Flowchart of the study design and results. C-G GFR indicates glomerular filtration rate estimated by the Cockcroft and Gault formula.
Comparison of serum creatinine level in the 55 men and women with severe renal failure (Cockcroft and Gault formula glomerular filtration rate [C-G GFR], ≤30 mL/min). Women had significantly lower mean serum creatinine values than men, despite a similar level of renal function. This is likely because of the smaller muscle mass of women compared with men, as suggested by the women's significantly lower mean weight. The asterisk indicates P<.005. To convert creatinine to milligrams per deciliter, divide by 88.4.
Investigation for causes and complications of severe renal failure (Cockcroft and Gault formula glomerular filtration rate, ≤30 mL/min) in the 40 nonreferred patients.
Swedko PJ, Clark HD, Paramsothy K, Akbari A. Serum Creatinine Is an Inadequate Screening Test for Renal Failure in Elderly Patients. Arch Intern Med. 2003;163(3):356-360. doi:10.1001/archinte.163.3.356
Serum creatinine is the most commonly used screening test for renal failure. We hypothesized that serum creatinine would underestimate the degree of renal failure in elderly people because they have a reduced muscle mass. If so, this would lead to underrecognition and thus suboptimal care of patients with severe renal failure.
We conducted a retrospective medical record review of all patients 65 years or older in an outpatient academic family medicine practice. The glomerular filtration rate was calculated using the Cockcroft and Gault formula and was used to evaluate the testing characteristics of serum creatinine for the detection of renal failure.
We screened 1510 patients, 854 (56.6%) of whom met the inclusion criteria. Renal failure (glomerular filtration rate, ≤50 mL/min) was present in 28.9% of the patients, and severe renal failure (Cockcroft and Gault formula glomerular filtration rate, ≤30 mL/min) was present in 6.4%. A serum creatinine level of greater than 1.7 mg/dL (>150 µmol/L) had a sensitivity of 12.6% and a specificity of 99.9% for the detection of renal failure. For the detection of severe renal failure, the sensitivity was 45.5%, with a 99.1% specificity. Only 15 (27.3%) of the 55 patients with severe renal failure were referred to a nephrologist. Moreover, 34 (85%) of the 40 nonreferred patients with severe renal failure were incompletely evaluated regarding the metabolic complications associated with kidney dysfunction.
Serum creatinine is a poor screening test for renal failure in elderly patients, leading to marked underinvestigation and underrecognition of renal failure in this population.
THE PREVALENCE of end-stage renal disease (ESRD) is increasing at an alarming rate. In Canada, the number of patients receiving renal replacement therapy was 17 807 as of December 31, 1996, and is projected to be 32952 by 2006.1 Data from the United States predict that the 304 083 patients with ESRD registered in 1997 will have doubled by 2010.2 Elderly patients have the highest incidence rates of ESRD, and their proportion of the population undergoing long-term dialysis is increasing.3,4 Early referral to a nephrologist, or to a multidisciplinary team specializing in ESRD care, has been associated with a reduction in health care costs, morbidity, and mortality in patients starting renal replacement therapy.5- 10 Such timely referral requires early detection of kidney disease, and in this respect the most common screening test for renal dysfunction is the serum creatinine.11 Recently published clinical practice guidelines12 acknowledge these facts by addressing the evaluation and referral of patients with elevated levels of serum creatinine.
Unfortunately, serum creatinine alone may be misleading when evaluating renal dysfunction. For instance, patients with a relatively lower muscle mass, such as women and/or elderly persons, can have a serum creatinine level within the laboratory-reported normal range, although their renal function is severely compromised.11 More accurate estimates of renal function can be obtained in clinical practice by measuring creatinine clearance from a timed urine collection or by using formulas, such as the Cockcroft and Gault formula (C-G),13 to calculate the glomerular filtration rate (GFR).11
We hypothesized that renal failure is not being detected in many elderly patients because their serum creatinine levels fall within or only slightly above the laboratory-reported normal range, and that these patients do not undergo an appropriate evaluation or receive referral in a timely manner. A study was, therefore, designed to evaluate the detection and management of renal failure in ambulatory elderly patients within the primary care setting.
