Combined intensive blood pressure and glycemic control does not produce an additive benefit on microvascular outcomes in type 2 diabetic patients

Abstract

A reduction of either blood pressure or glycemia decreases some microvascular complications of type 2 diabetes, and we studied here their combined effects. In total, 4733 older adults with established type 2 diabetes and hypertension were randomly assigned to intensive (systolic blood pressure less than 120mmHg) or standard (systolic blood pressure less than 140mmHg) blood pressure control, and separately to intensive (HbA1c less than 0.060) or standard (HbA1c 0.070–0.079) glycemic control. Prespecified microvascular outcomes were a composite of renal failure and retinopathy and nine single outcomes. Proportional hazard regression models were used without correction for type I error due to multiple tests. During a mean follow-up of 4.7 years, the primary outcome occurred in 11.4% of intensive and 10.9% of standard blood pressure patients (hazard ratio 1.08), and in 11.1% of intensive and 11.2% of standard glycemia control patients. Intensive blood pressure control only reduced the incidence of microalbuminuria (hazard ratio 0.84), and intensive glycemic control reduced the incidence of macroalbuminuria and a few other microvascular outcomes. There was no interaction between blood pressure and glycemic control, and neither treatment prevented renal failure. Thus, in older patients with established type 2 diabetes and hypertension, intensive blood pressure control improved only 1 of 10 prespecified microvascular outcomes. None of the outcomes were significantly reduced by simultaneous intensive treatment of glycemia and blood pressure, signifying the lack of an additional beneficial effect from combined treatment.

INTRODUCTION

Both glucose^1–5 and blood pressure (BP) lowering^6–8 reduce the risk of several microvascular complications of type 2 diabetes mellitus (T2DM), and epidemiological analyses suggest that the combined effects of BP and glucose control may be greater than the benefit of either intervention alone.^9,10 Two large clinical trials provide additional insight on the combined effects of BP and glycemic control. Intensive glycemic control significantly reduced microvascular complications in the UK Prospective Diabetes Study (UKPDS).^1,2 A subset of UKPDS participants with uncontrolled hypertension were randomized to intensive vs. standard BP control. Participants in the intensive arm achieved a BP of 144/82mmHg and experienced a 34% reduction in progression of retinopathy, a 29% reduction in onset of microalbuminuria (450 mg/l), and a 37% reduction in a composite measure of microvascular end points; however, there was no reduction in development of macroalbuminuria (4300 mg/l) or renal failure (RF).⁸ These investigators also reported that the incidence of ‘any diabetes-related end point’ was lowest in those in both the intensive BP and intensive glycemia arms of the trial.¹¹

The ADVANCE trial used a double blind factorial design to assess the effect of BP and glycemic control on cardiovascular and microvascular outcomes in patients with T2DM.⁶ The impact on microvascular complications of intensive BP control alone,⁶ intensive glycemic control alone,³ and their combined effect has been reported.¹² Rates for all renal events and for new onset of microalbuminuria were reduced by intensive BP control, intensive glycemic control, and both, with participants randomized to both intensive arms having the lowest hazard ratio (HR). However, the effect of each treatment was not significantly altered by the other. In aggregate, these data suggest a possible additive benefit of BP and glycemic control on prevention or progression of certain microvascular complications, particularly those related to microalbuminuria and macroalbuminuria. The ACCORD-BP trial¹³ reported that participants randomized to a systolic BP (SBP) <120mmHg (achieved mean SBP 119.3mmHg) vs.<140mmHg (achieved mean 133.5mmHg) had no significant effect on the composite cardiovascular end point or its components, and did not reduce the progression of retinopathy; however, a detailed analysis of microvascular outcomes of this trial has not been reported. Intensive BP control did lead to a 16% reduction of incident microalbuminuria, but not macroalbuminuria. Analysis of microvascular outcomes in the main ACCORD glycemia trial showed that intensive glycemic control did not affect a prespecified composite measure of advanced microvascular complications, including RF and retinopathy requiring photocoagulation and/or vitrectomy.⁴ However, intensive glycemic control significantly improved two of five secondary measures of nephropathy (microalbuminuria and macroalbuminuria), whereas one measure deteriorated with intensive glucose control.⁴ In addition, the incidence of one measure of diabetic eye disease and three measures of neuropathy were reduced with intensive glycemic control. In a substudy of ACCORD, intensive glucose control slowed progression of retinopathy, but intensive BP control resulted in no beneficial effect on retinopathy, with a nonsignificant worsening of moderate visual loss with intensive BP control.¹⁴

The present analysis of the ACCORD-BP trial has two aims as follows: (1) to report the results of intensive vs. standard BP control on 10 predefined microvascular outcomes and to examine the combined effects of intensive BP and glycemic control on these outcomes, and (2) to assess potential benefits of intensive BP and glycemic control on development and progression of microalbuminuria and macroalbuminuria, and on the development of RF.

