Association of Impaired Diurnal Blood Pressure Variation With a Subsequent Decline in Glomerular Filtration Rate FREE

Ambulatory 24-hour blood pressure (BP) monitoring has become an accepted method for evaluating hypertension or the efficacy of antihypertensive therapy. In most individuals, there is a physiologic decline in nocturnal systolic BP (SBP) of at least 10% from daytime values.¹ Patients lacking a nocturnal decline in BP have been termed nondippers, while those with normal diurnal BP variation are termed dippers.² Nondipping has been associated with chronic kidney disease (CKD),^3- 5 hypertension in older persons,^6- 7 obstructive sleep apnea,⁸ and metabolic derangements such as Cushing syndrome,⁹ types 1 and 2 diabetes mellitus (DM),^10- 13 and insulin resistance¹⁴ independent of mean 24-hour SBP. In addition, population-based studies have correlated nondipping with target organ damage, including cardiovascular morbidity and mortality,^1,15- 16 progression of preexisting renal disease,^17- 20 and cerebrovascular disease.^16,21- 23

The prevalence of CKD is increasing, paralleling the increasing prevalence of DM and hypertension among our aging population.²⁴ Loss of diurnal BP variation is common in patients with existing CKD.^{3- 4,17,19,25} In several prospective studies,^17- 20 nondippers with renal disease were shown to have a more rapid progression of their renal disease than dippers. Although systemic hypertension is clearly a risk factor for deterioration of renal function,²⁶ nondipping independent of mean 24-hour SBP has not been demonstrated to be a risk factor in patients without preexisting renal disease. The objective of the present study was to examine the prognostic effect of diurnal BP variability on subsequent renal function in a cohort of patients, most of whom had normal renal function at baseline.

We conducted a retrospective cohort study of outpatients referred to the Department of Hypertension and Nephrology at the Cleveland Clinic Foundation, Cleveland, Ohio, for ambulatory BP monitoring between December 29, 1994, and July 23, 2003. Consecutive patients 18 years and older were eligible for inclusion if they (1) did not have end-stage renal disease requiring renal replacement therapy, (2) had no history of renal transplantation, and (3) had an initial serum creatinine level noted at the time of their ambulatory BP recording and a follow-up serum creatinine level recorded at least 1 year from that date. We did not exclude patients based on their initial glomerular filtration rate (GFR). When multiple follow-up creatinine levels were available, we used the most recent value, excluding values that were associated with an acute illness or a hospitalization. The Cleveland Clinic Foundation Institutional Review Board approved the study.

Twenty-four–hour ambulatory BP and heart rate (HR) monitoring was performed using Spacelabs 90207 (Spacelabs Medical, Redmond, Wash). Monitors recorded HR, SBP, and diastolic BP readings every 15 minutes during the day and every 30 minutes during the night for the 24-hour period. The mean day (6 AM to 11 PM) and night (11 PM to 6 AM) SBPs (daytime SBP–nighttime SBP ratio) were compared. These time cutoffs were determined prospectively. We prospectively defined nondipping as a daytime SBP–nighttime SBP ratio higher than 0.90.¹

Clinical information was obtained from written and electronic medical records. This included height, weight, medical history, current medications, laboratory data, and smoking history.

Glucose, triglyceride, uric acid, total cholesterol, and high-density lipoprotein (HDL) cholesterol levels were determined by standard enzymatic methods (Roche Diagnostics, Indianapolis, Ind). Low-density lipoprotein cholesterol levels were calculated from total triglyceride, total cholesterol, and HDL cholesterol levels using the standard formula. High-sensitivity C-reactive protein levels were measured using a particle-enhanced immunonephelometric assay (Beckman Coulter, Brea, Calif). Glycosylated hemoglobin A_1c levels were measured using an ion-exchange high-performance liquid chromatography assay (Tosoh, Mendocino, Calif). The GFR was calculated using the Modification of Diet in Renal Disease 4-variable equation.²⁷ Echocardiograms were performed by the Department of Cardiovascular Medicine at the Cleveland Clinic Foundation using standard protocols.

