Achievement and Safety of a Low Blood Pressure Goal in Chronic Renal Disease

Study Design

The MDRD Study was a multicenter, randomized clinical trial conducted from January 1989 to January 1993 to determine the effect of dietary protein restriction and strict BP control on the progression of renal disease.^{20 23 24} Details of the study design, patient recruitment, and baseline characteristics have been published.^{23 25 26 27} The study was approved by the investigational review boards of all participating centers, and all participants gave informed consent.

In brief, patients were eligible to enter the MDRD Study if they were 18 to 70 years old and had chronic renal disease with serum creatinine ≥1.4 to 7.0 mg/dL (124 to 619 μmol/L) in men or ≥1.2 to 7.0 mg/dL (106 to 619 μmol/L) in women. Hypertension was not an inclusion criterion. Pregnant women, patients with diabetes requiring insulin, urinary protein excretion >10 g/d, or a history of renal transplantation were excluded, as were patients with unrepaired known renal artery stenosis or urinary tract obstruction and a number of systemic diseases that affect the kidney. Patients with congestive heart failure (New York Heart Association class III or IV) were excluded, as were patients with three or more hospitalizations within the previous 60 days.

During a 3-month baseline period, BP, dietary intake, and renal function were assessed.²⁶ BP was expressed in terms of MAP, computed as two thirds diastolic BP plus one third systolic BP. The MAP goal was ≤107 mm Hg for patients between 18 and 60 years old and ≤113 mm Hg for patients ≥61 years old. MAP ≤125 mm Hg was required for randomization. Patients with GFR 25 to 55 mL/min per 1.73 m² (0.42 to 0.92 mL/s per 1.73 m²) and dietary protein intake ≥0.9 g/kg per day entered in study A (n=585), and patients with GFR 13 to 24 mL/min per 1.73 m² (0.22 to 0.40 mL/s per 1.73 m²) entered into study B (n=255).

Twelve-lead electrocardiograms were obtained by standard techniques during the baseline period. Tracings were interpreted in a central electrocardiogram laboratory by two blinded readers separately and then jointly. A consensus reading obtained jointly became the official interpretation. Left ventricular hypertrophy was considered present when a repolarization abnormality was found in V₅ or V₆ as follows: depressed ST segment with upwardly convex descending limb of an inverted or diphasic T-wave, with or without a delayed intrinsicoid deflection.

Study participants had renal diseases of diverse causes, the most common being polycystic kidney disease (25% of all patients) and glomerular disease (24%). Only 3% had diabetic nephropathy. A total of 116 patients (14%) did not have hypertension (defined as a history of hypertension based on a review of the medical record and use of antihypertensive drugs).

For the diet intervention, patients in study A were randomly assigned to either a usual or low protein diet (goals of 1.3 or 0.58 g/kg per day, respectively), and patients in study B were randomly assigned to either a low or very low protein diet (goals of 0.58 or 0.28 g/kg per day, respectively). Goals for the diet interventions were based on standard body weights (the midpoint of weight-for-height tables by frame size).²⁸ For the BP intervention, the MAP goal was determined by the patient's age and assignment to either the usual or low BP group as specified in Table 1⇓. The target MAP of ≤107 mm Hg for the usual BP group is equivalent to a BP of ≤140/90 mm Hg, the usual accepted upper limit of normal BP.²⁹ The low BP target MAP of ≤92 mm Hg is equivalent to a BP of ≤125/75 mm Hg and is a more strict level of BP control than is usually recommended.

As indicated in Table 1⇑, only the upper targets for MAP were specified; the study protocol specified no lower limits for MAP in either BP group. As the study progressed, some patients in the usual BP group were found to have BPs that were below the goal of the low BP group. Accordingly, the MDRD Study Steering Committee instructed the clinical centers to allow MAP levels to rise in patients assigned to the usual BP group to a level above the goal for the low BP group, when this was acceptable to the patient and referring physician and caused no short-term adverse effect.