RESULTS

Table 1 summarizes the 10 predefined ACCORD microvascular outcomes and their frequencies of assessment. Baseline characteristics of participants stratified by the four intervention groups are shown in Table 2. Changes in BP and glycemic control occurred during the trial, as previously described^4,13 (see Supplementary Table S1 online). In addition, minor but significant differences in serum triglycerides, high-density lipoprotein cholesterol, serum creatinine, and body mass index were observed between the intervention arms (Supplementary Table S1 online). The present analysis did not adjust for parameters measured during the post-randomization study period.

Table 1.

Definition of ACCORD microvascular outcomes and their frequency of assessment

Outcome category	Label	Definition	Assessment frequency
Primary composite	Primary	Development of renal failure (initiation of dialysis or ESRD, renal transplantation, or serum creatinine 4292 mmol/l) or retinal photocoagulation or vitrectomy to treat retinopathy	Every 4 months
Nephropathy	Neph-1	Development of incident microalbuminuria (defined as urine albumin/creatinine ratio X3.39mg albumin/mmol creatinine and <33.9mg albumin/mmol creatinine)	Annually
	Neph-2	Development of incident macroalbuminuria (defined as urine albumin/creatinine ratio X33.9mg albumin/mmol creatinine)	Annually
	Neph-3	Development of renal failure (defined as initiation of dialysis or ESRD, or renal transplantation, or serum creatinine 4292 mmol/l in absence of an acute reversible cause)	Every 4 months
Diabetic eye complications	Eye-1	Retinal photocoagulation or vitrectomy to treat retinopathy	Annually
	Eye-2	Eye surgery for cataract extraction	Annually
	Eye-3	Three-line decrease in visual acuity (as measured using Log MAR visual acuity chart)	Biennially
Neuropathy	Neuro-1	Score of 42.0 on the Michigan Neuropathy Screening Instrument (MNSI)	Annually
	Neuro-2	Loss of vibratory sensation (tested using 128 Hz tuning fork)	Annually
	Neuro-3	Loss of light touch (as measured by 10 g force monofilament test)	Annually

Abbreviation: ESRD, end-stage renal disease.

Table 2.

Selected continuous and categorical covariates and components of microvascular outcomes at baseline, by blood pressure and glycemia trial arm assignment

