Screening for renal disease using serum creatinine: who are we missing?

Abstract

Background. Appropriate management and timely referral of patients with early renal disease often depend on the identification of renal insufficiency by primary care physicians. Serum creatinine (SCr) levels are frequently used as a screening test for renal dysfunction; however, patients can have significantly decreased glomerular filtration rates (GFR) with normal range SCr values, making the recognition of renal dysfunction more difficult. This study was designed to estimate the prevalence of patients who have significantly reduced GFR as calculated by the Cockcroft–Gault (C‐G) formula, but normal‐range SCr.

Methods. The study included 2781 outpatients referred by community physicians to an urban laboratory network for SCr measurement. GFR was estimated using the C‐G formula. Patients were grouped according to the concordance of SCr level abnormalities (abnormal >130 μmol/l) with significantly abnormal C‐G values (abnormal ⩽50 ml/min). The C‐G value of ⩽50 ml/min was chosen to reflect substantial renal impairment in all age groups. A further analysis of historical laboratory data was undertaken to determine if there were previously documented changes in renal function parameters in those patients who had overt renal dysfunction during the study period.

Results. Of the 2781 outpatients referred, 2543 (91.4%) had normal SCr levels. Of these patients, 387/2543 (15.2%) had C‐G calculated GFR ⩽50 ml/min, representing substantially impaired renal function. Among patients with normal SCr, abnormal C‐G values were identified in 47.3% ⩾70 years old, 12.6% 60–69 years old, and 1.2% 40–59 years old. Analysis of historical available laboratory data for patients with abnormal SCr and abnormal C‐G values showed that 2 years prior to the study period, 72% of this group had abnormal SCr, while 18% had normal SCr with abnormal C‐G values, and 10% had normal SCr with normal C‐G values.

Conclusions. This study documents the substantial prevalence of significantly abnormal renal function among patients identified by laboratories as having normal‐range SCr. Including calculated estimates of GFR in routine laboratory reporting may help to facilitate the early identification of patients with renal impairment.

Introduction

Early identification and appropriate nephrological management of patients with mild renal disease has been increasingly recognized as an important opportunity to delay the progression of renal disease and modify risk factors for comorbid diseases [1–5]. However, early recognition of renal disease and timely referral to nephrologists depend on identification of impaired renal function by primary care physicians. As most patients with mild renal disease are asymptomatic, early detection usually results from an incidental finding on laboratory testing, or routine screening tests targeted at high‐risk groups, including patients with diabetes or hypertension.

Many primary care physicians rely on serum creatinine (SCr) as a screening test for renal impairment; however, SCr levels can remain within the normal range even when renal function is significantly impaired [6]. Although not routinely reported by laboratories, a more accurate approximation of renal function can be obtained using formulae, such as the Cockcroft–Gault (C‐G) equation [7–10], to calculate an estimated glomerular filtration rate (GFR) from SCr and routine clinical data.

In this study we identified a large outpatient cohort who had had SCr concentrations requested by community physicians. The study was designed to determine the prevalence of patients who had significantly abnormal GFR values (i.e. below any expected age‐related declines) as calculated by the C‐G equation, but SCr levels within the normal range, which may make the recognition of renal dysfunction more difficult.

Subjects and methods

Patients

This cross‐sectional study included 2781 outpatients from a large urban area in British Columbia. The study included all patients aged 16 years or older who had had a SCr level ordered by a community physician and measured at the facilities of a single laboratory network during a 14‐day period (2–16 July 1998). Patient weights were measured by laboratory staff; genders and birth dates were obtained from health cards. SCr values from the July 1998 visit were used for the primary analysis. Retrospective SCr concentrations were obtained from the laboratory database for previous visits to the laboratory network in the last 4 years. If more than one SCr value was recorded per year, the value closest to year‐end was taken to be representative of that year. We screened 2879 patients, and after linkage with the provincial renal databases, patients known to have had a renal transplant or who received dialysis therapy were discarded from the analysis (n=98). Data from physician billing numbers showed that 96% of SCr levels in the study population were ordered by family physicians and general internists. The clinical indications for ordering SCr levels were not available. Additional laboratory testing such as urinalysis was not ordered for most patients and therefore was not included in the analysis. Information regarding the racial background of patients was also not available; however, the local population is primarily Caucasian and Oriental. All data were maintained anonymously.