The BP Intervention

Antihypertensive regimens were based on a modified version of the stepped-care approach defined in the 1988 Report of the Joint National Committee.³⁰ This approach considers nonpharmacological therapy, such as reduction in sodium intake, loss of weight, reduction in alcohol intake, and other changes in lifestyle, as the first step. Nonpharmacological therapy was incorporated with the protein prescription as long as it did not compromise adherence to the diet intervention. Pharmacological therapy was deemed necessary in virtually all hypertensive patients. During a 3-month baseline period, antihypertensive regimens were adjusted to achieve MAP goals as specified for the usual BP group in follow-up. In some cases, this adjustment required discontinuation of antihypertensive medications to allow the BP control achieved by the referring physician to be relaxed. After randomization, the regimens were modified to achieve either the low or usual BP goals. All antihypertensive medications were allowed, but ACE inhibitors and calcium channel blockers were encouraged as agents of first and second choice, respectively. In addition, the ACE inhibitor enalapril (Merck & Co) and the calcium channel blocker diltiazem (Marion Merrell Dow) were provided free of charge by the manufacturers. Patients without hypertension taking antihypertensive drugs for other indications (eg, β-blockers for angina) were allowed to continue these drugs. During baseline and follow-up, MAP levels were monitored by monthly clinic BP measurements made with a Hawksley random-zero sphygmomanometer (W.A. Baum, Inc), according to established methodology.³⁰ Patients were placed in a quiet room for 5 minutes before measurement. The sitting BP was taken three times by the same nurse or technician. The average of the last two measurements was used as the BP for the visit.³¹ Each technician was trained and certified annually in the proper techniques of BP measurement and recording, in the use of the equipment, and in the study's quality-control procedures. Antihypertensive medications were modified monthly, or more often if necessary, to achieve target levels. Adherence strategies included review of medications and possible side effects at each visit and simplification of medication regimens as possible. Pill counts were not performed.

Safety of the BP Intervention

Safety of the BP intervention was assessed by examining the frequency of changes in antihypertensive medications, the frequency of symptoms that could be attributed to low BP, and the occurrence of previously defined “action items” and “stop points,” hospitalizations, and deaths.

BP medications and changes in prescriptions were recorded at each monthly visit. Also each month, the patients reported the frequency of occurrence of 24 symptoms that might be attributed to low BP.

Conditions requiring modifications of the diet, antihypertensive regimen, vitamin prescription, or frequency of measurements (action items) were defined and reported as they occurred. These conditions included an action item for “symptoms of low BP,” which required physicians to allow an increase in BP goal by 5 mm Hg.

Conditions requiring withdrawal from the study (stop points) also were defined and reported as they occurred. These included malnutrition (weight loss to <75% of standard body weight or serum albumin <3.0 g/dL [<30 g/L]), rapid decline in GFR (study A only, to <50% of baseline if baseline GFR was ≤40 mL/min per 1.73 m² [0.67 mL/s per 1.73 m²] or to ≤20 mL/min per 1.73 m² [0.33 mL/s per 1.73 m²] if baseline GFR was >40 mL/min per 1.73 m² [0.67 mL/s per 1.73 m²]), end-stage renal disease requiring dialysis or transplantation, and serious intercurrent medical conditions.

Information on hospitalizations was obtained routinely during study visits or if a study visit was missed. Reasons for hospitalization were derived from ICD-9 codes.³² All codes in section 7 of the ICD-9 code list referring to cardiovascular or cerebrovascular diseases were used to classify cardiovascular or cerebrovascular hospitalizations. Only the first hospitalization was used in the analyses to avoid giving too much weight to patients with multiple hospitalizations.

Patient deaths were reported with a special form. Information about the cause of death was sought from several sources, including family members, hospital records, and the death certificate. Cause of death was classified by the clinical center principal investigators.

Statistical Methods

MAP measurements averaged about 1.4 mm Hg higher at GFR visits than at non-GFR visits throughout the follow-up period (paired t test, P<.001), perhaps because of water loading at GFR visits. This difference in MAP between GFR and non-GFR visits was similar for patients assigned to the usual and low BP groups, for patients with and without a history of hypertension, and for patients in both studies A and B. As a result of this systematic bias, mean follow-up MAP in the MDRD Study is defined as the mean of all follow-up MAP measurements obtained at non-GFR visits beginning with the third monthly follow-up visit. Mean follow-up systolic and diastolic BPs were defined similarly. The measurements were averaged to reduce the effect of short-term fluctuations in BP.