	Standard glycemia/standard BP				Standard glycemia/intensive BP				Intensive glycemia/standard BP				Intensive glycemia/intensive BP
Continuous factors	N	Mean	s.d.	Median	N	Mean	s.d.	Median	N	Mean	s.d.	Median	N	Mean	s.d.	Median
Age	1178	62.2	6.8	62	1184	62.3	7.0	62	1193	62.3	7.0	62	1178	62.1	6.6	61
Diabetes duration (years)	1169	10.8	7.7	9	1174	11.2	8.1	9	1180	11.3	8.4	10	1161	10.9	7.8	9
HbA1c (proportion)	1171	0.083	0.011	0.081	1181	0.084	0.011	0.082	1190	0.083	0.011	0.081	1178	0.084	0.011	0.081
Serum glucose (mmol/l)	1168	9.649	3.279	9.158	1175	9.772	3.184	9.324	1186	9.580	3.123	9.102	1174	9.773	3.218	9.269
Systolic BP (mm Hg)	1178	140.1	16.1	139	1184	138.6	16.2	137	1193	138.6	15.0	137	1177	139.4	16.0	139
Diastolic BP (mm Hg)	1178	76.3	10.3	76	1184	75.8	10.5	76	1193	75.7	10.2	75	1177	76.1	10.6	76
Total cholesterol (mmol/l)	1168	4.941	1.142	4.817	1175	5.042	1.173	4.895	1186	4.975	1.154	4.843	1174	5.012	1.165	4.869
LDL cholesterol (mmol/l)	1168	2.803	0.917	2.694	1172	2.899	0.978	2.745	1186	2.831	0.946	2.745	1174	2.859	0.961	2.720
HDL cholesterol (women; mmol/l)	550	1.319	0.345	1.295	552	1.336	0.348	1.295	571	1.337	0.392	1.269	569	1.320	0.344	1.295
HDL cholesterol (men; mmol/l)	618	1.090	0.324	1.036	623	1.090	0.293	1.062	615	1.087	0.318	1.036	605	1.052	0.284	1.010
Triglycerides (mmol/l)	1168	2.162	2.103	1.655	1175	2.176	2.077	1.639	1186	2.162	2.075	1.661	1174	2.225	1.943	1.684
Body mass index (kg/m2)	1177	32.1	5.3	31.6	1184	32.3	5.7	31.7	1193	32.1	5.6	31.5	1177	32.1	5.7	31.7
Weight (kg)	1178	92.0	17.5	91.6	1184	92.4	19.4	90.5	1193	91.7	17.8	90.7	1177	91.9	19.4	90.1
Serum creatinine (mmol/l)	1167	78.7	20.0	79.6	1175	78.9	20.5	79.6	1186	79.7	21.6	79.6	1175	79.5	21.3	79.6
Estimated GFR (ml/min per 1.73 m2)	1167	91.6	25.6	89.5	1175	92.5	34.4	90.3	1186	91.8	28.5	89.7	1175	90.7	25.6	89.5
Urine albumin:creatinine ratio (mg/mmol)	1164	11.34	38.04	1.681	1169	9.26	32.0	1.653	1188	11.33	46.90	1.512	1171	10.03	32.23	1.616
Number of oral anti-hyperglycemics	1178	1.3	0.9	1	1184	1.3	0.9	1	1193	1.3	0.9	1	1178	1.3	0.9	1
Number of oral anti-hypertensives	1178	1.6	1.0	2	1184	1.6	1.1	2	1193	1.5	1.0	1	1178	1.6	1.0	1
Average visual acuity score	1136	73.2	12.9	76.5	1144	73.1	12.1	75.5	1127	73.4	12.0	76	1135	72.9	11.9	75
Average Snellen fraction	1147	45.2	64.4	32	1149	43.0	46.6	32	1141	41.6	48.6	32	1142	44.5	51.7	32.5
MNSI neuropathy score	1145	2.0	1.7	2	1161	2.0	1.7	2	1166	1.9	1.7	2	1151	2.0	1.7	2

	Standard glycemia/ Standard BP			Standard glycemia/ intensive BP			Intensive glycemia/ Standard BP			Intensive glycemia/ intensive BP
Categorical factors	N	Frequency	%	N	Frequency	%	N	Frequency	%	N	Frequency	%
Positive history of clinical CVD	1178	391	33.2	1184	392	33.1	1193	398	33.4	1178	412	35.0
On oral anti-hyperglycemic meds	1178	918	77.9	1184	912	77.0	1193	905	75.9	1178	919	78.0
On insulin	1178	435	36.9	1184	458	38.7	1193	451	37.8	1178	401	34.0
On any anti-hypertensive meds	1178	1027	87.2	1184	1038	87.7	1193	1028	86.2	1178	1041	88.4
Smoking history	1178			1182			1192			1176
Never smoker		535	45.4		542	45.9		534	44.8		510	43.4
Former smoker		495	42.0		492	41.6		494	41.4		500	42.5
Current smoker		148	12.6		148	12.5		164	13.8		166	14.1
Race/ethnicity	1178			1184			1193			1178
White		701	59.5		695	58.7		666	55.8		719	61.0
Hispanic		86	7.3		79	6.7		85	7.1		80	6.8
African American		273	23.2		283	23.9		307	25.7		264	22.4
Asian		61	5.2		62	5.2		63	5.3		64	5.4
All other (including multi-racial)		57	4.8		65	5.5		72	6.0		51	4.3
On statin	1178	761	64.6	1184	774	65.4	1193	795	66.6	1178	735	62.4
On aspirin	1178	599	50.8	1184	644	54.4	1193	610	51.1	1178	620	52.6
On NSAID	1178	40	3.4	1184	46	3.9	1193	41	3.4	1178	39	3.3
Microalbuminuria (urine albumin:creatinine ratio X3.39 mg albumin/mmol creatinine mg/g)	1166	398	34.1	1173	383	32.7	1191	360	30.2	1165	364	31.2
Macroalbuminuria (urine albumin:creatinine ratio X33.9)	1166	83	7.1	1173	64	5.5	1191	69	5.8	1165	69	5.9

Abbreviations: BP, blood pressure; CVD, cardiovascular disease; GFR, glomerular filtration rate; HbA1c, hemoglobin A1c; HDL, high-density lipoprotein; LDL, low-density lipoprotein; meds, medicines; MNSI, Michigan Neuropathy Screening Instrument; NSAID, nonsteroidal antiinflammatory drug.

