Discussion
This large-scale retrospective study with a total of 124 651 patients with T2DM was selected for predicting the optimal values of major laboratory measurements with the best long-term clinical outcomes. The four risk factors (A1C, LDL-C, SBP, and DBP) as a combination associated with lowest predicted risks of clinical outcome were identified from regression models. The estimated optimal laboratory results were further analyzed for subgroups with various age, race, and BMI levels.
Most guidelines have actually moved toward ‘individualized goals’. The guidelines were vague in guidance to picking a particular number of A1C, BP, and LDL-C for each individual based on some kind of decision support system, which we think is needed. This is particularly true of A1c goals where ADA, American Association of Clinical Endocrinology (AACE), and American College of Physicians had different general goals but emphasized individualization.2 17 18 The goal setting is perhaps a bit easier in very high-risk patients such as the AACE ‘extreme risk’ group which may comprise individuals with diabetes following an event, or BP goals in those with nephropathy.17 However, it is harder in people with early diabetes, no complications, and treated in primary care, such as those included in our study. Our study attempted to report what we have observed as the best outcomes, as the optimal values among the subgroup population who achieved the levels seen may be useful as a guide (without being prescriptive) for such subgroups. When it comes to individualization, there is a still long way to go.
A1C control and its optimal value
In our findings, the optimal A1C was associated with lowest risk of microvascular/macrovascular complications. There is a U-shape relationship between A1C and microvascular/macrovascular complication. Explicit evidence has been found to support that lowering A1C by proper treatment can reduce the complication rates (online supplemental appendix E). However, studies rarely talked about how low the A1C should be. In this study, for lowering the risk of microvascular and macrovascular complications, A1C between 6.5% and 7.0% was optimal for the general population with T2DM. In addition, there is no one-for-all A1C target. Also, one more crucial finding is that too tight glycemic control (<5.8%) may be harmful for patients’ long-term clinical outcomes.
The microvascular complication was reduced significantly by intensive glycemic control in the ADVANCE and STENO-2 studies.15 19 ADVANCE study demonstrated the effect of tight A1C control on microvascular complication reduction with the A1C achievement of lowering from 7.5% to 6.5%, consistent with part of our results. The STENO-2 showed only <20% patients with intensive glycemic control reached the goal of A1C<6.5%. It implied that too stringent A1C level may not have a strong correlation with better clinical outcome of vascular complications. Part of our findings were inconsistent with epidemiological analyses of the DCCT and UKPDS studies.1 20 The relationship between A1C and microvascular complications was curvilinear in the epidemiological studies. Lowering A1C from 7% to 6% was associated with further reduction in the risk of microvascular complications, and the absolute risk reductions became much smaller. Also, in our findings, lowering A1C even after it reached 6.8% was inversely associated with higher risk of microvascular complications. Furthermore, the UKPDS study found that intensive glycemic control contributed to lower microvascular risk, mostly reduced retinopathy. As the UKPDS study aimed at lowering fasting blood glucose, A1C of the group with intensive treatment (7%) was lower than the control group (7.9%).
While A1C is the dominant determinant in microvascular complications, glycemic control remains important in macrovascular complications. Buse et al have demonstrated that HbA1c lowering explains most of the reduction in events in the LEADER trial.21 The ACCORD trial results led to confusion in some respects. Intensive glycemic control did reduce macrovascular events to a moderate degree, but this were overshadowed by the increase in mortality. The very low goals in ACCORD with aggressive medication titration, beyond what would have been done in clinical practice, may have contributed to the increase in mortality.10 11
In the subgroup analysis, black patients had higher optimal A1c level than white patients for achieving lowest risk of microvascular/macrovascular complication. Our findings were consistent with previous studies that higher A1C level has been found in African Americans than in white patients.22–24 The race differences in optimal A1C levels for controlling vascular complications between black and white patients have significant clinical implications on diabetes management.25 26 Our study also demonstrated that stringent A1C control is less appropriate for black patients with T2DM than white patients for the consideration of lowering risk of microvascular and macrovascular complications.
LDL-C control and its optimal value
This study identified the optimal LDL-C values for the veteran population with T2DM, and optimal LDL-C values were slightly higher than the commonly used goal (<100 mg/dL) and much higher than the stringent goal (<70 mg/dL). It implied that the risk of vascular complication might increase if patients achieved the old LDL-C target. White patients had higher optimal LDL-C value than black patients for achieving the lowest risk of microvascular/macrovascular complications.
The ADA guideline removed the LDL-C goal since 2015 and statin is recommended for all patients aged >40 years at different intensities.27 The shift in blood cholesterol management followed the changing in the ACC/AHA blood cholesterol guideline, which mentioned that statin treatment can be decided by risk evaluation instead of LDL-C level.28 Diabetes is considered as a Congenital Heart Disease (CHD) equivalent for lipid management. Therefore, patients with T2DM were widely recommended with lipid control treatment, without evaluating the level of LDL-C.