Box plots were used to illustrate the distribution of BP measurements over time between the usual and low BP groups. Student's t test was used to compare the BPs (systolic, diastolic, and MAP) between the usual and low BP groups.

Antihypertensive medications were categorized according to whether they were being used at more than or less than 50% of follow-up visits. Fifty percent was selected as the cut point because, for most patients, use of each class of antihypertensive agent was close to 0% or close to 100% of visits. The χ² test was used to compare the use of each medication class during follow-up between the usual and low BP groups.

Longitudinal analysis was used to relate BP measurements during follow-up to baseline patient characteristics to determine whether certain patient subgroups experienced a larger or smaller separation between the BP groups.³³ This analysis assumed a two-slope, mixed-effects spline model in which each patient was assumed to have an initial BP slope from baseline to 10 months and a different slope subsequently. This method was used to account for varying lengths of follow-up among the various patient subgroups. Two slopes were used to account for a greater rate of increase in the separation of mean MAP during the first 10 months of follow-up than in the remainder of follow-up.

Frequencies of initial hospitalizations were tabulated and the percent per patient year of follow-up was determined. To identify any association between BP and frequencies of deaths, hospitalizations, or stop points, the quartiles of the three BP measures (systolic, diastolic, and MAP) were computed. The frequencies (percent) of each measure of safety per patient year of follow-up were tabulated by BP quartile.

Time-dependent Cox proportional hazards regression³⁴ was used to relate the cumulative mean follow-up BP measurements to frequencies of initial hospitalizations. Risk ratios were computed to indicate the change in risk of hospitalization associated with a 10–mm Hg increase in each of the cumulative mean follow-up BP measurements. To assess the effects of traditional cardiovascular risk factors or other possible confounding conditions, risk ratios were also evaluated using a multivariate regression model with the following baseline covariates: age, sex, renal diagnosis (polycystic kidney disease, glomerular disease, or other and unknown), history of cigarette smoking, and serum concentrations of low-density and high-density lipoprotein cholesterols.

The indexes of follow-up BP considered in this article did not differ meaningfully between the diet groups among patients assigned to either the standard or low BP goals. There was also no significant interaction between the BP and diet interventions in the assessments of the safety of the BP intervention. Consequently, we present comparisons of the BP groups for all study participants irrespective of their dietary assignment.

All hypothesis tests are two-sided, without adjustment for multiple comparisons.

Baseline Characteristics

Selected characteristics of the patients at baseline are shown in Table 2⇓. Although 86% of patients had a history of hypertension, few patients had target-organ damage. Only 6.6% had a diagnosis of hypertensive nephrosclerosis; 9.6% and 1.5% had a history of coronary artery disease and cerebrovascular disease, respectively; and only 3.9% had evidence of left ventricular hypertrophy by electrocardiography. Funduscopic changes of severity equal to or worse than grade II (Keith-Barker-Wagner classification) were present in 11% of patients.

Use of antihypertensive drug use is also shown in Table 2⇑. At the first baseline visit, 83% of patients were taking antihypertensive medication, with a mean number of 1.6 medications per patient. The classes of antihypertensive agents prescribed in the two studies combined were as follows: ACE inhibitors, 37%; calcium channel blockers, 26%; β-blockers, 32%; diuretics, 40%; and other agents, 19%.

Medications During Follow-up

Patients were followed on their diet and BP interventions for an average of 2.2 years. By the end of follow-up, the mean number of antihypertensive drugs prescribed per patient in study A was 1.9 and 1.5 in the low and usual BP groups, respectively. In study B, the mean number per patient was 2.1 and 1.8, respectively. Table 3⇓ shows the number and percentage of patients who were prescribed each of the major classes of antihypertensive drugs, alone or in combination, for more than 50% of follow-up visits in studies A and B combined. ACE inhibitors, calcium channel blockers, and diuretics, but not β-blockers, were prescribed for more patients in the low BP group than the usual BP group. Overall, the three most common combinations of antihypertensive classes prescribed for more than 50% of follow-up visits were an ACE inhibitor plus a diuretic, an ACE inhibitor plus a calcium channel blocker, and an ACE inhibitor plus a calcium channel blocker plus a diuretic. In the low BP group, 9%, 4%, and 5% of patients received these three combinations, respectively. Corresponding proportions in the usual BP group were 9%, 3%, and 2%.

