Dr MC Smith, Department of Obstetrics and Gynaecology, School of Surgical and Reproductive Sciences, 3rd Floor William Leech Building, Medical School, University of Newcastle upon Tyne, Framlington Place, Newcastle upon Tyne, NE2 4HH, UK. Email firstname.lastname@example.org
It is now recommended practice to use estimated glomerular filtration rate (eGFR) values to screen for and monitor chronic renal disease. The most frequently used formula in the general population is that described following the Modification of Diet in Renal Disease (MDRD) study whereby serum creatinine is adjusted for age, gender and race. This study evaluates the performance of the MDRD formula in pregnancy by comparing eGFR with measured values obtained by inulin clearance studies in early and late normal pregnancy and in pregnancies complicated by renal disease or pre-eclampsia. Our results indicate that in all situations, MDRD substantially underestimates glomerular filtration rate during pregnancy and cannot be recommended for use in clinical practice.
Maternal renal adaptation is characterised by substantially increased glomerular filtration rate (GFR), typically 50% above the pre-pregnancy value.1 These changes, initiated prior to the first missed menstrual period, are fully established by the end of the first trimester and maintained thereafter. Most women, even with a background of impaired renal function, are able to adapt, but in a few women, renal function will become increasingly compromised, occasionally irreversibly. Furthermore, renal impairment may occur during pregnancy in women without pre-existing renal problems, particularly in the context of pre-eclampsia. It is important that renal function is closely monitored in these high-risk pregnancies to minimise maternal and fetal complications.
The ‘gold standard’ method to measure GFR is inulin clearance, but this is costly, cumbersome and not practical outside a clinical research setting. Creatinine clearance is commonly used to approximate GFR; however, this requires a timed urine collection, which is frequently incomplete, particularly in the outpatient setting, leading to clinically significant inaccuracies.
A noninvasive, rapid technique to estimate GFR in pregnancy has been sought by nephrologists but remains elusive. Several prediction formulas based on single-point serum parameters or novel serum markers have been used to estimate GFR with varying degrees of success, but there have been few attempts to employ these techniques in pregnancy. In the general adult population, the most widely used prediction equation was published following the Modification of Diet in Renal Disease (MDRD) study.2 Recently, the National Kidney Foundation in the USA and the National Service Framework for Renal Services in the UK have recommended laboratories routinely report estimated glomerular filtration rate (eGFR) whenever serum or plasma creatinine levels are determined, using the MDRD formula based on four parameters; serum creatinine, age, gender and race. The guidelines specifically exclude interpretation in pregnant women in the absence of validation.
Unlike many prediction formulas, the MDRD equation is not individualised for body surface area and is therefore theoretically attractive for use in the obstetric population in whom increased surface area is not a reflection of increased muscle mass and therefore of increased creatinine levels. Clinicians caring for obstetric patients may soon see eGFR routinely reported alongside the traditional biochemistry, and this study seeks to determine if this information might be clinically useful. We have therefore compared GFR measured by inulin clearance (Cin) with eGFR calculated by the MDRD formula in healthy pregnancies and in pregnancies complicated by pre-eclampsia and chronic renal disease.
Twenty-four healthy pregnant volunteers aged between 22 and 35 years were recruited. All subjects were normotensive with no history of renal, diabetic or vascular disease. GFR was serially determined using inulin clearance in early pregnancy (10–16 weeks of gestation), late pregnancy (33–39 weeks of gestation) and postpartum (>8 weeks following delivery). Ten women with pre-eclampsia (defined by International Society for the Study of Hypertension in Pregnancy criteria) were similarly studied in late pregnancy and postpartum. A third group of women known to have chronic renal impairment predating their pregnancy were studied on one or two occasions during pregnancy (15 studies in 10 women), women were included if eGFR was less than 90 ml/minute (the limit below which it is recommended that consideration should be given to the possibility of renal impairment).
On arrival at the research facility, an intravenous bolus of 10 ml inulin (25% Inutest; Fresenius Kabi, Graz, Austria) and 48 ml 0.9% normal saline (Baxter Healthcare, Newbury, UK) were given over 5 minutes followed by a sustaining infusion of inulin (70 ml) and normal saline (340 ml) at 60 ml/hour. After 60 minutes of equilibration, the volunteer voided and then three 20-minute urine collections were obtained, each with midpoint blood sampling. To minimise dead space error in the urine collections, an adequate diuresis was maintained throughout by ad libitum oral water intake. Analytical procedures and across-batch coefficients of variation for inulin and true creatinine have been described elsewhere, as have the formulas used in the clearance calculations.3 Measured GFR was compared with eGFR using the modified MDRD formula:
The mean and SD of GFR in each group using both inulin clearance and the MDRD formula was calculated. Agreement between the methods of measurement was assessed by calculation of the mean and SDs of the differences. Limits of agreement were used to help quantify the differences. A clinically acceptable level of agreement of 15 ml/minute was considered appropriate. Previous studies indicate that a difference of 10% can be expected between GFR measurements calculated using the gold standard technique of inulin clearance and the method used most commonly in clinical practice, creatinine clearance. Data previously published from our laboratory report GFR measured by inulin clearance in pregnancy to be 152 ml/minute (± 17.6). Sample size was determined using nomograms for paired continuous data. The study has 95% power to identify a difference in GFR of 15 ml/minute between techniques with a power of 95% at the 5% significance level during normal pregnancy (48 studies in 24 women). All volunteers gave informed consent for inulin clearance studies with protocols approved by the Joint Ethics Committee of Newcastle and North Tyneside Health Authority and the Universities of Newcastle upon Tyne and Northumbria at Newcastle.