There was limited evidence from RCTs on cholesterol-lowering effects. In the ACCORD lipid trial, LDL-C had no significant difference between treatment and control groups, and no significant CVD benefits were found either.29 In STENO-2, the risks of cardiovascular disease caused by microvascular and macrovascular complications were all reduced by multifactorial intensive intervention.30 31 However, the isolated effect of LDL-C lowering was unclear in the STENO-2. The ADDITION-Europe was another randomized trial with intensive multifactorial therapy but found no significant effect on clinical outcomes.32 33 Based on our findings and literatures, widely used statin (or other lipid-lowering agents) without careful examination of LDL-C is potentially harmful to patients and may increase the risk of long-term clinical outcomes. These inconsistences may be due to the fact that no large microvascular outcome trial has been done with microvascular events as the primary outcome, despite preliminary observations showing a benefit on nephropathy and retinopathy. In addition, researchers have not figured out all the answers yet and continued to grapple with the pathophysiology as to how diabetes and high cholesterol are related. One study found that blood sugar, insulin, and cholesterol all interact with each other in the body, and are affected by each other.34 However, our study was not suggesting these LDL-C levels as targets/goals. What we have observed is the best outcomes in people who achieved these levels and may be useful as a guide (without being prescriptive) for such people.
Previous studies showed African Americans had lower LDL-C test rate and lower proportion of achieving LDL-C goal, but no significant difference in LDL-C level across races has been found,35 36 while our study showed that white patients had higher optimal LDL-C value than black patients for achieving the lowest risk of macrovascular/microvascular complications.
BP control and its optimal value
Optimal value was only detected for SBP in the models fitting for macrovascular complication. Lower SBP was correlated to lower risk of microvascular complication. To minimize the risk of macrovascular complication, optimal SBP was found at 131 mmHg. Lower SBP may increase the risk of macrovascular complication in our study. BP of 143/82 mmHg was estimated with lowest risk of vascular events. However, the population used in this study was not patients with T2DM. A retrospective study found that an optimal SBP of 128 mmHg was associated with best outcome of diabetic nephropathy.37 Considering the risk of macrovascular complication, we found that higher SBP might be the best SBP value for the general population with T2DM.
Compared with the targets in guidelines, the optimal SBP values provided more valuable information. SBP <140 mmHg is recommended for the general diabetes population, while the lower target of <130 mmHg is recommended for healthier patients or who can tolerate. In subgroup analysis, to achieve lower risk of macrovascular complication, patients older than 60 years have higher optimal SBP than younger patients. The optimal SBP was much higher in white than in black patients. Also, the patients with normal weight had lower SBP for risk reduction than the overweight patients. Therefore, SBP target should be adjusted with respect to age group, race, and body weight. Using the SBP target of <140 mmHg may be not be suitable for obese patients. Younger black patients with normal weight can be recommended with SBP at around 120 mmHg. However, patients who are older, white, and/or with obesity should have less intensive SBP control plan.
Almost all the relationships between DBP and clinical outcomes were negatively associated in the general T2DM population. Higher DBP was associated with lower risk. In the univariate analysis, the risk of vascular complications was monotonic decreased with growth of DBP until it reached around 85 mmHg. When DBP was higher than 85 mmHg, the risk slightly increased when the DBP increased. Our findings were consistent with some other studies. The SPRINT trial found that patients with low DBP showed a significantly higher risk of cardiovascular events and nephrology outcomes.38 39 The EURODIAB Prospective Complications Study concluded that diastolic blood pressure less than or equal to 83 was an important predictor for progression to proliferative diabetic retinopathy.40 The diastolic J-shape phenomenon was still in debate.41–43 However, the diastolic J-shape phenomenon was observed in this study that either too low DBP or too high DBP is harmful.
Limitations
The study has some limitations. There is a lack of information about diabetes duration in our data. To minimize this problem, the patients with a history of microvascular and macrovascular complication at baseline period were excluded. This exclusion can reduce severity of hyperglycemia and complications, both of which are associated with DM duration. Due to the nature of the VA population, more than 90% of patients are male in our sample. Thus, the results should not be generalized to both genders. The optimal blood glucose, blood pressure, and lipid control levels may vary between genders, but unfortunately it cannot be assessed in this study. Although risk prediction is not the primary objective in this study, predicted risk was used for comparison and determination the relative optimal value of diabetes management by applying the splines on the predicted risks. The numbers of knots and degree of these splines may still have estimation errors from the true optimal values.
Finally, this study is based on data that were collected before the results of recent cardiovascular and renal outcome trials were known and subsequent changes in guidelines. However, those trials compared newly developed drugs with placebo, and none of them had optimization of risk factor goals in either the drug or placebo groups, with somewhat better control of glucose and BP in the drug group over a prolonged period, and the trials were done mainly in a population outside the USA with less impressive results in subgroups in this country.44 Indeed, the analyses have suggested that risk factor differences may explain the benefits of the drugs.21 45 A comparative effectiveness study between these drugs and optimized goals, and even further subgroup analyses among different races (white vs black patients) and biomarker levels, as in our data is needed. It is noteworthy that at least one analysis has demonstrated less of a benefit of these drugs in outcomes in subgroups particularly in the USA.44 However, such subgroup analyses may not have enough power.