Results of BP treatment arm assignment are shown in (Figure 1a). Over a mean follow-up period of 4.7 years, the primary microvascular outcome occurred in 527 of 4726 participants, including 11.4% in the intensive BP arm and 10.9% in the standard BP arm (HR=1.08, 95% confidence interval (CI): 0.91–1.28). Intensive BP control reduced the development of microalbuminuria (HR=0.84, 95% CI: 0.72–0.97, P-value=0.019), but the effect on incident macroalbuminuria was not significant (HR=0.81, 95% CI: 0.63–1.03, P-value=0.087). No effects of intensive BP control were observed on any of the other microvascular outcomes (Figure 1a). Intensive vs. standard glycemic control in participants in the ACCORD-BP trial reduced three of the microvascular outcomes but not the primary composite outcome for advanced microvascular disease (Figure 1b). However, there was no interaction between intensive BP and intensive glycemic control.

Figure 1.

Effect of intensive versus standard blood pressure (BP) and glycemic control on microvascular outcomes. Forest plot depicting effects of randomized blood pressure (a) and glycemia (b) treatments on selected microvascular events with P-values for tests of two-way interactions between treatments (“Interact P-value” column).

Analyses of possible two-way interactions are provided in Supplementary Tables S2 and S3. In participants assigned to intensive BP control (Supplementary Table S2 online), intensive glycemic treatment reduced the incidence of macroalbuminuria compared with standard glycemic control (HR=0.58, P-value=0.004); however, there was no significant interaction. In participants assigned to intensive glycemic control (Supplementary Table S3 online), intensive BP control resulted in a reduction in incident microalbuminuria compared with standard BP control (HR=0.76, P-value=0.013) and resulted in a near-significant reduction in the risk of macroalbuminuria (HR=0.68, P-value= 0.052), again without a significant interaction. Potential two-way interactions between BP and glycemia treatment arm assignments and baseline history of clinical cardiovascular disease were also determined for each outcome and are summarized in Supplementary Tables S4 and S5; interactions were not significant for any of the outcomes.

Relationships between BP and glycemia interventions and development of macroalbuminuria were explored after stratifying by baseline microalbuminuria status (Table 3). In participants with no microalbuminuria at baseline, intensive BP control was not associated with a reduction in incident macroalbuminuria (HR=1.14, 95% CI: 0.66–1.96). In contrast, among participants with microalbuminuria at baseline, intensive BP treatment was associated with reduced incidence of macroalbuminuria (HR=0.71, 95% CI: 0.54–0.93). However, the interaction test for differential treatment effects across the strata of baseline microalbuminuria was not significant (interaction P-value=0.134); thus, the pooled estimate of the HR for BP treatment (HR=0.84, 95% CI: 0.63–1.03, P-value=0.09; Figure 1a and b) remained the more appropriate relative risk measure. Assignment to intensive glycemic control was associated with a reduction in incident macroalbuminuria among participants without microalbuminuria at baseline (HR=0.55, 95% CI: 0.32–0.97), and to a lesser extent among participants with microalbuminuria at baseline (HR=0.77, 95% CI: 0.59–1.01; Table 3). Again, the test for interaction between glycemic control and baseline microalbuminuria was not significant (P-value=0.2997); hence, the pooled estimate is more appropriate (HR=0.68, 95% CI: 0.53–0.87, P-value=0.002; Figure 1a and b).

Table 3.

Effect of randomized treatments on follow-up incident macroalbuminuria, stratified by baseline microalbuminuria

Baseline microalbuminuria	Intensive BP arm	Standard BP arm	HR	95% CI	P-value	Interaction P-value
No	27/1473 (1.8%)	25/1501 (1.7%)	1.14	0.66–1.96	0.6394	0.1340
Yes	90/565 (15.9%)	122/557 (21.9%)	0.71	0.54–0.93	0.0121
	Intensive glycemia arm	Standard glycemia arm
No	19/1512 (1.3%)	33/1462 (2.3%)	0.55	0.32–0.97	0.0403	0.2997
Yes	89/531 (16.8%)	123/591 (20.8%)	0.77	0.59–1.01	0.0603

Abbreviations: BP, blood pressure; CI, confidence interval; HR, hazards ratio.