Laboratory data

SCr levels were measured using the Boehringer–Mannheim Jaffe method on a Hitachi 727 autoanalyser. The normal range for SCr in the laboratory was 30–130 μmol/l. These limits represent the 97.5th percentile of the distribution of SCr levels in healthy male and female volunteers, and were determined using standard laboratory protocols.

Calculations

The C‐G formula: (140 −age) (weight in kg)/(SCr×0.81) with a gender correction factor of 0.85 for females, was used to estimate GFR. A stable weight was assumed for calculation of retrospective GFR. Abnormal renal function was very conservatively defined as a calculated GFR ⩽50 ml/min for the purposes of analysis.

Statistical analysis

Patients were divided into four groups based on the concordance of abnormal and normal SCr levels with abnormal and normal GFR as calculated by the C‐G formula. The four groups were: normal SCr/normal C‐G (overt normal renal function); abnormal SCr/normal C‐G (covert normal renal function); normal SCr/abnormal C‐G (covert renal dysfunction); abnormal SCr/abnormal C‐G (overt renal dysfunction). The group with covert renal dysfunction was the group of primary interest. Descriptive analysis of the population based on age, gender, and laboratory values was performed. Chi square and t‐tests were performed on categorical and continuous variables as appropriate.

Results

Of the 2781 patients in this outpatient population sent for SCr measurement by community physicians, 2156 (77.5%) had normal SCr concentrations and a normal GFR calculated by the C‐G equation, while 182 (6.5%) had overtly abnormal renal function with abnormal SCr and C‐G values (Table 1). Interestingly, 443 patients (15.9%) demonstrated discordance between SCr and C‐G values, with 387 (13.9%) having normal SCr levels, but C‐G values ⩽50 ml/min (covert renal dysfunction). Comparison of the covert renal dysfunction group with the normal SCr/normal C‐G group showed statistical difference in SCr levels (99±16 vs 90±14, P<0.001) and C‐G values (41±7 vs 85±26, P<0.001).

The demographic characteristics of the patients grouped by concordance of SCr and C‐G values are shown in Table 1. The discordance between SCr and C‐G‐calculated GFR values was most pronounced in the older age groups (Table 2), with 47.3% of patients 70 years or older with normal SCr levels having abnormal C‐G values. However, abnormal C‐G values were also found in 1.2% of patients in the 40–59 year‐old age group with normal SCr levels.

Historical data

Analysis of historical laboratory data was undertaken to characterize specifically the ‘renal history’ of patients with overt renal dysfunction (abnormal SCr and abnormal C‐G) in 1998. The purpose of this analysis was to determine if these patients had previous abnormalities in renal function, including abnormal C‐G values with normal SCr levels, which may have heralded the development of overt kidney dysfunction evident in 1998.

Two‐year historical data (1996) was available for 55% of patients (n=100) with overt renal dysfunction in 1998. In 1996, 72% of this group (72 of 100 patients) had abnormal SCr, while 28% (28 of 100) had normal SCr (Table 3). Of the 28% (n=28) with normal SCr in 1996, 36% (10 of 28) had normal C‐G values, while 64% (18 of 28) had abnormal C‐G values, as defined as ⩽50 ml/min. Thus, over half of the group with normal SCr had abnormal renal function 2 years prior to having abnormal SCr in 1998. Differences in SCr between patients with normal renal function and patients with abnormal renal function (with normal range SCr), are statistically significant (104±15 μmol/l vs 119±9 μmol/l, P=0.003) (Table 4). Similar results were obtained with historical data from 1994, 1995 and 1997 (Tables 3 and 4).

