Nephrology literature watch

Abstract

This month, I review three recent studies from the literature addressing issues important to the care of the peritoneal dialysis (PD) patient. A number of core questions related to modality choice center on whether PD offers specific patient-centered outcomes benefits or whether specific PD prescriptions might result in improved hard (non-surrogate) patient-centered outcomes. When considering whether an intervention results in changes in outcomes, the randomized controlled trial (RCT) has been considered the “holy grail,” as when it is performed rigorously the opportunity for the introduction of bias or confounding is minimized. However, the challenges of conducting RCTs with dialysis modality choice are well documented.1

In this literature watch, I review an RCT that highlights additional limitations that might arise from a well-designed RCT of the small size typical of many RCTs in nephrology. Specifically, I consider what might happen if the baseline risk of individuals randomly assigned to the experimental and control treatments differ significantly. Bias is reduced by randomization if sufficient numbers of patients are included in the study such that by chance alone all important baseline prognostic (known and unknown) factors that might influence the outcome are distributed equally to both groups, and that the groups differ only in the intervention under investigation. In small RCTs the equivalence of prognostic characteristics in each arm of the study cannot be assumed.2, 3

Additionally, when addressing the issue of whether a particular clinical finding might predict poorer outcomes or even result in harm, the RCT is not the ideal study design. To identify a factor as potentially harmful and likely to be causal of an adverse outcome, observational findings evaluated in the context of the Bradford-Hill considerations are preferred. In addition to the ethical constraints on conducting an RCT on a question of causation or harm, well-designed observational/epidemiological studies may be most informative because they are conducted under real-world conditions and may include patients expressing the full spectrum of baseline risk. Here too the size of the study population is likely to matter. In this literature watch I review two recent observational studies that interrogate patient databases to provide evidence about potential harm related to a clinical feature or the lack thereof. The first study evaluates the use of PD for initiation of unplanned dialysis compared with an initiation with hemodialysis (HD). In the second observational study, a study exploiting a very large observational database, the authors investigate whether depressed serum albumin levels are similarly associated with mortality in HD and PD patients.

Citation: Takatori Y, Akagi S, Sugiyama H, et al. Icodextrin increases technique survival rates in peritoneal dialysis patients with diabetic nephropathy by improving body fluid management: A randomized controlled trial. Clin J Am Soc Nephrol. 2011;6:1337–1344.

Analysis: Takatori and colleagues report the results of an RCT evaluating whether use of icodextrin as the osmotic agent in PD fluid results in preservation of the PD technique in patients with newly diagnosed end-stage renal disease (ESRD) and diabetes. They define technique preservation largely around the ability of the PD to adequately remove fluid.4, 5 As a secondary outcome, they evaluated preservation of renal function and peritoneal membrane function.

It has been established in multiple clinical trails that use of icodextrin in place of dextrose solutions (2.5%) results in improved net ultrafiltration and control of volume. The novel finding in this study is that these previously reported findings extend to incident dialysis patients with diabetes. The study was pre-registered in the Japanese clinical trials registry (JPRN registry WMIN00001040) with the control of volume as the primary outcome and with the intended enrollment of 100 patients. The primary finding of this study was that icodextrin resulted in preservation of function defined as the ability to achieve adequate volume removal when compared with standard PD with 2.5% Dianeal.

Validity and threats to validity: The optimal study design to assess a question regarding a therapeutic intervention remains the RCT or a systematic review of high-quality RCTs. When well conducted, an RCT can reduce the risk of bias. There are significant limitations to RCTs when performed less than optimally that may distort the findings reported. Some of these are widely recognized and include non-masking of group assignment, non-blinding leading to secondary interventions impacting outcomes, misclassification of outcomes, and so on. Recently, M.W. Walsh and colleagues have attempted to quantify the risk of important imbalances in baseline prognostic characteristics that might occur by chance in RCTs that are too small to ensure that a balance has occurred as a result of random group allocation (personal communication, June 2011).

