The assessment of kidney function by general practitioners in Australian patients with type 2 diabetes (NEFRON-2)

The National Evaluation of the Frequency of Renal Impairment cO-existing with Non-insulin dependent diabetes mellitus (NEFRON) study was an incident-driven, clinic-based clustered and stratified survey of patients with type 2 diabetes in Australian primary care.4 Investigator selection and representation of Australian general practice is described elsewhere.4 In brief, expressions of interest were invited from all registered GPs across Australia, and investigators were then randomly selected (n = 500) proportional to the census population.5 The selected NEFRON investigators were requested to recruit 10–15 consecutively presenting adult patients with type 2 diabetes, irrespective of the reason for the consultation.

Data collection took place between April and September 2005. A de-identified case report form was completed for each eligible patient. Case report forms captured demographic information, results of physical examination and the most recent laboratory test results, including serum creatinine level and urinary albumin to creatinine ratio.6,7 At the time of the study, most GPs had access to resources that enabled the estimation of kidney function using the formula proposed by Cockcroft and Gault. This formula incorporates measures of serum creatinine, age, sex and weight to estimate creatinine clearance.8 Consequently, for the purposes of our study, creatinine clearance for each patient was also estimated using the Cockcroft–Gault formula from data recorded in the case report forms. No attempt was made to standardise results from different laboratories or regions, but rather to reflect the “raw” (unstandardised) results on which practitioners currently base their assessment and management.

On the same case report forms for each patient, GPs were asked whether they routinely estimated kidney function or whether their clinical laboratory provided an estimation of kidney function. On the basis of each patient’s current status, GPs were asked to state whether kidney function was impaired. Finally, GPs were also requested to provide their best estimate of their patients’ kidney function in mL/min. To reflect resources available to GPs, no instruction was provided on how kidney function should be estimated or defined.

The Royal Australian College of General Practitioners (RACGP) National Research and Evaluation Ethics Committee gave approval for the study. All participating patients provided written informed consent to take part in the study.

Subanalyses for nominal variables consisted of either one-way analysis of variance (ANOVA) for single variables or two-way ANOVA for comparison of three groups. Subanalyses for categorical variables involved Pearson χ2 analysis of proportions between independent parameters. Because of incomplete data recording, different patient numbers are presented for different endpoints. Missing variables were not imputed. The predictive association between patients’ variables and the perception of impaired kidney function by their GP was assessed by the area under the receiver operating characteristic curve (ROC curve) using SAS software, version 9.1 (SAS Institute Inc, Cary, NC, USA).

At the time of this survey, an estimation of kidney function was provided by the local laboratory for 2.4% of patients (n = 92/3783). The mean creatinine clearance in these patients was similar to that in patients whose kidney function was not reported by a laboratory (77 v 78 mL/min). Seventy-two per cent of the results estimated by laboratories were above 60 mL/min, a level at which the accuracy of predictions of kidney function are suboptimal and thus not currently recommended.3

At the time of this survey, one in four patients (24.0%, n = 904/3766) with type 2 diabetes had their kidney function routinely estimated by their GP. The clinical characteristics of these patients are detailed in Box 1. The likelihood of patients having their kidney function estimated by their GP was not influenced by the patient’s age, sex, or duration of diabetes (Box 1), although these are all key risk factors for impaired kidney function.4 Patients with a urinary albumin to creatinine ratio in the macroalbuminuric range were marginally more likely to have routine estimation of their kidney function (P < 0.01) by their GP; however, only 30% of patients with a urinary albumin to creatinine ratio in the macroalbuminuric range had their kidney function routinely estimated by their GP. Patients receiving metformin or sulfonylureas with active metabolites were as likely as patients not receiving these medications to have their kidney function routinely estimated by their GP (24% and 26%, respectively), despite the risk of side effects of these drugs in patients with impaired kidney function.

An estimation of kidney function was recorded by GPs in 85% of patients (n = 3347/3893). There was a good statistical correlation between practitioner-estimated values and those derived from data on case report forms (R², 0.72), even in patients in whom kidney function was not routinely estimated (R², 0.70). This meant that, when prompted, practitioners were able to identify individuals with a creatinine clearance < 60 mL/min in over 83% of cases, with a specificity of 90%.

Practitioners were also invited to record whether, on the basis of current results, their patients had impaired kidney function (99%, n = 3868/3893). Impaired kidney function was reported to be present in 34.4% of men and 36.4% of women (P = 0.2). These figures do not accord with kidney function estimations derived from data on the case report forms using the Cockcroft–Gault formula (Box 2A), or with the level of kidney function estimated by the GPs themselves (Box 2B). In about half of all individuals categorised by GPs as having impaired kidney function, kidney function estimates using the Cockcroft–Gault formula (52%) or by the GPs themselves (49%) were < 60 mL/min.

When kidney function was recorded by GPs as < 60 mL/min, impaired kidney function was perceived to be present in 74% of patients. Those patients in whom GPs estimated kidney function at < 60 mL/min, but in whom impaired kidney function was not perceived to be present (Box 2), were less likely to have an elevated urinary albumin to creatinine ratio (30%), compared with those in whom impaired kidney function was considered to be present (63%, P < 0.01). A documented history of CKD, macrovascular disease or hypertension was also less common in these individuals (all P < 0.01). The presence of impaired kidney function as perceived by the GP made it marginally more likely that kidney function would be routinely estimated (30% v 23%, P < 0.01).

In the NEFRON study, men and women with creatinine levels over 120 μmol/L and 110 μmol/L, respectively, were perceived by their GP to have impaired kidney function in 95% of cases (Box 2C). Indeed, in determining the perception of impaired kidney function, the raw (unstandardised) creatinine level was as strong a determinant as creatinine clearance estimated by the GP (area under the ROC curve, 0.82 v 0.81, P = 0.9). Serum creatinine levels were also more likely to be above these levels in male and female patients correctly perceived as having impaired kidney function (50%) than in those in whom GP-estimated kidney function — < 60 mL/min — was not seen as commensurate with impaired kidney function (19%, P < 0.001). However, the value of the serum creatinine level in distinguishing normal from abnormal function in individual patients is problematic, as individuals with a serum creatinine level in the normal range may have abnormal creatinine clearance. Indeed, in the NEFRON study, 55.6% of all those with a creatinine clearance < 60 mL/min had a serum creatinine level < 120 μmol/L in men or < 110 μmol/L in women (Box 3).