This study, approved by the Research Ethics Board of The Ottawa Hospital, is a retrospective medical record review, with data collected between August 1 and November 1, 2000.
All patients 65 years or older in an academic family medicine practice associated with The Ottawa Hospital, a tertiary care teaching hospital, were identified using a computer-generated list obtained from the main practice database. Patients were excluded from the study for the following reasons: (1) a serum creatinine level had never been documented in the medical record, (2) the patient's weight had never been documented in the medical record, (3) the elapsed time between the measurement of the serum creatinine level and the measurement of the patient's weight was longer than 10 years, (4) the patient was undergoing hemodialysis or peritoneal dialysis at the time of review, or (5) the medical record was not available for review during the study period.
The following data were recorded for each patient: age, sex, most recent serum creatinine level, and weight. The C-G was used to calculate the GFR.13 This formula has been validated in elderly populations to accurately predict the measured creatinine clearance and the glomerular filtration rate.14- 16 The C-G GFR has also been superior to measured creatinine clearance as a predictor of glomerular filtration rate in several studies.17- 19
Renal failure was defined as a C-G GFR of 50 mL/min or less. This value was chosen to represent a decreased GFR that would not be attributable to normal aging alone.20 Severe renal failure was defined as a C-G GFR of 30 mL/min or less. At this level of renal function, prompt referral to a nephrologist for ongoing management has proved beneficial.5- 10,12
Additional data were sought in the severe renal failure category: (1) blood pressure, (2) history of diabetes mellitus, and (3) whether the patient was referred to a nephrologist. We also determined whether these patients underwent the following investigations: the determination of hemoglobin, serum bicarbonate, serum calcium, serum phosphate, and serum albumin levels; chemical urinalysis; measured creatinine clearance; and abdominal ultrasonography.
The primary outcome evaluated was the value of serum creatinine as a test for renal failure in an elderly primary care population. We selected a serum creatinine level of 1.7 mg/dL (150 µmol/L) as a clinically relevant cutoff value, and any higher value was considered a positive test result for renal failure. This conservative cutoff value was selected based on previous studies,21,22 which demonstrated that family physicians were unlikely to refer patients to a nephrologist when the serum creatinine level was 1.7 mg/dL or less. The serum creatinine level was compared with the C-G GFR to determine its characteristics as a test for renal failure.
Secondary outcomes were the prevalence of renal failure in this population, the pattern of nephrology consultation for patients with severe renal failure, and the pattern of investigation and management of causes and complications of severe renal failure.
Results are presented as mean ± SD. Statistical analysis was performed using Statistical Product and Service Solution software, version 10 (SPSS Inc, Chicago, Ill). The t test for continuous variables and the χ2 test for categorical variables were used where indicated. Logistic regression was performed to identify characteristics that predicted referral to a nephrologist. P<.05 was considered statistically significant.
There were 1510 patients 65 years or older with active medical records, 1315 of which were available for review. Of these records, 461 were excluded (277 did not have a documented creatinine level, and 184 had a documented serum creatinine level but did not have a documented weight), leaving a study population of 854 patients (Figure 1). The mean age and the male-female ratio were not significantly different in the included and excluded patients (P = .16 for age; P = .09 for sex; data not shown).
The characteristics of the 854 patients are summarized in Table 1. There were 380 men (44.5%) and 474 women (55.5%). Women had a significantly lower C-G GFR than men, despite having a lower serum creatinine level.
Renal failure (C-G GFR ≤50 mL/min) was present in 28.9% of the patients, while severe renal failure (C-G GFR ≤30 mL/min) was present in 6.4% of the patients (Figure 1). The prevalence of renal failure was significantly higher in women than in men (164 [34.6%] of 474 women vs 83 [21.8%] of 380 men; P<.001). There was also a higher prevalence of severe renal failure in women than in men, but the increase did not reach statistical significance (37 [7.8%] of 474 women vs 18 [4.7%] of 380 men; P = .10).
