Use of the Modification of Diet in Renal Disease Equation for Estimating Glomerular Filtration Rate in the Urologic Literature

Introduction

The modification of diet in renal disease (MDRD) equation is a commonly used formula to estimate the glomerular filtration rate (GFR) of the kidneys using the variables of serum creatinine (SCr), gender, age, and race. GFR is considered the most accurate manner in which to estimate clinical renal function. However, direct measurements of GFR cannot be easily obtained in clinical practice, prompting the development of several equations to estimate this clinical parameter. These include the Cockcroft–Gault formula, ¹ MDRD ² equation, and a newer set of equations from the chronic kidney disease epidemiology collaboration (CKD-EPI) study group ³ (equations are listed in Supplementary Data S1; Supplementary Data are available online at www.liebertpub.com/end). Given its improved precision over previous equations, the MDRD equation has become the most commonly used formula to estimate GFR. ² In today's clinical practice, more than 70% of laboratories report estimated glomerular filtration rate (eGFR) with all adult SCr results. ⁴

As an expected correlate to this use in clinical practice, eGFR derived from the MDRD equation has been used commonly in the urologic literature as a clinical endpoint for urologic outcomes. This has been done, however, despite considerable literature demonstrating that the MDRD equation carries important deviation from true, iothalamate GFR when estimating GFR greater than 60 mL/min per 1.73 m². ⁵ While reasonably accurate at values of eGFR less than 60 mL/min per 1.73 m², estimates of GFR above 60 mL/min per 1.73 m² tend to be inaccurate and systematically underestimate GFR. ⁶ The estimates of GFR based on MDRD have been noted to vary from the true GFR by as much as 29% to 39%, and as the eGFR estimates rise, MDRD calculated levels become more unstable and can be off by as much as 40%. ^6,7 For example, a healthy, 55-year-old Hispanic woman with an SCr of 0.7 mg/dL would have an MDRD eGFR of 87 mL/min per 1.73 m², but were she to have an SCr measured the following day and it was 0.8 mg/dL, we would attribute no great significance to this because this could be due to random error, but that represents a 15% drop in eGFR to 74 mL/min per 1.73 m². Based on this and similar reports, it is widely recommended that clinical laboratories not report actual numerical values of eGFR greater than 60, and instead report “GFR >60 mL/min per 1.73 m².” ⁸ Despite this finding, many clinical studies have continued to use specific eGFR values based on the MDRD equation as endpoints in clinical research, regardless of whether the values are above or below 60 mL/min per 1.73 m². Furthermore, many of these studies have made comparisons of eGFR claiming superiority of one therapy or another at preserving kidney function when the claimed difference is actually less than the error of the measurement.

In the current study, our goal was to evaluate the urologic literature regarding the use of the MDRD equation. We sought to evaluate the appropriateness of equation use and whether the limitations of the equation were discussed in the articles that elected to use MDRD to estimate GFR.

Materials and Methods

A medical librarian searched Medline (using the Ovid interface) on December 18, 2013, for articles that included the phrases “MDRD” or “modification of diet in renal disease” in the title, abstract, or keywords. The time frame for the search was 2004 to 2013. The journals included in our search were European Urology, Journal of Urology, Urologic Oncology–Seminars and Original Investigations, BJU International, World Journal of Urology, Urology, and Journal of Endourology. These journals were chosen as they had ISI impact factors >2.0 and were urology focused.

From this query, 54 articles were identified (references listed in Supplementary Data S2). Articles were initially reviewed for the appropriateness of inclusion into the study. Articles were included if they involved the comparison of eGFR values among two or more groups of patients, or if the eGFR was used in a multivariate model. Three articles were excluded because the primary objective was to compare MDRD to alternative methods of estimating GFR and were not comparing GFR values. Three articles were excluded because MDRD was only used as a component of a nomogram or as demographic descriptor of a patient population without comparison of values. Finally, one article was excluded because it involved a pediatric patient population. In sum, 47 articles were included in the analysis.

Each of the included articles was reviewed in its entirety by two independent reviewers (J.R.Z. and G.L.) and graded on the appropriateness of its use of MDRD to estimate GFR through the use of a grading scale developed by the authors for the current article. The full details of this grading scale are outlined in Table 1. In short, inappropriate use was defined as using GFR calculations based on the MDRD equation to make comparisons and conclusions with the majority of the GFR values >60 mL/min per 1.73 m². Grade 1 articles appropriately used MDRD, Grade 2 articles used MDRD inappropriately, but the primary claim of the article did not rely on these results, and Grade 3 articles inappropriately used MDRD for the primary claim. We also evaluated whether each article acknowledged the limitations of the MDRD equation for calculating eGFR. Any discrepancy in the grading performed by the initial reviewers was re-reviewed by a third reviewer (C.J.W.) and the grade assigned by two of three reviewers was used for analysis. The ISI Impact Factor for each journal article was complied. Differences in impact factor between the Grade groups were evaluated using ANOVA. SPSS was used to perform comparative statistical analysis and evaluate outcomes.

Results

Forty-seven articles published between 2004 and 2013, from the journals outlined above, were reviewed. Seventeen of the articles were Grade 1 (36%), 20 were Grade 2 (43%), and 10 were Grade 3 (21%), as established by the criteria outlined in Table 1. With regard to whether the articles contained a discussion or acknowledgment of the limitations of the MDRD equation to estimate GFR, we found that, overall, only 34% of articles (16 of the 47) acknowledged limitations. When stratified by grade, 24% of Grade 1 articles (4/17) and 40% of Grade 2 articles (8/20) acknowledged MDRD limitations. When focused on Grade 3 articles (those most inappropriately using MDRD), only 40% (4 of 10) of articles acknowledged the limitations of using MDRD to estimate GFR. There was no significant difference in impact factor of the journals in the three grade groups (p = 0.959).

