Does Treatment of Non-Malignant Hypertension Reduce the Incidence of Renal Dysfunction? a Meta-Analysis of 10 Randomised, Controlled Trials
Total Page:16
File Type:pdf, Size:1020Kb
Journal of Human Hypertension (2001) 15, 99–106 2001 Nature Publishing Group All rights reserved 0950-9240/01 $15.00 www.nature.com/jhh ORIGINAL ARTICLE Does treatment of non-malignant hypertension reduce the incidence of renal dysfunction? A meta-analysis of 10 randomised, controlled trials CY Hsu Division of Nephrology, University of California, San Francisco, CA, USA Objective: It remains controversial whether non-malig- dysfunction outcome varied among trials but within nant ‘benign’ hypertension causes renal dysfunction. each trial was applied similarly to both treatment and The effect of lowering blood pressure on the incidence control groups. Drug treatment consisted mostly of of renal dysfunction among patients with non-malignant diuretics and adrenergic blockers. Overall, treated hypertension is not clear. This meta-analysis was con- patients had lower blood pressure and fewer cardio- ducted to determine whether antihypertensive drug vascular events. There were a total of 317 cases of renal therapy reduces the incidence of renal dysfunction in dysfunction. Patients randomised to antihypertensive patients with non-malignant hypertension. therapy (or more intensive therapy) did not have a sig- Methods: Randomised, controlled trials of antihyperten- nificant reduction in their risk of developing renal dys- confidence interval %95 ;0.97 ؍ sive drug therapy of more than 1 year duration that function (relative risk .(0.77 ؍ reported renal dysfunction as an outcome were iden- (CI) 0.78–1.21; P tified through MEDLINE search and literature review. A Conclusions: Among patients with non-malignant random effects model was used to obtain summary esti- hypertension enrolled in randomised trials, treated mates. patients did not have a lower risk of renal dysfunction. Results: Ten trials were identified, involving 26 521 indi- The 95% CI suggests that a 25% or more true protective viduals and 114 000 person-years. All excluded subjects effect of antihypertensive drugs is unlikely. with advanced baseline renal disease. Definition of renal Journal of Human Hypertension (2001) 15, 99–106 Keywords: therapy; high blood pressure; kidney disease Introduction lowering blood pressure decreases the incidence of renal dysfunction among patients with non-malig- It remains controversial whether non-malignant nant hypertension is a randomised controlled trial. 1–5 ‘benign’ hypertension causes renal dysfunction. However, renal disease has not been a major focus in Although elevated blood pressure predicts subsequent 6,7 the large traditional antihypertensive drug treatment development of end-stage renal disease, it is poss- trials. Furthermore, there are too few renal end- ible that pre-existing renal disease caused both the 8,9 points within any one study to provide sufficient hypertension and the eventual renal failure. power and to make definitive conclusions. There- The effect of lowering blood pressure on the inci- fore, this meta-analysis was conducted to integrate dence of renal dysfunction among patients with the results of previous trials into a more precise esti- non-malignant hypertension is also not known. mate of treatment effect. Specifically, it asks Some observational studies have shown that for whether in randomised, controlled trials, drug treat- hypertensive patients, treated blood pressure level ment of non-malignant hypertension reduces the did not protect against subsequent rise in serum cre- incidence of renal dysfunction. atinine.10,11 The strongest study design to address whether Methods Correspondence: Dr Chi-yuan Hsu, Division of Nephrology, Identification of trials University of California, San Francisco, Room 672 HSE, 513 Parnassus Avenue, San Francisco, CA 94143-0532, USA. E-mail: Randomised controlled trials of antihypertensive hsuchiȰmedicine.ucsf.edu drugs among patients with non-malignant hyperten- Received 28 June 2000; accepted 10 July 2000 sion were identified using prior meta-analyses, bib- Treatment of non-malignant HTN CY Hsu 100 liographies in recent articles and textbooks and MED- example, the paper describing the main findings of LINE searches. Locating relevant studies was the United States Public Health Service Hospitals facilitated by the fact that numerous meta-analyses of (USPHS) study listed ‘renal insufficiency’ as an this literature have been performed that used similar exclusion criterion and as a secondary outcome.20 trial selection criteria.12–16 However, these published An earlier paper defined ‘renal insufficiency’ as systematic analyses have mostly focused on cardio- measured creatinine clearance Ͻ80 ml/min/ vascular disease outcomes. Extra effort was expended 1.73m2.24 on locating recent publications