For the detection of renal failure, a serum creatinine level of 1.7 mg/dL had an overall sensitivity of only 12.6% (Table 2). In fact, 87.4% of the patients with renal failure had serum creatinine values of 1.7 mg/dL or lower. The serum creatinine performed better in the detection of severe renal failure; however, the sensitivity was only 45.5% (Table 2). Also, the test was inconsistent across sex. Interestingly, although the mean serum creatinine level of women with severe renal failure was significantly lower than that of men, the C-G GFR was similar (Figure 2). Of the 55 patients with severe renal failure, 30 (54.5%) had serum creatinine values of 1.7 mg/dL or less.
Only 15 (27.3%) of the 55 patients with severe renal failure were referred to a nephrologist. Patients with higher serum creatinine levels were significantly more likely to be referred to a nephrologist; the odds of being referred increased by 2.4 for each 0.23-mg/dL (20-µmol/L) increase in serum creatinine level (P = .001). Not a single patient with severe renal failure and a serum creatinine value of 1.7 mg/dL or less was referred to a nephrologist. Men were more likely to be referred (odds ratio, 10.0; 95% confidence interval, 2.5-40.3; P = .001), as were diabetic patients (odds ratio, 5.3; 95% confidence interval, 1.4-19.8; P = .01), although the referral rate for diabetic patients was still only 50% (8 patients referred of 16 patients with diabetes mellitus).
The data from the 40 patients with severe renal failure who were not referred to a nephrologist were analyzed further, to examine the management of severe renal failure in the primary care setting (Figure 3). Of these 40 nonreferred patients, 27 (67.5%) underwent a urinalysis, 14 (35%) underwent abdominal ultrasonography, and only 7 (17.5%) had a 24-hour urine collection for creatinine clearance. Hemoglobin was measured in all 40 (100%) of these patients, serum calcium level in 14 (35%), serum bicarbonate level in 18 (45%), serum albumin level in 11 (27.5%), and serum phosphate level in 3 (7.5%). Anemia (defined as a hemoglobin level <13.0 g/dL in men and <11.5 g/dL in women) was present in 3 (37.5%) of 8 men and 6 (18.8%) of 32 women who were not referred.
The mean blood pressure values for all 55 patients with severe renal failure were 144 ± 21 mm Hg (systolic) and 74 ± 9 mm Hg (diastolic). There were no significant differences between the mean blood pressure values of referred and nonreferred patients (145/73 vs 143/74 mm Hg; P = .76 for systolic; P = .51 for diastolic). Suboptimal blood pressure control, defined as a mean blood pressure higher than 130/80 mm Hg,23 was noted in 43 (78.2%) of the patients with severe renal failure.
These data convincingly demonstrate the poor sensitivity of serum creatinine alone for the detection of renal failure in ambulatory elderly patients. We also observed that family physicians tended not to investigate or refer many patients with severe renal failure, underscoring the problems of relying on this poorly sensitive test. Of further interest was the high prevalence of renal failure in our study population, with an intriguing female preponderance.
We based our sensitivity analysis on a serum creatinine level of 1.7 mg/dL, although the selection of a lower cutoff value would have increased the sensitivity of the test for the detection of renal failure. The goal of this study was to evaluate the performance of the serum creatinine level in its clinical context, as primary care practitioners are applying it. This cutoff value was chosen based on 2 previous studies, by Mendelssohn and colleagues21 and by Wilson and colleagues.22 They found that physicians were most likely to refer a patient to a nephrologist when the serum creatinine level was greater than 3.39 mg/dL (300 µmol/L), and tended not to refer patients whose serum creatinine level was 1.35 to 1.69 mg/dL (120-150 µmol/L).
The investigation for causes and complications of severe renal failure by the family physicians in our study was poor. Only 15 (27.3%) of the 55 patients with severe renal failure were referred to a nephrologist, and few of the nonreferred patients were investigated for the metabolic complications of renal failure, suggesting that the renal failure was not recognized. One of the major determinants of referral in patients with severe renal failure (C-G GFR, ≤30 mL/min) was the serum creatinine level; the odds of nephrology referral increased significantly with higher serum creatinine values, yet there were no nephrology referrals made for patients with serum creatinine levels of 1.7 mg/dL or less. The family physicians seemed to base their management decisions on the serum creatinine levels irrespective of the patients' true renal function.
