Control of glycemia and blood pressure in British adults with diabetes mellitus and subsequent therapy choices: a comparison across health states

Choosing Wisely (https://​www.​choosingwisely.​org), the “Do not do” recommendations from NICE (https://​www.​nice.​org.​uk), and various guideline bodies have raised the issue of whether some individuals with type 2 diabetes mellitus may be over-treated in light of observational studies suggesting that the relationship between HbA1C levels and poor outcomes is U-shaped [1, 2]. Certainly, beyond slowing of diabetic retinopathy progression, the benefits of intensive glucose control are uncertain for patients with type 2 diabetes, and are offset by increased risks of hypoglycemic events [and even all-cause mortality in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial], especially in the elderly or those with substantial comorbidities [3‐5]. Although diabetes guidelines recommend individualization of glycemic control with HbA1c goals of less than 7.0% in healthy individuals (and < 7.5% in healthy patients older than 65 years) with less stringent targets in patients with comorbidities or frailty, several recent analyses from the United States demonstrated that many adults with type 2 diabetes had HbA1C ≤ 7% with little difference in glycemic control across health states [6‐8]. Moreover, only a minority of patients who were potentially over-treated subsequently had their glycemic therapy deintensified [9‐11].

In a similar vein, current hypertension guidelines [12, 13] advise individualization of antihypertensive therapy and caution against overtreatment in patients with type 2 diabetes since the ACCORD trial did not demonstrate any reduction in major cardiovascular events with intensive control of systolic blood pressure (SBP) but instead an increase in serious adverse events [14]. A recent systematic review of 49 antihypertensive trials in diabetes also demonstrated that benefit from treatment was only present in those patients with SBP > 140 mmHg [15]. Despite this trial evidence and studies suggesting risk with lower SBP, especially in older patients with comorbidities [16‐18], a recent report from the US Veterans Health Administration demonstrated that half of patients with type 2 diabetes had SBP less than 130 mmHg and over one-third had SBP less than 120 mmHg [9].

The purpose of this study is to examine risk factor control (glucose and SBP) in a large cohort of patients with diabetes cared for by UK primary care physicians, to examine whether control varied by health status, and to explore whether patterns of treatment deintensification varied by risk factor level and/or health status.

Of 406,649 individuals with diabetes, 297,589 had a HbA1C drawn and 292,703 had a SBP measured at least 6 months after their diabetes diagnosis; the 292,170 with both HbA1C and SBP in the Thin dataset after at least 6 months of diabetes formed the sample for this study (Fig. 1). Mean age was 61.7 years, median time to the index measurements was 354 days (IQR 254–701) after diabetes diagnosis, and the median number of primary care physician visits in the year prior to the index measurements was 14—ranging from 13 in otherwise fit individuals with diabetes up to 39 in those with moderate or severe frailty (Table 1). Of note, the median time between the index HbA1C and index SBP measurements was 12 days.

Most patients in the study sample were otherwise fit (270,068, 92.4%), 21,448 (7.3%) met the definition for mild frailty, 642 (0.2%) moderate frailty, and 12 were severely frail (we lumped moderate and severe frailty together in subsequent analyses). Comorbidities and health resource use were substantially higher as degree of frailty increased (Table 1). Diabetes complications (most commonly retinopathy) were not uncommon, even in the relatively healthy subgroups: 5.2% of otherwise fit patients and 10.3% of those classified as having mild frailty had a diabetes complication noted by the time of the index measurements.

We found that glycemic and SBP control was similar for otherwise fit adults with diabetes as for frail patients, and that while nearly one-third of all diabetic individuals in this cohort of primary care-managed NHS patients had treated HbA1C < 6.5% or SBP < 130 mmHg, nearly half had HbA1C > 7.5% or SBP > 140 mmHg. This is in contrast to US Medicare data suggesting that overtreatment is more common than undertreatment in elderly Americans [23]. We also found that, although deintensification rates in this UK cohort are nearly double the 14–19% reported in US-based studies [8‐10, 23], therapy was reduced in less than half of patients with low HbA1C or SBP levels, even if they were frail, despite frequent primary care follow-up (median 14 visits per year). On the other hand, it should be noted that even in otherwise fit individuals with diabetes, the proportions with HbA1C or SBP above recommended targets also differed little when re-measured 6 months later. This echoes an earlier study from the UK Clinical Practice Research Datalink which found that for the vast majority of patients with a new diagnosis of diabetes the HbA1C levels changed little after the first 6 months (followup was out to 24 months) and only 40% of patients with HbA1C levels above target had any therapy intensification in the subsequent year [24]. A recent analysis of 833 elderly individuals with type 2 diabetes mellitus in Canadian primary care practices reported findings similar to ours: approximately half of healthy older patients had HbA1C and SBP levels above target range, nearly half of frail patients exhibited potential overtreatment, and medication adjustments in either situation were uncommon [25]. Recent reports from China and a registry from 410 sites in 32 other countries also documented that less than half of patients with type 2 diabetes have their HbA1C and SBP controlled within recommended target ranges, although they did not distinguish between potential over- vs. under-treatment [26, 27].