Achieved BP in the Low and Usual BP Groups

In studies A and B combined, the mean±SD MAP throughout follow-up was 93.0±7.3 mm Hg in the low BP group and 97.7±7.7 mm Hg in the usual BP group. The mean MAP for the low BP group is near the goal (≤92 mm Hg for patients ≤60 years old and ≤98 mm Hg for patients ≥61 years old). In the usual BP group, mean MAP was well below the goal (≤107 and ≤113 mm Hg in patients ≤60 or ≥61 years old, respectively).

In the low BP group, the percentage of patients with MAP measured at a single visit below the assigned target was 38% immediately before the initiation of the intervention and increased to 54% by the third month of follow-up and to more than 65% during the third year of follow-up. In the usual BP group, the percentage of patients with MAP at a single non-GFR visit less than the assigned target was 87% immediately before initiation of the intervention and remained steady at approximately 88% for the remainder of follow-up. When the MAP was averaged over all non-GFR visits, the proportion of patients in the low and usual BP groups with mean follow-up MAP less than the targets for the respective groups was 61% and 95%, respectively.

In both groups, BP was well controlled according to usual criteria. The percentage of patients with a MAP higher than the assigned target for the usual BP group did not exceed 20% at any non-GFR visit during follow-up in either the usual or the low BP group.

Difference in MAP Between Low and Usual BP Groups

In studies A and B combined, the difference in mean MAP between the BP groups was 4.7 mm Hg, which was statistically significant (P<.001). However, as shown in the Figure⇓, there was substantial overlap in the distribution of MAP measurements between the groups. This overlap was apparent for systolic and diastolic BPs as well and persisted even after exclusion of 116 patients (14%) without a history of hypertension.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

MAP (top), systolic BP (middle), and diastolic BP (bottom) for patients in the low (dotted line) and usual (solid line) BP groups in the MDRD Study. Left panels show all patients (n=840); right panels show subgroup of patients with a history of hypertension (n=724, 86.2%). BP measurements are shown at the second baseline visit (B2) and 5th, 11th, 19th, 27th, and 35th follow-up (F) visits (non-GFR visits, see text). Vertical lines give 25th and 75th percentiles of BP at each visit. Lines connect median values. Numbers of patients in the usual BP (n1) and low BP (n2) groups are given for the different visits. P values show the level of statistical significance for BP difference between groups at each visit. MAP was calculated as two thirds diastolic BP plus one third systolic BP.

The difference in mean±SE MAP values increased over time between the low and usual BP groups. There was virtually no difference between the two groups at baseline. During follow-up, separation of mean MAP values developed progressively, and the difference in mean MAP between the two groups increased to 6.0±0.5 mm Hg at 9 months and varied between 4.2 and 7.8 mm Hg for the remainder of follow-up.

Longitudinal analysis was used to relate BP measurements during follow-up to baseline patient characteristics to determine whether certain patient subgroups experienced a larger or smaller separation between the BP groups (Table 4⇓). Table 4⇓ provides the mean MAP at baseline and estimated mean MAP at 12 months of follow-up for selected patient subgroups. At baseline, mean MAP was significantly higher in each of the following groups: patients with a history of hypertension, patients with polycystic kidney disease or glomerular diseases, patients with urinary protein >1 g/d, and black patients. Irrespective of BP group, mean MAP at 12 months of follow-up was significantly higher in patients with mean baseline MAP >92 mm Hg in each of these subgroups and also in patients ≥61 years old.

The separation in mean MAP between the low and usual BP groups at 12 months was statistically significantly greater in patients with a history of hypertension (5.6 mm Hg) than in those without a history of hypertension (2.0 mm Hg), in patients with polycystic kidney disease (5.5 mm Hg) or glomerular disease (6.4 mm Hg) than in those with other or unknown renal diagnoses (3.4 mm Hg), and in patients entering the study with MAP >92 mm Hg (6.5 mm Hg) than in those with lower BPs (2.8 mm Hg).