Renal failure developed in 124 (2.6%) of 4726 participants in the ACCORD-BP trial (Figure 1); 17 exhibited serum creatinine 4292 mmol/l on a single blood draw, the remaining 107 initiated dialysis. Among 3182 participants without microalbuminuria at baseline, 63 (2.0%) developed RF, whereas 33 of 1216 (2.7%) with microalbuminura at baseline developed RF, and 26 of 285 participants with macroalbuminuria at baseline (9.1%) developed RF (two participants with RF did not have baseline urinalyses). Thus, approximately half (63/122) of all participants who developed RF had neither microalbuminuria nor macroalbuminuria at baseline. Figure 2 shows the relationship between baseline and follow-up albuminuria and subsequent RF among the 4232 participants who had baseline and at least one follow-up urine albumin measurement before occurrence of RF; 74 participants from this group developed RF. Among 2894 participants with no albuminuria at baseline, 36 developed RF, and 33 of the 36 never exhibited albuminuria during the trial. Among 1101 participants with microalbuminuria at baseline, 22 developed RF, and among 237 participants with macroalbuminuria at baseline 14 developed RF. Less than half (33/74 or 45%) of the participants without baseline albuminuria who developed RF did not have albuminuria during the trial, and only 19 of the 74 participants (26%) who developed RF had prior macroalbuminuria either at baseline or during the trial.

Figure 2.

Follow-up urine albuminuria and renal failure (RF, dialysis or serum creatinine 4292 lmol/l) in 4232 ACCORD-BP Trial participants with at least one follow-up urine albumin measurement before occurrence of RF. Follow-up microalbuminuria (urine albumin-to-creatinine ratio or UACRX3.39 mg albumin/mmol creatinine) and macroalbuminuria (UACRX33.9 mg albumin/mmol creatinine) was determined by urinalysis performed every 2 years. Participants with microalbuminuria or macroalbuminuria at baseline, but a maximum urine albumin-to-creatinine ratio value o3.39 mg/mmol at all subsequent assessments regress to ‘No microalbuminuria or macroalbuminuria’. Participants with macroalbuminuria at baseline but a maximum urine albumin-to-creatinine ratio value X3.39 mg/mmol and <33.9 mg/mmol at all subsequent assessments regress to ‘Microalbuminuria’. Of 124 participants who experienced RF during the course of the trial, 2 did not have baseline urinalysis performed, 6 did not have post-randomization urinalysis performed, and 42 had onset of RF before the first post-randomization urinalysis and are not included in the figure.

Table 4 shows results of a proportional hazard regression model for time to RF during the trial as a function of baseline renal function variables, BP and glycemia treatment arm assignments, and history of clinical cardiovascular disease. Among the baseline variables listed, a higher serum creatinine (HR=1.64, 95% CI: 1.21–2.23 for every 44.2 mmol/l increase; P<0.001) and macroalbuminuria (HR=4.42 vs. no albuminuria, 95% CI: 2.73–7.14, P<0.0001) were associated with increased risk of RF during the trial. Baseline microalbuminuria (vs. no albuminuria), history of cardiovascular disease, and inclusion in the intensive or standard glycemia or BP arms did not predict development of RF.

Table 4.

Proportional hazards regression model for time development of renal failure (nephropathy outcome #3; defined as initiation of dialysis or ESRD, or renal transplantation, or rise of serum creatinine >292 μmol/l in absence of an acute reversible cause) including baseline renal measures as covariates

Parameter	HR	95% Cl	P-value
Baseline serum creatinine (44 μmol/l increase)	1.64	1.21–2.23	0.0014
Baseline urine albuminuria
Microalbuminuria (UAlb:Cr ratio ≥ 3.39 mg albumin/mmol creatinine vs. no albuminuria	1.34	0.88–2.04	<0.0001a
Macroalbuminuria (UAlb:Cr ratio ≥33.9 mg albumin/mmol creatinine) vs. no albuminuria	4.42	2.73–7.14

Abbreviations: BP, blood pressure; Cl, confidence interval; ESRD, end-stage renal disease; HR, hazards ratio; Ualb:Cr, urine to albumin-to-creatinine ratio.

