed that minor or major, as well as major-only, amputation rates were 26%–51% and 3%–51%, respectively, lower among those who met the HbA1c indicator and 14%–47% lower among those who met the cholesterol indicator for our primary outcome. However, we did not find evidence of association between blood pressure indicator attainment and amputation. Sensitivity analyses for both outcomes were indicative that the incremental gains associated with HbA1c and cholesterol indicator attainment may be lower when greater numbers of QOF indicators have been met overall, in keeping with the idea that relative risk reduction is greater where baseline absolute risk is higher. Substantially lower amputation rates across both outcomes were also observed among those who completed a greater number of NDA care processes, indicative of benefits of comprehensive care. Although formal trial evidence regarding the association between HbA1c control and amputations is relatively sparse in view of amputations being a relatively infrequent outcome,21 22 our HbA1c indicator findings correspond with results of a number of observational studies.23–26 These suggest a relatively linear relationship between HbA1c and amputations, at least within the limits of glycemic control considered optimal in terms of other important diabetes complications.23 24 26 Previous observational and randomized studies are similarly in agreement with our findings relating to the cholesterol indicator.27 28
Regarding blood pressure control, although this plays an essential role in the management of type 2 diabetes as part of comprehensive secondary prevention, after adjustment for other study covariates, we did not find an association between blood pressure (at the QOF indicator threshold) and lower limb amputation. This corresponds with the findings of some,29 but not all,30 previous studies. Residual confounding may be relevant to the apparent lack of association. Hypertension is a well-established risk factor for peripheral arterial disease.31 However, the interplay between peripheral arterial disease, blood pressure, and foot perfusion is not clearly understood, and it may be the case that relatively low blood pressure is unfavorable in the context of compromised vascular supply (ie, those with existing vascular disease). There may also be an element of reverse causation, as autonomic neuropathy can lead to relative hypotension, and neuropathy is part of the pathological process that results in amputation among those with diabetes.32
Overall, our findings generally support use of the current QOF and NDA indicators in reducing the risk of amputations. The reduction in amputation rates seen over the last decade4 may be due to the emphasis placed on secondary prevention through QOF over the last 17 years, given the associations between QOF indicator achievement, care process delivery, and amputation incidence that we demonstrate in our study. More refined analyses among those in whom relative hypotension does not reflect underlying disease would aid interpretation of the blood pressure findings. And indeed, the QOF blood pressure indicator was amended in 2019–2020, such that it no longer applies to those with moderate-severe frailty, who are at potentially greatest risk of blood pressure effects of comorbidities.33 Regarding the potentially more limited benefit of meeting QOF indicators where relatively comprehensive care is already achieved, this makes theoretical sense and is supported by the trend in point estimates observed in our sensitivity analyses. However, these analyses will also have been relatively underpowered (reflected in the larger CIs), hence we would not conclude that control to indicator levels is not beneficial where other indicators are met. Even if this were the case, we have previously observed that indicator attainment in this context is associated with lower occurrence of mortality and hospital admissions (manuscripts currently under review).
In terms of mechanisms to further type 2 diabetes prevention activities to help mitigate against ongoing increases in lower limb amputation events, it will be important that this is balanced against the disutility associated with comprehensive diabetes care (particularly as this is potentially not dissimilar to that associated with the conditions that predispose to amputation, and amputation is a relatively rare outcome).3 Strategies to reduce the disutility associated with diabetes care would be helpful. These should be guided by those with diabetes, but as examples, adaptations could include facilitating safer and less financially costly access to dietary and physical activity options, providing easy-to-negotiate follow-up mechanisms, and inclusive care with strong continuity.
The strengths of this study include the sample size and that the sample is likely to be reasonably representative of the population of interest. The CPRD employs routine quality assurance checks,34 35 and CPRD data have been shown to offer clinical predictive value and similar incidence to other sources of UK health data.34 35 They have previously been successfully used for validation of amputation risk predictive modelling among those with type 2 diabetes,24 and the dataset has a linkage to non-primary care data including hospital admissions on a prospective basis. The dataset enabled adjustment for many potential confounders. QOF incentivizes use of standard recording methods, and HES data are widely used and considered to be of high quality. HES data are subject to regular national audits, and a systematic review has evaluated its accuracy as high for both diagnoses and procedures.36
There are, nonetheless, important limitations to our study. An element of miscoding and misclassification will have occurred as a result of inconsistency in general practitioner coding practices. That we defined our exposure using 2010–2011 data, and did not assess variation over time, will also have led to some heterogeneity in clinical status as related to exposure classification and potential dilution of effect. For the major-only secondary outcome, some individuals had prior minor amputations during the study period. Our study population excludes those with amputation history at baseline, who may experience different incidence rates. Additionally, the events and rates reported correspond to first amputations (whether overall (primary analysis) or major-only (secondary analysis)), where censoring occurs at the time of first amputation, rather than overall amputation rates during the study period for this cohort. It is also possible that the exposures could, to an extent, reflect level of engagement with care more generally, as much as their specific physical correlates. Our interpretation of the analyses of NDA care processes is limited by their examination as a ‘count’ measure, rather than individual processes. It would be anticipated that some processes, such as foot examination, would be particularly relevant to amputation prevention.37–39 Finally, mortality-driven censoring could be more prevalent among those with higher amputation risks,40 with different distributions between exposed and unexposed groups. Thus, the Cox proportional hazards assumption of non-informative censoring may not apply to all individuals.
Both comprehensive diabetes care, and HbA1c and cholesterol control to the levels investigated, are associated with considerably lower risks of lower limb amputation. This information may help inform individual diabetes management and medical decision-making, as well as diabetes-related resource allocation. It supports widening access to comprehensive diabetes care. Inclusive discussions around reducing care disutility may help promote both equity in access and care uptake.