Meta-Analysis: Low-Dose Dopamine Increases Urine Output but Does Not Prevent Renal Dysfunction or Death

Dopamine is a catecholamine with dose-dependent effects on the systemic and renal vasculature. In healthy participants, low-dose dopamine increases renal blood flow and promotes natriuresis through stimulation of renal D1, D2, and D4 receptors and thus may protect the kidney from acute tubular necrosis (1). The concept of low-dose or renal-dose dopamine has persisted since the first clinical description of its use in patients with congestive heart failure (2). Few controlled trials have demonstrated any benefit, and as a result, several editorials have discouraged its use (3-6). Nevertheless, recent surveys have documented dopamine's continued popularity. For example, 17 of 24 New Zealand intensive care units (7) and 18 of 19 pediatric and neonatal intensive care units in the Netherlands (8) use low-dose dopamine to treat renal dysfunction or oliguria. Moreover, this therapy continues to attract substantial research resources: Several randomized trials have been published each year, and at least 1 large trial is ongoing (9). Two recent systematic reviews have addressed low-dose dopamine. Both reviews (10, 11) had several methodologic limitations. Neither review analyzed dopamine's effects on urine output and adverse events. Other authors (12) have discussed the limitations of 1 review (10) and have called for a rigorous, updated meta-analysis conducted by independent researchers (13). Given the limited scope and methodologic concerns of previous systematic reviews and the ongoing widespread use and study of low-dose dopamine, we conducted a systematic review and meta-analysis by using a comprehensive search strategy to determine its effect on a broad range of clinical and renal physiologic outcomes and adverse events. Methods Search Strategy We searched the OVID versions of MEDLINE (1966January, week 4, 2005), EMBASE (1980week 5, 2005), CANCERLIT (1975October 2002), CINAHL (1982January, week 3, 2005), and CENTRAL (Cochrane Central Register of Controlled Trials, The Cochrane Library, fourth quarter, 2004). We also searched the Renal Health Library (available at www.update-software.com/publications/Renal/) on 3 February 2005. Two authors conducted independent search strategies. The first MEDLINE search strategy retrieved citations containing the subject heading dopamine (limited to the publication types clinical trial and meta-analysis) or the text words low dose dopamine or renal dose dopamine. The second MEDLINE search strategy retrieved citations containing the subject heading dopamine combined with exploded subject headings describing renal disease (kidney diseases and kidney) and physiology (kidney function tests, urine, and renal circulation) or text words describing low-dose dopamine appearing in close proximity to each other (low, renal, kidney, dose, and dopamine). We limited citations from the second search to randomized, controlled trials using a maximally sensitive strategy (14). We modified these searches for other databases. Full details of both search strategies are available from the authors. We screened reference lists from all retrieved articles and from recent review articles (8, 10, 11, 15-24) to identify additional studies. There were no language restrictions. Study Selection and Characteristics We selected parallel-group randomized, controlled trials that included any patient sample, compared low-dose dopamine (5 g/kg of body weight per minute) with placebo or no therapy, and recorded any of the following outcomes: all-cause mortality, requirement for renal replacement therapy, renal physiologic variables (urine output, serum creatinine level, or measured creatinine clearance on days 1, 2, or 3 after starting therapy), or adverse effects. We also included trials in which patients were allocated in alternating fashion or by hospital registry number (quasi-randomization) and trials with pharmacologic co-interventions (such as mannitol, diuretics, or diltiazem) that were equally applied to both groups. We defined a priori adverse effects of interest likely to be detected by routine patient monitoring: arrhythmias, myocardial ischemia, and limb or cutaneous ischemia. Data Abstraction Two reviewers independently screened studies for inclusion; retrieved all potentially relevant studies; and extracted data on study sample, intervention, prespecified outcomes, and methods from included trials. In both phases, disagreements between reviewers that remained after contacting study authors were resolved by consensus. We assessed agreement between the 2 reviewers on the selection of articles for inclusion by using Cohen's (25). Validity Assessment We extracted methodologic information important for the assessment of internal validity: method of allocation and concealment of the randomization schedule, blinding of caregivers and outcomes assessors, and the number of and reasons for postrandomization withdrawals. We determined whether fluid and diuretic