and a rigorous MED- In all trials, whenever possible, intention-to-treat LINE search strategy designed by the Cochrane Col- analysis outcome data were used as input data for laboration to identify trials of pharmacological treat- the meta-analysis. In the European Working Party on ment of hypertension16 was conducted for the period High Blood Pressure in the Elderly Trial (EWPHE), 1995 through 1998. A total of 2617 records were pro- serial serum creatinine measurements were perfor- duced.Thetitleandabstractofeachrecordwere med only during the randomised part of the study, evaluated and full text of 31 articles obtained for but mortality and cause of death data continued to further evaluation. No additional trials met eligibility be collected on all patients after withdrawal from criteria for this review. randomisation.28 During randomisation, one treated patient died of renal disease and four treated and one control patients had doubling of serum creatin- Inclusion criteria ine. In the overall intention-to-treat analysis, four Trials of randomised antihypertensive drug therapy treated and one control patient died of renal dis- included in the meta-analysis are those published ease.28,29 Because ascertainment of serum creatinine before 1 January, 1999 that: (1) lasted over 1 year; doubling may be more reliable than nosologist’s (2) explicitly described random allocation of treat- coding of cause of death, the ‘on treatment’ data ment (ie, did not use alternate allocation or other were used. In the Systolic Hypertension in the Eld- potentially biased methods); and (3) reported renal erly Program (SHEP), a rise in serum creatinine to dysfunction as an outcome. Trials that compared the Ͼ265 mol/l (3 mg/dl) was explicitly listed as a sec- effects of two specific antihypertensive therapies ondary endpoint.30 In addition, it was reported that with the same blood pressure goal were excluded. two patients each in the treatment and control arm Trials of multiple risk factor interventions were died of renal disease, which was not a stated end- excluded because of potential confounding by other point. These were therefore not counted. Performing therapies. Also excluded were trials that enrolled the meta-analysis using these alternative endpoints only patients with known renal insufficiency or did not significantly change the final conclusions of established renal parenchymal disease17–19 since the study. this study was not designed to examine the effect of In two instances, after publication of the paper blood pressuring lowering in patients who already describing the main trial findings, secondary analy- had renal dysfunction. ses of trial data were performed which provided more information on renal function changes.21,31 The meta-analysis was repeated to incorporate the Extraction of data results of these secondary analyses. The author used a standardised abstraction sheet to extract the following data from each study: year of Statistical analysis publication, number of subjects, mean age, percent male, baseline systolic and diastolic blood The effect of antihypertensive medication on the pressures, types of medication used, blinding, use of development of renal dysfunction was determined placebo or a less intensely treated group as control, in each trial using the relative risk (RR), calculated duration of follow-up, and mean difference in by the ratio of the incidence of renal dysfunction in observed blood pressure between treatment and con- the treatment vs control group. In cases where there trol groups. The proportion of black (or non-white20) were no events, a 1/6 arithmetic correction was participants was also abstracted. This was hypo- added to all cells.32 A random-effects model as thesised a priori to be a possible source of hetero- described by DerSimonian and Laird was used to geneity among studies, as black subjects are believed obtain the summary estimates of log (RR) from the to be particularly susceptible to renal damage from included studies.33 Effect homogeneity was evalu- hypertension.21,22 ated by the Q-statistic.33 Stata Release 6 (College Information on whether and how patients with Station, TX, USA) was used to perform the analysis renal disease were excluded from entry into each and to produce graphic output (command ‘meta’). trial was abstracted. Also noted was how new cases of renal dysfunction were defined in the paper Results presenting the main findings of each trial. In some instances, renal exclusion criteria and definition of Ten trials were identified that met the inclusion cri- renal dysfunction were found in earlier publications teria20,27,28,30,34–39 (Table 1). These encompassed a detailing the study protocol or pilot study.23–26 For total of 26 521 individuals and 114 000 person-years Journal of Human Hypertension Treatment of non-malignant HTN CY Hsu 101 Table 1 Characteristics