Discussion

The measurement of SCr and the subsequent estimation of GFR have become prominent in everyday medical practice as the primary method of estimating renal function. MDRD is currently the most commonly used equation to estimate GFR and is frequently used in the medical literature. In this study, we identified a sample of articles from the urologic literature using the MDRD equation. In total, we found that 64% of articles were inappropriately using the MDRD equation (the majority of the calculated values of GFR >60 and conclusions about changes in renal function were made). Furthermore, of those using this equation, only 34% discussed and/or acknowledged the limitations of the equation.

The MDRD study produced an equation to eGFR based on the variables of SCr, age, gender, and race. At the time of its initial publication, this equation provided a more accurate estimate of GFR than the measured creatinine clearance or other previously used equations ² and as a result became the predominant equation for estimating GFR. However, considerable literature subsequently identified variable accuracy in the equation among various study populations, especially among patients with normal renal function. ^6,7 Given that the MDRD equation was based on a study population with known chronic kidney disease, it is not surprising that this equation is most accurate in patients with low GFR. Subsequent large studies of more diverse populations identified that the MDRD equation indeed exhibits increased inaccuracy above GFR estimates of 60 mL/min per 1.73 m² and tends to underestimate GFR in these patients. ^5,6 Given this concern regarding the accuracy of the equation to estimate GFR values >60 mL/min per 1.73 m², it is clear that MDRD should be used with caution in the literature, particularly to study patients with normal renal function. This is of particular importance as studies comparing various types of nephron sparing surgical techniques become increasingly common.

The suboptimal performance of the MDRD equation has led to further research to identify alternative estimates of GFR. The CKD-EPI subsequently identified such an alternative set of equations that were more accurate than the MDRD equation. ³ Because the CKD-EPI equations were developed using a more diverse patient population, they yielded more accurate results with less deviation from the true GFR, particularly in patients with normal renal function. In one direct comparison between MDRD and CKD-EPI, the CKD-EPI equations were found to be more accurate across most subgroups, and particularly with eGFR >60 mL/min per 1.73 m², enough so that the authors suggested that these values could be reported with SCr in clinical practice. ⁹

Direct comparisons between these equations have also been applied to the urologic population. Lane et al. compared MDRD and CKD-EPI equations for estimating GFR before and after nephrectomy. With measured GFR through ¹²⁵I-iothalamate renal clearance, these authors identified CKD-EPI equations more precise and accurate than the MDRD or CG equations. ¹⁰ Shikanov et al. also compared MDRD and CKD-EPI equations for patients undergoing partial nephrectomy and found that CKD-EPI equations identified 7% fewer patients as having Stage III CKD or worse ¹¹ (likely correlated to less underestimation of GFR). Novel markers such as serum cystatin C have also been evaluated to develop equations for eGFR measurement that may provide further improvements in accuracy that are not linked to muscle mass. ¹² However, these newer measurements remain infrequently used in standard urologic practice. In short, a growing body of literature has identified the limits of the MDRD equation. While CKD-EPI is not without its own potential inaccuracies, it appears to provide a better estimate of GFR in many clinical circumstances. Despite this, the MDRD equation remains a commonly used mechanism in the literature to estimate GFR.

Regardless of which equation is used, using any SCr assay to estimate GFR is also rife with possible sources of error. These may include inaccuracies of SCr assays themselves secondary to measurement imprecision and/or calibration bias that can lead to uncertainty in GFR estimate. ^7,13 In addition, daily variations in SCr measurements secondary to hydration status, medications, and dietary intake may lead to inaccuracies in solitary measurements of creatinine for GFR calculation. Thus, any calculation of GFR based on a solitary measurement of creatinine may not accurately reflect the true GFR of a given patient. In this context, using such values to declare clinical differences in a research setting must be interpreted with caution.

There are several limitations to this study that must be discussed. Our method of article selection only applied our search criteria to title, keywords, and abstracts of articles rather than the entire body of text. In addition, we chose a selection of seven journals with higher impact factors rather than including all urologic journals. While this limits the number of articles for review and does not provide a complete review of all literature, we felt this provided an appropriate representative sample of the most commonly read and cited urologic journals. We also devised a novel grading scale of the appropriateness of MDRD application that has not been externally validated as a method for evaluating the accuracy of individual article results. However, we felt this provided a descriptive framework for the purposes of evaluating how MDRD was being used in these articles and is not intended in any way to invalidate or call into question the results of any particular article. We acknowledge that many of the articles reviewed have made valuable contributions to the urologic literature and used MDRD in appropriate ways. However, the estimation of GFR using MDRD and the use of this value for drawing clinical conclusions when eGFR is >60 mL/min per 1.73 m² should be done with caution.

Conclusion

The majority of the articles reviewed inappropriately used MDRD to report and/or compare renal function values of eGFR >60 mL/min per 1.73 m². Furthermore, despite considerable evidence that the MDRD equation provides estimates that demonstrate substantial deviation from true GFR in patients with normal renal function, less than half of the articles that inappropriately applied this equation discussed these limitations. We suggest that authors should consider more accurate equations for eGFR estimation such as CKD-EPI equations and at the least should discuss the limitations of these equations, as readers should be aware of these limitations and interpret results appropriately.

Authors' Contribution

J.Z.: Project development, data collection, data analysis, manuscript writing. G.L.: Project development, data collection, manuscript editing. J.K.: Project development, manuscript editing. D.L.: Project development, manuscript editing. J.K.A.: Project development, manuscript editing. C.J.W.: Project development, data analysis, manuscript writing/editing.

Compliance with Ethical Standards

This article was prepared without the use of patient information or human participants. The authors declare no directly relevant conflicts of interest, but all potential conflicts and associations are listed below.