therapies were standardized or equally applied in dopamine and control groups. We attempted to contact all authors of trials that met our inclusion criteria. Quantitative Data Synthesis For each outcome of interest, we pooled all studies reporting the outcome on the basis of the a priori expectation of similar direction and magnitude of treatment effect. For mortality and renal replacement therapy, we combined studies reporting these outcomes at any time after randomization. For studies with 2 or more dopamine groups receiving different doses, we combined data from all doses to determine an overall outcome measure for the dopamine group. We used Review Manager 4.2 (The Cochrane Collaboration, Oxford, United Kingdom) to aggregate data for each outcome by using a random-effects model (26), and we considered a P value of 0.05 or less to be statistically significant. All pooled effect estimates are presented with 95% CIs, and all P values are 2-sided. Results for binary outcomes (mortality, need for renal replacement therapy, and adverse effects) are reported as relative risks. For studies with no events in one arm, 0.5 is added to all cells. (Studies with no events in either arm are not included in the pooled analysis.) The summary relative risk is calculated on the natural logarithm scale. The weight of each study is calculated as the inverse of the variance of the natural logarithm of its relative risk. In the presence of between-study heterogeneity, each study's weight is adjusted (26). Because clinical outcomes occurred infrequently in many trials, we also analyzed these outcomes by using other effect measures, including Peto's odds ratios and random-effects risk differences (by using Review Manager 4.2) and exact odds ratios (by using StatXact 6.1 [Cytel Software Corp., Cambridge, Massachusetts]). These alternate methods gave very similar pooled treatment effects and CIs, and we therefore present only the random-effects relative risk data. Some papers reported 2 of the renal physiologic outcomes, urine output and creatinine clearance, in units that were adjusted for body weight or body surface area. However, trials did not consistently provide average patient weights, which are required to convert these measures into values with identical units, and they enrolled highly variable study samples that included adults, children, and neonates. Identical measurement units are necessary to statistically analyze the results by using weighted mean difference as the measure of treatment effect. One alternative approach would have pooled these measures by using the standardized mean difference (the absolute treatment effect in pooled SD units). However, we chose to summarize the treatment effect for each continuous outcome by using the relative change in the dopamine group compared with the control group. This approach provides a more clinically meaningful summary of treatment effect than the standardized mean difference. For each continuous outcome, we calculated the ratio of the mean value in the dopamine group to the mean value in the control group for each study and calculated a standard error for the natural logarithmtransformed ratio (see Appendix). We aggregated the natural logarithmtransformed ratios across studies by using the generalized inverse variance method (27). In studies where investigators obtained the total urine output over 24 hours by addition of urine outputs over several time periods, we calculated the variance of the total urine output by using the method of Follmann and colleagues (28). We assumed a moderate correlation ( ) of 0.4 between urine outputs at different time periods. Sensitivity analyses using correlations of 0 and 0.8 did not change the results. We considered only first-order correlations. We assessed between-study homogeneity for each pooled comparison by using the Cochran Q-test (29), with a P value of 0.10 or less indicating significant heterogeneity (30). We also report the I2 statistic, which is the proportion of total variation among studies that is explained by between-study heterogeneity rather than chance (31, 32). Substantial heterogeneity exists when I2 exceeds 50%. We developed several a priori hypotheses to explain statistically significant heterogeneity: 1) populationgreater treatment effect in trials enrolling surgical patients (fewer comorbid conditions) compared with medical patients; 2) baseline riskgreater treatment effect when low-dose dopamine was given for treatment rather than prevention of acute renal failure; 3) doseresponse relationshipgreater treatment effect with a dopamine dose 3 g/kg per minute vs. <3 g/kg per minute; 4) duration of therapygreater treatment effect in trials where dopamine was given for the entire time period before measurement of the outcome; and methodssmaller treatment effect 5) in studies with adequate allocation concealment (vs. all other studies) and 6) in studies with blinding of caregivers. For each hypothesis, we statistically tested the difference in estimates of treatment