^aTwo degree-of-freedom test.

Model also includes terms for glycemia treatment arm assignment, BP treatment arm assignment, and baseline history of cardiovascular disease (CVD).

DISCUSSION

Although the primary aim of the main ACCORD trial was to study the effect of intensive vs. standard glycemic control on cardiovascular outcomes, ocular, renal, and neural outcomes were systematically measured as well.⁴ In the ACCORD-EYE study, development or progression of retinopathy was shown to be reduced by 33% by intensive glycemic control but not by intensive BP control.¹⁴ In addition, as described previously,⁴ in the entire ACCORD population of 10,251 participants, intensive glycemic control did not alter the occurrence of the primary composite microvascular end point, but reduced the incidence of microalbuminuria by 21% at the end of the 3.5 (mean) years of randomized treatment and by 15% after B1.5 years of further follow-up.⁴ In addition, reductions of several other microvascular end points (incident macroalbuminuria, cataract surgery, Michigan Neuropathy Screening Instrument score 42 for neuropathy, loss of ankle jerk, and loss of pressure sensation) were evident after 5 years of follow-up.⁴

Our analysis of microvascular outcomes in 4733 participants in the ACCORD-BP trial reported here adds to these prior findings.¹³ Over a mean follow-up period of 4.7 years, intensive BP treatment achieved mean SBP of 119.3mmHg in contrast to 133.5mmHg with standard treatment. As previously reported, this intervention led to no significant effect on the composite cardiovascular end point or its components,¹³ nor did it reduce the progression of retinopathy in the ACCORD-EYE study.¹⁴ In the present analysis, intensive vs. standard BP control was associated with a 16% reduction of new microalbuminuria; development of macroalbuminuria was nominally reduced by 19%, but this effect fell short of statistical significance, possibly due to the smaller number of events. The lack of significant association between intensive BP control and reduction in microvascular complications was due to small observed effects for some outcomes and a lack of statistical power for others (Supplementary Table S6 online); in particular, the study was clearly underpowered to detect development of incident macroalbuminuria and overt RF.

The lack of effect of intensive vs. standard BP control on the composite cardiovascular outcome, use of medications, and rates of adverse events in the ACCORD-BP study were reported previously.¹³ Our finding of relatively modest effects of maintaining a substantial (B14mmHg) reduction in SBP on microvascular outcomes deserves comment. One possible explanation is that the patient population studied in the ACCORD-BP trial had, on average, long duration of diabetes, and thus a substantial burden of preexisting microvascular complications. In such a population, the individuals most vulnerable to developing early complications may have already done so, and those with more advanced microvascular damage might be unresponsive to further intensification of treatment because of irreversible structural changes. Another interpretation suggests that, in this specific population, additional lowering of BP beyond the level achieved in the control arm of the trial is not beneficial. The smaller-than-expected benefit of lower SBP in ACCORD is similar to that predicted by an epidemiological analysis of the UKPDS population,¹⁵ where 10mmHg lower SBP was associated with only a 13% lower risk of pooled microvascular end points (retinopathy requiring photocoagulation, vitreous hemorrhage, fatal or non-fatal RF). The UKPDS investigators raised the possibility that the disparity between this epidemiological association and the treatment effect observed in the randomized BP study (37% reduction of microvascular end points) might have been due to favorable effects of the antihypertension agents used (mainly captopril and atenolol) in addition to their effects on lowering BP alone. In ACCORD, both the standard and intensive BP treatment strategies used multiple classes of antihypertension medications, a mean of 2.1 with standard and 3.4 with intensive treatment within 1 year of randomization.¹³ At the end of the study, 80% of participants in the standard and 90% in the intensive arm were taking an angiotensinconverting- enzyme inhibitor or an angiotensin-receptor blocker, with 56% and 80% on diuretics, and 43% and 61% on beta-blockers, respectively. Thus, the putative beneficial effects of medications derived from presumed non-BP-reduction mechanisms may have been less in ACCORD than in other trials. Further analysis is necessary to explore the potential effects of medication usage on these outcomes.