Other researchers have commented on the risks of using serum creatinine alone to detect renal disease in elderly patients. Duncan and colleagues20 used a cutoff value of 1.47 mg/dL (130 µmol/L) or greater to define an elevated serum creatinine level, and found that 47.3% of those 70 years and older with a "normal" serum creatinine level had a C-G GFR of 50 mL/min or less. A similar study by Papaioannou and colleagues24 focused on patients 65 years and older living in long-term care facilities, and found that their mean C-G GFR was only 43 mL/min while their mean serum creatinine level was 0.97 mg/dL (86 µmol/L), within the normal range. Our study confirms and extends these observations by demonstrating the underinvestigation and lack of referral of elderly patients with renal disease. This is all the more important given the increasing incidence of ESRD in this population.
Of interest was the high prevalence of renal failure in elderly women. Papaioannou and colleagues,24 who used the C-G GFR and not the serum creatinine to detect renal failure, found that the prevalence of a C-G GFR of 50 mL/min or less was 44% in men and 69% in women 65 years or older. These results conflict with earlier studies4,25,26 demonstrating a much lower overall prevalence of renal failure, with a male preponderance. Each of these earlier studies used the serum creatinine level to detect renal failure in their respective study populations. Interestingly, we found that the serum creatinine level was much less sensitive for the detection of renal failure in women than it was in men. The true prevalence of renal failure in women was likely underestimated in previous studies because the serum creatinine level, and not the glomerular filtration rate, was used as a marker of kidney function.
We observed that elderly women with severe renal failure were significantly less likely to be referred to a nephrologist. In Canada and the United States, more men than women start dialysis each year.27,28 Several researchers29- 32 have explored the existence of sex bias in access to dialysis and transplantation, but the potential sources of this bias are not well delineated. It is conceivable that renal failure is being underrecognized in elderly women because of the use of the serum creatinine level and, therefore, these women are not referred for dialysis or transplantation. Clearly, this is an area that requires further study.
One limitation of this study is the failure to use a gold standard measurement of glomerular filtration rate, such as inulin clearance, a limitation predicated, among other things, by the retrospective nature of the study. We, therefore, selected the C-G GFR as our standard measurement of renal function. The Cockcroft and Gault formula estimates glomerular filtration rate from the age, weight, sex, and serum creatinine level of the patient, and has the advantage of being easily applied in the clinical setting.13 It has been validated in elderly patients, and accurately predicts the measured creatinine clearance and the glomerular filtration rate.14- 16 Some researchers33,34 dispute the use of the C-G GFR in elderly patients, based on a few negative study results; these studies involve few patients and use the 24-hour urine collection for creatinine clearance as the reference standard. The C-G GFR has been shown to be superior to 24-hour urine measured for creatinine clearance as a predictor of glomerular filtration rate in several studies.17- 19
We conclude that serum creatinine, as it is being used by primary care practitioners to screen geriatric populations, is a seriously flawed test. This leads to a lack of appropriate investigation and referral of many elderly patients with renal failure, which may increase their risk of morbidity and mortality. The estimation or measurement of glomerular filtration rate should be the preferred screening method for the detection of renal failure in elderly patients.
Corresponding author and reprints: Ayub Akbari, MD, Division of Nephrology, The Ottawa Hospital, Riverside Campus, 1967 Riverside Dr, Ottawa, Ontario, Canada K1H 7W9 (e-mail: firstname.lastname@example.org).
Accepted for publication June 14, 2002.
This study was supported by a grant from Ortho Biotech Inc, Raritan, NJ.
We thank Kevin D. Burns, MD, and Marshall D. Lindheimer, MD, for their expert advice, and the physicians of the Family Medicine Center, The Ottawa Hospital, Civic Campus, Ottawa, Ontario, for their help.