To test differences between proportions, we used the χ² test or the Fisher exact test. t Tests were used to compare continuous data across groups. Variables with highly skewed distributions were log transformed before all analyses. Univariate logistic regression analysis was used to calculate odds ratios associating nondipping with clinical variables. Adjusted odds ratios were calculated using multiple logistic regression analysis. For continuous variables, we expressed the odds ratio per 1-SD increase in the variable (or its log-transformed value). Because many variables were associated with the main independent variable (nondipping), we used a propensity score method to adjust for confounding, which is discussed in the next paragraph. Univariate and multiple logistic analyses were used when the dependent variable was continuous (eg, percentage GFR change). In determining which variables (other than nondipping) were significantly associated with the mean GFR change, we dichotomized continuous variables at the median to avoid bias introduced by choosing optimal cutoff values retrospectively. The Brown-Forsythe test was used to assess whether variances were unequal.

To create a propensity score for each subject, we constructed a nonparsimonious multiple logistic regression model with nondipping as the dependent variable. We incorporated the following clinical variables into the model as independent variables: age, sex, active smoking, body mass index, race (white vs nonwhite), gout, DM, sleep apnea, obstructive lung disease, congestive heart failure, history of hypertension (defined before the ambulatory BP recording), use of any antihypertensive medication, specific use of various antihypertensive medications (including diuretics, calcium channel antagonists, β-adrenergic receptor antagonists, angiotensin-converting enzyme [ACE] inhibitors, or angiotensin II receptor blockers), baseline laboratory test values (GFR and fasting glucose, log triglyceride, uric acid, total cholesterol, HDL cholesterol, low-density lipoprotein cholesterol, glycosylated hemoglobin A_1c, log high-sensitivity C-reactive protein, and baseline serum urea nitrogen and creatinine levels), left ventricular ejection fraction, echocardiographic left ventricular hypertrophy (present, absent, or unavailable), diastolic dysfunction (present, absent, or unavailable), 24-hour mean arterial pressure, and 24-hour mean SBP, diastolic BP, and HR. For the propensity score model, missing values were assigned the mean value for the given variable. Because urinary microalbumin data were available for only 45 patients and because this variable could not be assumed to be missing entirely at random (because this test is primarily ordered among patients at risk for renal disease), it was not included in the propensity score model. Variables reflecting diurnal hemodynamic variation (daytime and nocturnal HR and BP) were excluded from the propensity score model. Propensity scores were divided into deciles, assigning each patient a score between 1 and 10, reflecting the likelihood of being a nondipper.

The C statistic for the propensity score model (using deciles) was 0.82, indicating a very good fit. Multiple logistic regression analysis was then used to calculate adjusted odds ratios using the propensity score decile as an independent (adjuster) variable. We did not use any other variables as adjusters in the propensity score–adjusted models because the propensity score was successful in attenuating all systematic (adjusted) differences between dippers and nondippers. We did not impute missing values for models other than the propensity score model; for other models, subjects with missing data were omitted from analysis.

Three hundred twenty-two patients met the inclusion criteria, 137 dippers (mean ± SD daytime SBP–nighttime SBP ratio, 0.85 ± 0.04) and 185 nondippers (mean ± SD daytime SBP–nighttime SBP ratio, 0.97 ± 0.06). The median follow-up was 3.2 years. Characteristics of dippers and nondippers are given in Table 1. The clinical indication for ambulatory BP measurement was most often to assess control of ongoing antihypertensive therapy or to evaluate white-coat hypertension. Ninety-two dippers (67.2%) and 162 nondippers (87.6%) had a diagnosis of hypertension. Compared with dippers, nondippers tended to be older and have greater use of antihypertensive agents, worse end-of-study GFRs, higher fasting glucose levels, higher triglyceride levels, lower HDL cholesterol levels, higher serum urea nitrogen and creatinine levels at baseline, higher 24-hour mean arterial pressures and SBPs, lower nocturnal drops in HR, and higher frequencies of preexisting hypertension, coronary artery disease, and heart failure. After propensity score adjustment, only those variables excluded from the propensity score model (nocturnal drops in HR and BP, end-of-study GFRs, and number of patients with rapid GFR decline) differed significantly by dipping status, suggesting that the propensity score adjustment was successful in controlling for the other clinical variables.