Thus, our data highlight the importance of personalizing targets and multiple potential targets for quality improvement in primary care management of type 2 diabetes. For example, although the emphasis in diabetes professional education programs and guidelines is usually on glucose lowering options [4, 28], far more patients in our cohort exhibited uncontrolled hypertension than poor glycemic control—consistent with earlier studies in Canada, the United States, and the Netherlands [22, 28‐31]. This pattern is concerning since blood pressure is the strongest driver of cardiovascular outcomes in diabetic individuals (with an attributable risk nearly three fold greater than glycemic levels in the Framingham study) [32]. The benefits of blood pressure lowering (at least to < 140 mmHg) [15, 17] exceed those of glucose lowering [3‐5], and antihypertensive therapy is the most cost-effective of the cardiovascular prevention therapies in type 2 diabetes [33]. Moreover, it should be acknowledged that in all health states uncontrolled hypertension was far more common than over-controlled blood pressure. Thus, our study highlights the need for increased attention to the role of blood pressure control in improving outcomes for individuals with diabetes.

On the other side of the quality equation, although clinical performance measures have until recently almost exclusively focused on under-treatment and failure to meet targets [28, 34], there is increasing recognition of the potential harms that can arise from over-treatment. Intensive lowering of glucose and blood pressure have both been shown to be potentially harmful in individuals with diabetes [1‐3, 11, 15]. Although we did not have access to complete data on emergency room visits or hospitalizations for our cohort, other studies have demonstrated that hypoglycemia is one of the most common presenting diagnoses in emergency departments for elderly individuals and results in more frequent and more severe hospitalizations than does hyperglycemia [35, 36]. Moreover, achieving intensive glycemic or SBP control requires polypharmacy which puts patients at increased risk for drug–drug interactions, poorer medication adherence, and reduced quality of life [16]. While efforts through Choosing Wisely, NICE, and the Veterans Affairs Hypoglycemia Safety Initiative to improve physician awareness of this issue are a welcome addition to the quality improvement landscape in diabetes, educational interventions alone are inadequate to overcome clinical inertia [37]. In fact, preliminary evidence suggests that clinical inertia is an even greater barrier against deintensifying therapy than introducing novel therapies [38, 39].

Although our study has several strengths due to the availability of detailed clinical data in a large population-based sample of patients with new diagnoses of diabetes, there are some limitations to our data. First, to the extent that the primary care clinical records may have under-captured some comorbidities (particularly likely for conditions like dementia, chronic pain, or depression), we may have underestimated the proportion of individuals with multiple comorbidities and thus overstated the proportion who were otherwise fit. Second, we may have underestimated rates of deintensification as we don’t know if insulin doses had been decreased or if patients were told to split pills after receiving their prescriptions. Third, although some might see it as a weakness that we used prescription records rather than pharmacy dispensation records, we would argue that this is a strength since it permits us to capture physician intent without the data being confounded by primary non-adherence (patients not filling their prescription). Fourth, while there are various scales for evaluating patient frailty, cumulative deficit models (like the eFI we used) have been shown to better predict functional and mortality outcomes than some of the other scales more commonly quoted in the literature [40]. Fifth, although some antihypertensive therapies such as ACE inhibitors or angiotensin receptor blockers are recommended in diabetic individuals with left ventricular dysfunction or significant proteinuria regardless of their SBP, only a very small minority of our cohort had these conditions. Finally, we did not have data on other factors that could influence clinical decision making such as socioeconomic status, specialist involvement in patient care, or patient preferences. Without the fuller picture this data would provide, we elected not to perform multivariate analyses to try to define whether particular comorbidities or patient factors were more likely to be associated with out-of-range HbA1C or SBP levels than other factors to avoid potentially spurious findings.

In conclusion, although there were statistically significant (but clinically minor) differences in glycemic and hypertensive control across health states, medication changes were uncommon regardless of their health status. Nearly one-third of all diabetic individuals in this cohort had treated HbA1C < 6.5% or SBP < 130 mmHg with little difference in proportions when re-measured 6 months later. On the other hand, nearly half of this cohort exhibited poor control of glucose (HbA1C > 7.5%) or SBP (> 140 mmHg), with again little change in those proportions 6 months later despite frequent primary care visits. Our study thus highlights the need for personalizing treatment targets in adults with type 2 diabetes and implementing diabetes quality improvement strategies [41] that are flexible enough to focus on both the potential over-treatment of multimorbid patients as well as potential under-treatment of healthier patients and to consider both blood pressure as well as glycemic targets. To that end, it is worth noting that a recent Monte Carlo-based Markov Model decision analysis confirmed that care to achieve individualized treatment targets was cost-saving and associated with a small improvement in quality of life [42].