Discussion

Recently there has been increasing emphasis on appropriate treatment and timely referral of patients with early renal disease [1–4]. Serum creatinine is an important screening test for evaluating renal function; however, in this study we found that 13.9% of patients with normal range SCr levels had substantially abnormal calculated GFR, with C‐G values ⩽50 ml/min. The significant renal impairment in this group of patients may remain unrecognized by primary care physicians who rely on SCr abnormalities to identify renal insufficiency.

For this study, we chose a very conservative cut‐off for abnormal renal function of ⩽50 ml/min to reduce the likelihood of elderly patients being erroneously classified has having abnormal renal function. It has been suggested that renal function decreases over time as part of the normal ageing process [11,12]. However, even with the mean decrease in creatinine clearance of 0.75 ml/min/year found with normal ageing in the Baltimore Longitudinal Study of Aging [11], a normal patient with a GFR of 120 ml/min at age 30 [13] should still have a GFR of >80 ml/min at age 80. Thus ‘abnormal’ renal function in this study (calculated GFR ⩽50 ml/min) is not likely to be solely attributable to normal ageing. Furthermore, there is some data to suggest that those patients with GFR <50 ml/min are at risk of progressive decline in renal function [14]. Thus, our categorization of abnormal renal function using this GFR cut off is rational on that basis as well.

There are limitations to this cross‐sectional study. Namely, we were unable to differentiate transient renal dysfunction from early chronic renal insufficiency in the current cohort. Thus, it is unclear what proportion of the group with normal SCr and abnormal C‐G values will actually progress to overt renal failure. In an attempt to gain some insight into this issue, we performed a historical review of laboratory data for patients with overt renal dysfunction at the time of the primary study period. In each of the previous 4 years, 10–20% of patients who eventually developed overt renal dysfunction had normal SCr, but C‐G values ⩽50 ml/min, representing over half of the patients with normal SCr at these time points. The data therefore suggest that a substantial number of the patients with overt renal dysfunction at the time of the study did indeed pass through a transient stage with abnormal C‐G values, but normal SCr levels. This is in keeping with known physiology in that substantial reductions in GFR are necessary before changes in serum creatinine are noted. Clearly, identification of patients in this earlier, transient stage of ‘covert renal dysfunction’ may represent an important opportunity for intervention.

A second shortcoming of our study is the lack of a ‘gold standard’ measurement for GFR, which, due to cost and complexity, is difficult to obtain in a large screening population. We chose the C‐G equation to estimate GFR because the equation is well known and requires only readily available laboratory and demographic variables. The C‐G formula has previously been shown to correlate well with [^99mTc]‐DTPA‐measured GFR over the GFR range of 14–100 ml/min, with a correlation coefficient of 1.01, r=0.92 [15]. Couchoud et al. [16] found that the C‐G formula somewhat underestimates inulin‐clearance‐measured GFR in the range of 60–80 ml/min; however, an inulin‐clearance cut‐off of 60 ml/min/1.73 m², which represents significant renal impairment, could be approximated by a C‐G value cut‐off of 54±5 ml/min, similar to the cut‐off used in this study. While a number of calculations exist to estimate renal function [9,10], none has received the widespread use or recognition of the C‐G formula. This is probably due to its ease of use in clinical practice, and reasonable performance in approximating renal function. For the purposes of population screening, this measure then is very useful.

Inexpensive, simple methods of estimating GFR are critical to heighten the awareness of primary care physicians to abnormalities in renal function that exist while SCr remains within the normal range. The current epidemic in the end‐stage renal disease population requires identification of patients with kidney disease at a time sufficient to ensure proactive care to delay progression or further damage. There has been recent additional focus on the need for early referral to nephrologists, but this is predicated on the assumption that early renal disease is identifiable by primary care physicians. Nephrologists can help to facilitate this process by emphasizing renal function rather than absolute SCr levels in discussion and guidelines regarding early renal disease. Laboratory reporting, which simplifies identification of abnormal renal function, is another important step towards the appropriate medical management of early progressive renal insufficiency.

References