Among patients with initial GFRs greater than 65 mL/min per 1.73 m² (median baseline GFRs in this subset, 85.3 mL/min per 1.73 m² for dippers and 86.3 mL/min per 1.73 m² for nondippers; P = .32), 30 (24.0%) of 125 nondippers vs 4 (3.7%) of 108 dippers developed CKD using the standard GFR cutoff of 60 mL/min per 1.73 m² as defined by the National Kidney Foundation (P<.001 and propensity score–adjusted P = .006). A greater variance in baseline GFRs was noted among nondippers than among dippers (SD, 24.1 vs 18.8 mL/min per 1.73 m²; P = .005). Similarly, the variance in GFR decline was substantially greater among nondippers than among dippers (percentage GFR change standard deviation, 24.2% among nondippers and 17.2% among dippers, P<.001; with similar findings for actual GFR change in milliliters per minute per 1.73 m², P<.001), suggesting that nondippers tended to be more heterogeneous in their GFR trajectories than dippers or that intraindividual GFR fluctuations were more pronounced among nondippers. The Figure shows the percentage change in serum creatinine levels among dippers and among nondippers, illustrating the variability in GFR decline among nondippers. Dippers were more likely than nondippers (78.1% vs 40.0%, P<.001) to have stable serum creatinine levels, defined by an increase of less than 10% during follow-up, whereas nondippers demonstrated substantial heterogeneity in creatinine level changes over time.

We determined the univariate relationships between all variables in Table 1 and GFR decline. All variables with significant associations with GFR decline are included in Table 2, which gives the mean percentage GFR decline among various patient subsets. In univariate analysis, GFR decline was associated with age, DM, dipping status, measured and reported hypertension, congestive heart failure, coronary artery disease, hypertriglyceridemia, and glycosylated hemoglobin A_1c levels. None of these variables was significantly associated with the duration of follow-up. Some of these relationships were attenuated by adjustment for dipping status. Further adjustment for the use of drugs that affect the GFR (which might be used among patients with congestive heart failure and DM in particular) did not affect these findings, as summarized in Table 2.

Table 3 gives the results of a subgroup analysis in which nondipping predicted GFR decline in all prespecified subsets of patients (those with DM vs nondiabetics, those with baseline hypertension vs normotension, those taking antihypertensive drugs that affect the GFR vs those not taking such drugs, and those with baseline CKD vs normal renal function, using the standard GFR cutoff of 60 mL/min per 1.73 m²). In all subsets, nondipping was significantly associated with a decline in renal function, and there were no significant interactions between these variables and dipping status in predicting GFR change, suggesting that the heterogeneity in GFR decline among nondippers (reflected by the higher variance in GFR change) was not readily explained solely by the presence or the absence of other important clinical variables.

Although the correlation of nondipping with existing CKD has been previously reported,^{3- 5,17- 19} we present (to our knowledge) the first longitudinal data showing that nondipping is associated with a subsequent decline in renal function independent of baseline renal function, mean 24-hour SBP, and other risk factors for renal impairment. In our patient sample, the dippers had a mean GFR change of 1.3% (95% confidence interval, –1.6% to 4.2%) from baseline, suggesting no overall change in their renal function, because it is unlikely that GFRs actually increased. In contrast, the nondippers exhibited a mean GFR change of –15.9% (95% confidence interval, −19.4% to −12.3%). This finding persisted after rigorous adjustment for confounding variables using a propensity score method.

Impaired diurnal BP variation is common among patients with several renal diseases. In one study,³ nondippers constituted 82% of patients on hemodialysis, 78% of patients with elevated serum creatinine levels, 75% of patients on peritoneal dialysis, and 74% of renal transplant recipients. Similarly, Csiky et al¹⁷ found that 93% of a cohort of hypertensive patients with IgA nephropathy were nondippers. Normotensive nondippers in this cohort had significantly higher serum creatinine levels at the end of 36 months of follow-up than normotensive dippers. In another study,⁴ patients with hypertension and autosomal dominant polycystic kidney disease had blunted diurnal BP patterns compared with patients with hypertension alone. Among patients with type 1 DM, nondipping was shown prospectively to be a risk factor for the development of microalbuminuria.¹⁸ Finally, Timio et al¹⁹ followed up 48 patients with renal disease of various etiologies, 20 of whom were dippers and 28 of whom were nondippers. The nondippers had a faster decline in renal function during a 3-year follow-up.