Outside of pregnancy, eGFR was not significantly different to measured GFR (Cin), and the observed bias (11.9 ml/minute) was within a priori clinically relevant limits (15 ml/minute) (Figure 1, Table 1). During pregnancy, MDRD underestimated GFR (111.6 ± 19.6 ml/minute and 153.8 ± 28.2 ml/minute, MDRD versus Cin, respectively) and the bias (41.65 ml/minute) was considerable. These trends persisted even when the results were analysed by trimester.
Table 1. Values are GFR (ml/minute) (± SD) measured by inulin clearance (Cin) and estimated with the MDRD formula and the difference between the two measurements
95% limits of agreement
−25 to 49.5
−39.1 to 63.0
−12.7 to 96.0
−26.7 to 110.2
−13.6 to 78.8
−30.7 to 95.9
−6.3 to 107.8
−27.4 to 128.9
−19.4 to 66
−42.8 to 89.4
−31.1 to 85.6
−44.4 to 98.9
Ten women also had creatinine clearance measured to provide some comparison with a method clinicians are readily familiar with. Creatinine clearance was a better indictor of GFR than MDRD during pregnancy (bias 6.8 ± 34 ml/minute; 95% limits of agreement between −60 and 74 ml/minute, 95% CI −90.7 to 104.2).
In women with pre-eclampsia or established renal disease, MDRD underestimated GFR (86.6 ± 12.4 ml/minute and 104.5 ± 19.8 ml/minute, MDRD versus Cin, respectively), although to a lesser extent than in healthy late pregnancy.
In normal pregnancy, we have demonstrated that eGFR using the modified MDRD formula is not a useful measure of GFR.
Although the MDRD formula was derived from an adult population with chronic kidney disease most of whom had a GFR less than 60 ml/minute/1.73 m2, it has since been validated for use in the general population. Previous authors have reported that GFR tends to be underestimated in healthy individuals. In our study, MDRD consistently underestimated GFR in the absence of severe renal impairment, a finding which has been reported by other groups.4 It has been suggested that MDRD is a poor predictor of GFR in healthy individuals because fluctuations in serum creatinine are more dependent upon dietary protein intake and muscle mass than on kidney disease where GFR is the main determinant. This reasoning, however, cannot explain why eGFR is a poorer predictor of GFR during pregnancy than postpartum (nonpregnant) because creatinine turnover is not altered by pregnancy. Lower serum creatinine levels observed in pregnancy are a reflection of haemodilution due to increased plasma volume as well as hyperfiltration. Thus, the association between serum creatinine and renal function in pregnancy is weakened. Plasma volume continues to increase until 34 weeks of gestation, and this may in part explain the greater discrepancy between measured and eGFR in late pregnancy compared with early pregnancy.
In women with pre-eclampsia, the MDRD formula more closely predicted GFR. This is perhaps expected as these women have a significantly contracted plasma volume, and therefore, serum creatinine is more dependent upon renal function. Although MDRD was a better predictor of GFR in women with pre-eclampsia than in healthy pregnancies, GFR was still substantially underestimated. In women with known renal disease, similar results were obtained from studies during pregnancy. Exaggerated deterioration of GFR may prompt inappropriate concern and untimely intervention. Thus, the use of the MDRD equation during pregnancy to estimate GFR in women who have known renal impairment prior to pregnancy or who develop renal complications during pregnancy cannot be recommended.
There are concerns in the literature about bias with GFR prediction equations because of poor calibration of serum creatinine assays across laboratories. This was not a factor in our study because all the creatinine assays were performed on a single creatinine analyser. Our sample size was relatively small compared with many nonpregnant studies, but the validity of our findings is strengthened because women were serially studied and GFR was contemporaneously measured and calculated. The scatter of difference between estimated and actual GFR in pregnancy was much greater than in nonpregnant subjects and a larger population would need to be studied to achieve statistical significance; however, the statistical methods clearly demonstrate that eGFR is not a clinically useful measure of renal function in pregnancy.
In the UK, some laboratories plan to only report calculated eGFR values less than 89 ml/minute/1.73 m2, a threshold believed to be discriminatory in identification of renal disease in the general population. Values above this will be reported as greater than 90 ml/minute/1.73 m2. In common with any other measure of GFR, nonpregnant normal ranges are not applicable to pregnant women. Women who initially have normal renal function may experience clinically significant deterioration without eGFR falling less than 90 ml/minute/1.73 m2 even after allowance is made for substantial underestimation of GFR.
Women with renal disease are being increasingly monitored using eGFR, so it is inevitable that this will become the usual measurement of renal function available to the maternal medicine team at booking or at pre-pregnancy counselling. Discussion of the likely impact of the renal disease on the pregnancy should continue to be based on the absolute serum creatinine levels and coexisting morbidities. Subsequent monitoring of GFR during pregnancy5 should be based on creatinine clearance, which more closely agrees with true GFR. Although there is currently considerable interest in the potential use of novel serum makers, particularly cystatin C, these should only be applied in a research setting until the techniques are validated further and are more widely available.
In conclusion, MDRD significantly underestimates GFR during both healthy and pre-eclamptic pregnancies. It is less accurate than creatinine clearance and is not sensitive enough to use as a screening test for renal disease in this population. MDRD cannot be recommended as a useful clinical tool in obstetric practice.
Portions of this study were presented in abstract form at the 11th Annual Conference of the British Maternal and Fetal Medicine Society (Cardiff April 2006) and published in abstract form in the proceedings of that meeting (Journal of Obstetrics and Gynaecology 26 [Supplement 1] S47, 2006). We gratefully acknowledge the laboratory assistance of Elizabeth A. Shiells and Maureen Kirkley. The work was supported by Northern Counties Kidney Research Fund.