In this analysis, as in the UKPDS and ADVANCE trials,^11,12 no significant interactions were found between simultaneous intensive BP and glycemic control for microvascular end points. That is, the joint effect of intensive BP and glycemia interventions were neither synergistic nor differential in nature, and there were no outcomes for which both interventions were simultaneously significantly efficacious. For example, in the intensive glycemia arm of this study, the incidence of microalbuminuria and macroalbuminuria appeared to be lower in participants who were in the intensive compared with the standard BP treatment arm, suggesting that it might be clinically advantageous to pursue a joint intensive strategy to prevent these microvascular outcomes. However, the interaction P-value for the two interventions was not significant, indicating that this BP effect in the intensive glycemia arm was not significantly different from the BP effect in the standard glycemia arm. Moreover, the results observed in the intensive glycemia treatment arm were not statistically different from those in the standard glycemia arm. It is remarkable that two such different interventions targeting dissimilar risk factors appear to have no significant additive effect on microvascular outcomes.

Of particular interest in this analysis is the impact on renal outcomes. Albuminuria is an established risk factor and surrogate for diabetic kidney disease and RF.^16,17 Poor glycemic control has been associated with microvascular complications in T2DM,^18,19 and results of randomized controlled trials have shown that intensive control of BP and glycemia reduces albuminuria.^1,3–5 In agreement with these studies, we also observed a reduction in albuminuria with intensive BP and glycemic control. However, whether albuminuria is a frequent predictor for progressive renal disease in T2DM is less certain. In the UKPDS, only half of the participants who developed chronic kidney disease exhibited albuminuria either before or after the onset of chronic kidney disease, and less than 20% of those with chronic kidney disease exhibited albuminuria before developing chronic kidney disease. ²⁰ Similar findings have been reported by other investigators.²¹ Our present analysis shown in Figure 2 concurs and shows that 33 out of 74 (45%) of the participants who developed RF did not have microalbuminuria or macroalbuminuria at baseline or during the trial, and that only about 20% of participants who developed RF had prior macroalbumiuria. These observations suggest that microalbuminuria is not a constant marker for the development of RF in T2DM.

The lack of a beneficial effect of intensive BP or glycemic control on incident RF is consistent with findings in the larger ACCORD trial,^4,22 and is similar to results reported by the VADT and ADVANCE trials.^3,5 Potential reasons for this apparent lack of benefit include (1) small number of events, (2) a very long duration of exposure to hypertension and hyperglycemia might be necessary for the development of RF, (3) the duration and/or the intensity of treatment was inadequate, and (4) treatments were initiated late in the course of the disease when pathological changes were irreversible. In this context, it is worth noting that rates of RF were not affected by intensive glycemic control in participants with newly diagnosed T2DM who were enrolled in the UKPDS and followed up for 10.7 years.¹ The nearly equal percentage of participants who developed RF in the intensive and standard glycemic and BP treatment arms of all the above-mentioned studies suggests that the potential beneficial effect of intensive treatment may be small. It should be acknowledged that in patients with both T2DM and hypertension, the respective contribution of either condition to the development of RF is unknown. Finally, a kidney biopsy study reported that 23% of diabetic patients with 2 g of albumin excretion per day had nondiabetic causes of glomerulopathy and had no retinopathy.²³ Regardless, the multifactorial nature and the complex set of mechanisms underlying the pathogenesis of T2DM suggests that other factors, in addition to hypertension and hyperglycemia, may have important roles in the development of RF in this disease.

This study has several strengths, including the large sample size, the randomized controlled design, and analysis according to the intention-to-treat principle. The study has a number of limitations, including the fact that neither the investigators nor participants were blinded to treatment arm assignment; there were few RF events and thus low power to detect treatment effects for that outcome (although neither treatment strategy showed a potential of being efficacious); diagnosis of albuminuria was based on the albuminto-creatinine ratio measured on spot-urine samples instead of timed albumin excretion rate; and the assessment of multiple outcomes in a single analysis means that the experiment-wide type I error rate for all outcomes combined was greater than the nominal 5%. A Bonferroni-type adjustment applied to 10 outcomes would require significance at the 0.005 a-level in order to ensure an experiment-wide type I error rate of 5%. Only the effect of the glycemica intervention on development of macroalbuminuria and loss of pressure sensation met that level of significance.