Safety of the BP Intervention: Comparison of Randomized Groups

As described above, safety of the BP intervention was assessed by comparing a number of adverse outcomes between usual and low BP groups: the number of changes in antihypertensive medications, frequency of symptoms that could be attributed to low BP, and occurrence of action items, stop points, hospitalizations, and deaths.

During follow-up, antihypertensive medications or doses changed at 3840 patient visits (18.7% of patient visits): 2587 visits (17.9%) in study A and 1253 visits (20.8%) in study B. Changes occurred more frequently in the low BP group than the usual BP group: The mean percentages of visits per patient with changes were 21.0% and 16.3%, respectively, in study A (P<.001) and 18.7% and 16.4%, respectively, in study B (P=.038).

Of the 24 adverse symptoms, only the frequency of “feeling faint” differed significantly between the low and usual BP groups in both studies A and B. In study A, this symptom was reported in an average of 15% of follow-up visits per patient in the low BP group compared with 12% of follow-up visits per patient in the usual BP group (P=.012). In study B, the corresponding percentages were 18% and 12%, respectively (P=.009).

The total number of action items did not differ significantly between the BP groups in either study A or B. However, of the 17 patients (2%) who had an action item related to persistent symptoms of hypotension requiring a reduction in antihypertensive medications, 14 (3.2%) were in the low BP group and 3 (0.7%) in the usual BP group (P=.01).

Table 5⇓ compares the BP groups for the number and rates of stop points, initial hospitalizations for all causes, initial hospitalizations for cardiovascular and cerebrovascular disease, and deaths occurring while patients remained on the assigned interventions. The usual and low BP groups did not differ significantly in the occurrence of these events. However, the MDRD Study was not designed to detect a difference in the rate of these events, so the power for these analyses may be low. For example, combining patients in studies A and B, the risk ratio (and 95% confidence intervals [CI]) of adverse events in the low BP group compared with the usual BP group was 1.04 (0.79-1.39) for first hospitalizations due to all causes, 1.03 (0.59-1.79) for first hospitalizations due to cardiovascular or cerebrovascular disease, and 0.86 (0.65-1.15) for stop points. These events are described below for both BP groups combined.

Of 187 stop points reached during follow-up, 166 were the result of end-stage renal disease or rapidly declining GFR (study A only). The remaining 21 stop points were caused by serious medical complications, including the need for insulin (n=5); carcinoma (n=4); pregnancy (n=2); cerebral hemorrhage (n=2); malnutrition (n=2); and liver disease, abdominal aneurysm, emphysema, severe arthritis, cardiomyopathy, and acute renal failure (n=1 each). In the 2 patients who died from cerebral hemorrhage, there was no evidence that this complication was related to treatment-induced changes in BP. None of the other medical condition stop points appeared to be related to antihypertensive therapy. No patient reached a stop point from complications of hypotension.

There were 189 initial hospitalizations out of 275 total hospital admissions. The cause of 50 of the 189 initial hospitalizations (26%) was specified as cardiovascular or cerebrovascular according to the primary ICD-9 diagnosis code.

Overall, 16 patients died before reaching a stop point, and 14 additional patients died after a stop point but before their scheduled close-out visit. Of the 30 deaths, 15 were in study A (2.6%) and 15 in study B (5.9%). The causes of death were cardiovascular diseases (n=17), cancer (n=5), trauma (n=3), sepsis (n=2), renal failure (n=1), respiratory failure (n=1), and cerebrovascular accident (n=1). As discussed above, neither the overall nor cause-specific mortalities differed significantly between the BP groups.

Safety of the BP Intervention: Association Between Adverse Events and Achieved BP During Follow-up