Although this study appears to be a randomized trial with concealment of allocation, a number of the features of the current study may be problematic in the interpretation of the results and their application. First, the study is very small; significantly smaller than the size indicated as necessary in the pre-study plan. No formal power calculations are included in the published report. In such a small study, if prognostic characteristics are not balanced the results might deviate significantly from the “truth.” A recalculation of the effect size using non-parametric statistical tests and changing the outcome of one subject in each group would result in a significant change of the primary finding reported. Importantly, the study is underpowered to evaluate the patient-centered outcomes of peritoneal membrane and renal function survival or any major side effects including mortality and infection. It appears that group assignment may not have been masked after the initial randomization, so that the clinicians could have intervened in other ways (e.g., education, diet) that might have influenced volume control independent of the PD fluid interventions under study. Patients in the control arm were not treated with higher percentage glucose solutions as might be the case in the U.S. for patients who failed to achieve adequate net volume removal.

Application of the results and the clinical bottom line: This is a randomized controlled trial that demonstrates improvement in volume management using icodextrin to perform PD as compared with glucose-containing solutions. These findings reiterated multiple other RCT and observational trial findings in renal-failure patient populations. This study does not provide any new evidence about whether icodextrin might result in improvements in peritoneal membrane or renal survival. Before the conversion to icodextrin as the PD fluid of choice can be recommended, additional RCTs of sufficient duration and size need to be conducted. These RCTs need to determine if patient-centered outcomes can be improved upon significantly with the substitution of icodextrin for glucose-based PD solutions.

Citation: Koch M, Kohnle M, Trapp R, Haastert B, Rump LC, Aker S. Comparable outcome of acute unplanned peritoneal dialysis and haemodialysis [published online ahead of print May 28, 2011]. Nephrol Dial Transplant. doi:10.1093/ndt/gfr262.

Analysis: The issue of whether patients requiring urgent renal replacement therapy (RRT) can be safely managed with PD has here-to-fore not been rigorously investigated. The current study by Koch and colleagues begins to investigate this question. Ideally, an RCT comparing urgent PD to HD would most unambiguously address this question. The ability to conduct such a study despite equipoise has been restricted, however, by a strong clinical bias in the nephrology community that urgent PD cannot be conducted safely in most clinical circumstances. In such an environment, a well-designed observational study can provide evidence supporting the safety (lack of harm) of PD for urgent initiation of dialysis opening up the possibility for the appropriate RCT. Koch and colleages have exploited their unique clinical environment that allows them to provide PD urgently in a closely observed hospital setting to compare their experience with urgent PD versus urgent HD.

Validity and threats to validity: As an observational study, it is impossible to exclude bias that might have influenced the results. The most important of these is a selection bias where healthier patients are systematically more likely to receive one versus the other treatments being compared. In the case of the current study, it appears that patients with more severe cardiac disease were more likely to be encouraged to choose PD as treatment. Such an imbalance would be expected to negatively impact the outcomes (mortality, hospitalization rates, etc.) amongst patients undergoing PD. The absence of an observed difference (possibly even a trend favoring PD) can be attributed to the study being underpowered. Alternatively, the absence of a difference in outcomes might be due to an imbalance in prognostic factors, which would be expected in the case of this study to make a superior treatment option such as PD appear less favorable in comparison. Statistical methods to manage the differences in important prognostic factors between the two groups are imperfect.

Application of the results and the clinical bottom line: Importantly, this study may provide the necessary evidence of safety with the use of PD in urgent initiation of RRT and, therefore, open up the possibility of an RCT that will test the use of PD for emergency initiation of dialysis. The study results support the conclusion that urgent PD is safe and can be implemented equally effectively as HD for urgent initiation of RRT. If safe, a potential strategy based on PD for urgent RRT warrants further study as a means of reducing HD catheter-related infections—a significant cause of morbidity and mortality among patients new to dialysis. Treating a larger fraction of incident ESRD patients with PD might have other favorable consequences on morbidity, mortality, and quality of life yet to be determined.