In summary, the ACCORD-BP trial found that an intensive strategy achieving mean SBP below 120 mg Hg reduced new development of microalbuminuria by 16%, but had no significant effect on other microvascular end points or composites. No interactions between the glycemic and BP interventions were found. Furthermore, there were no microvascular outcomes for which both BP and glycemia treatment arm assignment were simultaneously significant; thus, neither intensive intervention was found to provide significant further reduction in risk of microvascular outcomes in the presence of the other. We also note that albuminuria was not a consistent surrogate marker for the development of RF. Taken together, we conclude that although older patients with long duration of T2DM, hypertension, and high cardiovascular risk may obtain protection from worsening albuminuria by intensification of either BP or glycemic control, additional benefit from simultaneous intensive management was not apparent.

MATERIALS AND METHODS

Design overview, setting and participants, and randomization and follow-up

Eligibility, consort diagrams, processes of recruitment, randomization, and masking for the ACCORD main trail and ACCORD-BP trial have been described previously.^4,13,22,24 Targets of the ACCORD-BP trial with 4733 participants were SBP <120mmHg vs. <140mmHg, and the targets for the glycemia trial were HbA1c <0.060 vs. 0.070–0.079.13,15 The study was approved by the Institutional Review Boards. Study investigators were masked to results of interim analyses but unmasked to individual treatment group assignment and prescribed therapies.

As part of their glycemic and BP interventions, participants received instructional materials and behavioral counseling, and were provided with all prescribed glucose-lowering and antihypertension medications. Study investigators adjusted therapy based on randomized assignment and response to therapies. Adverse effects were carefully monitored both locally and centrally to ensure participant safety.²⁵

Outcomes

This analysis is restricted to the subgroup of 4733 ACCORD participants who were enrolled in the ACCORD-BP trial. Surveillance for and measurement of prespecified microvascular outcomes have been described in detail previously.⁴ The primary composite outcome was the first occurrence of advanced kidney or eye disease as manifested by the development of RF (initiation of dialysis or end-stage renal disease, renal transplantation, or increase in serum creatinine 4292 mmol/l) or retinal photocoagulation or vitrectomy to treat retinopathy; this outcome was intended to approximate the primary microvascular outcome of the UKPDS study.¹ The 10 outcomes listed in Table 1 were selected from the original list of 14 prespecified outcomes based on glycemia treatment effect or clinical importance.⁴ Individual components of the microvascular outcomes reported here were measured or assessed using standard procedures, as reported,⁴ and are described in the study Manual of Procedures; more details can be found in the Supplementary Information online.

Statistical analysis

Participant characteristics at baseline were calculated by glycemia and BP intervention group assignments. Descriptive statistics (medians and inter-quartile ranges) of a subset of continuous factors related to treatment in each trial were calculated by glycemia and BP treatment arms; differences in these factors were assessed using the Wilcoxon signed-rank test (Supplementary Table S1 online).

Occurrence of microvascular events was determined for each predefined outcome listed in Table 1. For each participant and outcome, the observation was censored at the last surveillance time if no event was discovered. If an event was discovered, an event time was assigned using the midpoint between the time of event discovery and the most recent prior surveillance time.²⁶

The effect of treatment arm assignment on time to occurrence of the first microvascular event of each type was analyzed using proportional hazards regression models to estimate HRs and assess statistical significance. Graphical depiction of time to event was performed using product-limit plots. The primary statistical test for each outcome was taken from a proportional hazards regression model, which included BP treatment arm assignment, glycemia treatment arm assignment, and an indicator of history of clinical cardiovascular disease at baseline. The ACCORD-BP trial was designed to test the (marginal) effect of intensive vs. standard BP treatment in diabetics in a background of intensive vs. standard glycemic control; however, the factorial design allowed the testing of differential effects of combined BP and glycemia treatments, which was assessed by examining the two-way interaction between treatments. Log(-log(survival)) plots revealed that the assumption of proportional hazards was appropriate.

All analyses were performed by intention-to-treat where participants were analyzed based on treatment arm assignment regardless of treatments received or adherence to therapies. Similarly, the entire follow-up period, which included follow-up after participants assigned to intensive glycemic management were transitioned to standard therapy, was used for all models. All tests of significance were performed at the two-sided 5% a-level. Because 10 outcomes were examined simultaneously, the probability of one or more significant findings was approximately 40%, substantially higher than the nominal 5%. We chose not to perform a Bonferroni-type adjustment because that would require significance at the 0.005 a-level in order to ensure an experiment-wide type I error rate of 5%. Analyses were performed using the SAS version 9.2 software (SAS Institute, Cary, NC). All outcomes presented in this manuscript were prespecified in the study protocol.