Tables 6 and 7⇓⇓ examine the relationship between mean follow-up values of achieved systolic, diastolic, and mean arterial BPs with serious adverse events. In Table 6⇓, patients in studies A and B are combined and then divided into quartiles based on mean follow-up systolic, diastolic, or mean arterial BP. The quartiles are then compared for the number of patients with an event and the event rates for death, initial hospitalization from any cause, initial hospitalization from cardiovascular or cerebrovascular disease, and stop points. As shown, patients with higher systolic BP or MAP had higher rates of stop points, initial hospitalizations from all causes, and hospitalization from cardiovascular or cerebrovascular disease. For example, the rate of initial hospitalizations from all causes was 9.0% per patient year of follow-up in the lowest quartile of follow-up systolic BP (<119 mm Hg) compared with 19.3% per patient year of follow-up in the highest quartile (>138 mm Hg). Similarly, the rate of initial hospitalizations from cardiovascular or cerebrovascular disease increased gradually from 0.4% to 2.2% to 2.9% to 5.8% per patient year of follow-up over successively higher quartiles of systolic BP. In contrast, patients with higher diastolic BPs had fewer hospitalizations from all causes. The rate decreased gradually from 13.9% to 12.2% to 9.4% to 6.3% per patient year over successively higher quartiles of diastolic BP. However, there was no apparent relationship between hospitalizations from cardiovascular or cerebrovascular disease and diastolic BP.

Table 7⇑ shows the results of univariate regression analysis relating the risk of first hospitalization from all causes or from cardiovascular or cerebrovascular disease to BP during follow-up. MAP during follow-up was not significantly related to the risk of initial hospitalizations from all causes (P=.43). Higher systolic BP correlated significantly with an increased risk of hospitalizations from all causes in studies A and B combined (P=.001). The risk ratio was 1.17 per 10–mm Hg increase in systolic BP in both studies. However, controlling for baseline factors (age, sex, renal diagnosis, smoking, and serum levels of low-density and high-density lipoprotein cholesterols) in the multivariate regression analysis, the relationship was weaker and no longer significant (risk ratio=1.08, CI=0.97-1.20, P=.15). In univariate analysis, a decrease in the risk of first hospitalizations from all causes with higher diastolic BP also bordered on significance; the risk ratio was 0.83 per 10–mm Hg increase in diastolic BP (P=.07). However, this result was not significant in multivariate analysis (risk ratio=0.95, CI=0.77-1.77, P=.63).

The risk of initial hospitalizations for cardiovascular or cerebrovascular disease was directly related to follow-up MAP in univariate analysis (risk ratio=1.61, P=.009 in studies A and B combined). The risk was also significantly related to higher systolic BP (risk ratio=1.35, P<.001 in studies A and B combined). After controlling for baseline covariates, the risk ratios were reduced slightly (1.43 and 1.23 for MAP and systolic BP, respectively), but the results remained statistically significant (P=.02 and P=.03, respectively). Interestingly, in univariate analysis, there was no significant relationship between higher diastolic BP and increased risk for first hospitalization for cardiovascular or cerebrovascular disease (risk ratio=1.11, P=.60). In multivariate analysis, there was a trend (P=.07) for a relationship (risk ratio=1.44, CI=0.97-2.13).

Because the primary ICD-9 codes for 12 of the 50 initial hospitalizations for cardiovascular or cerebrovascular disease included the label “hypertension” in their definition, it is possible that some primary diagnoses were given this designation based on high BP measurements recorded on the patients' charts rather than on evidence of a cardiovascular or cerebrovascular condition per se. After exclusion of these 12 hospitalizations, the univariate associations between systolic BP and risk of hospitalizations remained significant in studies A and B combined (P=.04), marginal in study B alone (P=.07), and not significant in study A. The association between follow-up MAP and risk of hospitalization was not significant.

Analyses of Safety in Patients With Baseline Urinary Protein >1.0 g/d

Because the beneficial effect of the low BP goal in slowing the progression of renal disease was greater in patients with greater baseline urinary protein excretion, we repeated the analyses of the above two sections for the 266 patients (32%) in studies A and B with baseline urinary protein >1.0 g/d. Among these patients, in the low BP group, the numbers of deaths, initial hospitalizations, and initial cardiovascular or cerebrovascular hospitalizations were 4, 31, and 10, respectively. In the usual BP group, the corresponding numbers of these events were 1, 34, and 9. The numbers of these events did not differ significantly between the randomized BP groups.

The relationships between achieved follow-up BP and adverse events reported above for the entire group were also seen for the patients with baseline urinary protein >1.0 g/d. We observed a significant univariate relationship between higher systolic BP and rates of first hospitalization for all causes (risk ratio=1.45, P<.001) and for cardiovascular and cerebrovascular disease (risk ratio=1.33, P=.001).