Citation: Mehrotra R, Duong U, Jiwakanon S, et al. Serum albumin as a predictor of mortality in peritoneal dialysis: Comparisons with hemodialysis [published online ahead of print May 19, 2011]. Am J Kidney Dis. doi:10.1053/j.ajkd.2011.03.018.

Analysis: Mehrotra and colleagues exploit a large observational database to investigate whether depressed serum albumin levels are similarly associated with mortality in HD and PD patients. This study is important for two major reasons: First, if the impact of a low albumin is similar in PD and HD patients, PD patients may be placed in higher risk from excessive peritoneal protein losses and therefore, incentives and quality measures designed to prevent hypoalbuminemia might be warranted; and second, if interventions are to be tested or advocated to correct the hypoalbuminemia, the optimal target for serum albumin in PD versus HD patients should be established. This may be seen as an important precursor to studying interventions to alter albumin and to stratify the study populations according to who is most likely to benefit. In the current study, the authors have used the DaVita dataset containing the clinical parameters and outcomes for all patients receiving RRT by DaVita over a five-year period. They demonstrated a significant adjusted risk of mortality and cardiovascular mortality among all patients receiving RRT who were significantly hypoalbuminemic. Importantly, they demonstrated that the increase risk is not seen in PD patients until their serum albumin levels are observed to be below 3.8 g/dL. In contrast, in HD patients the threshold for increased risk with a depressed albumin begins at values below 4.0 g/dL.

Validity and threats to validity: Prior to initiation of RCTs to test interventions to normalize serum albumin levels in patients undergoing RRT, it should be firmly established that there is an increased mortality risk associated with the lower serum albumin and whether this risk is modified by treatment modality. This study provides substantial evidence of this association and that the risk might be different for patients treated with PD versus HD. The power of this study rests in that the observations are made using a very large database representing the full spectrum of patients and their comorbidities. The interrogated database represents a long enough period of observation of sufficient duration that it would be reasonable to expect to observe an impact of hypoalbuminemia on mortality. The study cannot, however, prove a causal relationship between a low albumin and mortality. In particular, despite the large size of the population, the robustness of the data allowing for morbidity adjustments, and the precision of the estimates, confounding cannot be excluded. The authors note these limitations. It is, however, fair to note (as the authors do) that despite the limitations of the evidence, agencies that monitor healthcare quality often chose to measure quality using parameters that arise from such observational studies. The rigor of this observational study and the precision of the estimates of the threshold make the findings from this study most compelling.

Application of the results and the clinical bottom line: While it is uncertain whether hypoalbuminemia itself is causal for some of the observed increased cardiovascular and all-cause mortality in ESRD patients, the current study by Mehrotra and colleageus adds significantly to our current understandings about serum albumin and nutrition in ESRD patients by more precisely describing the impact of a low albumin on different classes of ESRD patients. This study should provide evidence that will help in the design of clinical trials investigating interventions to correct low serum albumin levels in ESRD patients. Since the decision to switch patients from PD to HD is often influenced by the persistence of a lower serum albumin in PD patients, the results of this study might provide rationale—pending confirmation by an RCT—for a strategy that results in fewer patients switching off of PD and moving to HD. At a minimum, this study should raise the possibility that a slightly higher albumin achieved by switching a PD patient to HD might not translate into a significant survival advantage. This hypothesis requires further testing.

The two observational studies reviewed above provide significant insights into safety and harm or risk. As such, these observational studies may be informative for clinical practice. Thus, well-conducted observational studies can provide important insights especially related to risk or harm. In contrast, the first study reviewed above highlights some of the limitations presented by RCTs of small size—sizes typical of the nephrology literature. While the RCT is the optimal study design to investigate a therapy, the RCT reviewed here demonstrates that the results of even a well-designed and well-conducted RCT may, at times, need to be interpreted with caution. The plethora of small RCTs in nephrology and the difficulty of conducting larger trials in ESRD patients should not provide justification for our failure to conduct large, sufficiently powered RCTs on many of our current therapies for the complications of ESRD.