Mechanistically, it has been proposed that nondipping reflects increased sympathetic nervous system (SNS) tone,^29- 30 which may result from and contribute to kidney dysfunction. The link between nondipping and increased SNS activity has been demonstrated in different populations. Among patients with essential hypertension and normal renal function, spectral analysis demonstrated blunted HR variability in nondippers. Daytime low frequency–high frequency ratios were significantly higher in nondippers vs dippers, and nighttime high frequency of nondippers was significantly lower than that of dippers.³¹ In addition, normotensive patients with type 2 DM and normal renal function had a significantly higher prevalence of nondipping and blunted HR variability than control subjects.³² Increased renal SNS activity diminishes renal sodium excretion and blood flow, decreases the GFR via renal vasoconstriction, and may exacerbate albuminuria in patients with DM.^33- 34 In turn, increased systemic SNS activity is common in patients with CKD and may be may be an important contributor to the high rate of cardiovascular events among patients with renal disease.^24,35- 36

The prevalence of insulin resistance among nondippers may have contributed to our findings. Recent data from the Third National Health and Nutrition Examination Survey demonstrated that patients with the metabolic syndrome had a greater risk of developing CKD than patients without the metabolic syndrome.³⁷ In previous work, insulin resistance as determined by the triglyceride–HDL cholesterol ratio²⁸ was linked to nondipping.¹⁴ In the present cohort, patients who had triglyceride–HDL cholesterol ratios of at least 3.0 had significantly greater GFR decline compared with the other subjects during follow-up (−12.0% [95% confidence interval, −16.1% to −7.8%] vs −3.6% [95% confidence interval, −7.0% to −0.3%]). However, after adjustment for nondipping, that relationship was attenuated and was of borderline statistical significance (Table 2).

Although we cannot make therapeutic recommendations based on this observational study, we speculate that treatments for nondipping should be directed at hypertension and associated cardiovascular risk factors (rather than BP patterns per se) or at increased sympathetic tone. In particular, ACE inhibitors and angiotensin II receptor blockers are renoprotective in populations with various causes of renal disease and may be reasonable first-line agents for hypertensive nondippers. In contrast, despite the relationship between SNS activity and nondipping, β-adrenergic blockade is less efficacious than ACE inhibition in retarding nephropathy, and α-adrenergic blockade appears to be inferior to other antihypertensive medications in preventing major cardiovascular end points.^38- 39 Whether the known sympatholytic effects of ACE inhibitors and angiotensin II receptor blockers contribute to their renoprotective effects remains unproven.^40- 42 Finally, it is not known whether nondippers may benefit from lower target BPs to preserve renal function.

A few limitations of our study deserve mention. Because this was a retrospective study, longitudinal assessments of hypertension control and changes in antihypertensive therapy were beyond the scope of the investigation. It is possible that GFR-affecting drugs were preferentially added to the regimens of nondippers following their ambulatory BP recordings (compared with dippers), although the statistical adjustment for DM and mean 24-hour BP makes this explanation unlikely. Also, multiple assessments of ambulatory BPs were not performed.⁴³ However, intraindividual variation in dipping status would probably tend to weaken our findings. Although care was taken not to use serum creatinine values that reflected an episode of acute illness, even minor fluctuations in creatinine levels can lead to substantial changes in the calculated GFRs, especially using the Modification of Diet in Renal Disease equation, which has limited accuracy for calculating GFR values in the normal range.⁴⁴ An imprecision in GFR measurements would also bias toward negative findings, but the greater variance in GFR change among nondippers vs dippers may be a reflection of measurement imprecision. Perhaps the most important limitation of our study is that urinary albumin measurements were available for fewer than 15% of our subjects (n = 45). Because microalbuminuria often precedes the deterioration in renal function, it would have been useful to have these data for all subjects. Although we examined this variable in a univariate fashion, we did not have enough data to optimally control for it. Finally, the duration of follow-up in our study was brief. Investigators in future studies may be able to ascertain whether nondipping predicts the development of end-stage renal disease among patients with normal renal function at baseline.

In summary, although the physiologic mechanism for our findings remains incompletely understood, it is clear that nondipping is a risk factor for deterioration of renal function, independent of mean 24-hour SBPs per se. This relationship appears to be independent of DM, baseline CKD, and antihypertensive treatments. We suggest that this relationship may be mediated through increased renal and systemic SNS activity and perhaps via insulin resistance. Further elucidation of the mechanisms responsible for impaired diurnal BP variation among patients with and without renal disease may clarify the pathogenesis of CKD in the general population and the associated cardiovascular morbidity.

Correspondence: Daniel J. Brotman, MD, Department of Medicine, The Johns Hopkins Hospital, Jefferson Room 242, 600 N Wolfe St, Baltimore, MD 21287 (brotman@jhmi.edu).

Accepted for Publication: October 16, 2005.

Financial Disclosure: None.