Utility of Captopril Renal Scans for Detecting Renal Artery Stenosis

Background  Captopril renal scanning (CRS) is commonly recommended as a noninvasive method for detecting renal artery stenosis (RAS), based on performance characteristics determined in research settings. Scant data are available, however, regarding the utility of CRS in clinical practice.

Methods  We evaluated the performance characteristics (sensitivity, specificity, and predictive values) of CRS in a consecutive series of 90 patients who underwent both CRS and renal arteriography within a 6-month period (January 1, 1991, through December 31, 1995) at a university hospital.

Results  Among 86 eligible patients (and 169 kidneys), the prevalence of RAS was 43%. The sensitivity of CRS was 74% (95% confidence interval [CI], 62%-83%); the specificity was 59% (95% CI, 49%-69%); the positive predictive value was 58% (95% CI, 47%-68%); and the negative predictive value was 75% (95% CI, 64%-84%). Also, there was evidence of spectrum bias, because the sensitivity and specificity (as well as the positive and negative predictive values) were different for groups with and without vascular disease.

Conclusions  The results of CRS were substantially worse in a clinical practice setting than previously reported in research settings, despite a similar prevalence of RAS. Captopril renal scanning should not be used as an initial screening test for diagnosing RAS, even among patients with high clinical likelihood of disease.

RENOVASCULAR hypertension due to atherosclerotic renal artery stenosis (RAS) is one of the most common types of secondary hypertension in adults.^1-3 The natural history of RAS is progressive, and treatment options include surgical revascularization or angioplasty with stenting.^4-8 Current guidelines⁹ recommend using captopril renal scanning (CRS), duplex Doppler flow studies, or magnetic resonance angiography as initial noninvasive screening tests⁹ in patients who are suspected of having renovascular disease. Although no consensus exists regarding the single "best" test,^1,10-13 many clinicians rely on CRS as the preferred initial approach because of its purported role in identifying patients who would benefit (in terms of blood pressure control) from revascularization.^14-17

The proposed rationale for CRS as a diagnostic test for RAS relates to the physiologic effects of angiotensin II on renal autoregulation of glomerular filtration rate (GFR), involving differential effects on afferent and efferent arteriolar tone^5,10 as well as constriction of mesangial cells. Patients with afferent RAS are dependent on regulation of efferent arteriolar tone to maintain GFR. Angiotensin-converting enzyme inhibitors interfere with this angiotensin II–mediated vasoconstriction, resulting in a decrease in GFR. Accordingly, patients who demonstrate a change in GFR on their renal scan after administration of the angiotensin-converting enzyme inhibitor captopril are thought to represent cases in which the RAS is more likely to be hemodynamically significant.^14-17

Captopril renal scanning has been reported^14-17 to be highly sensitive (91%-94%) and highly specific (84%-95%) in detecting RAS, although some investigations have demonstrated less impressive results.^18,19 Informal observations from the Yale Hypertension and Referral Service suggested that the test might not perform as accurately in clinical practice as had been reported in an earlier study¹⁴ at the same institution. We therefore conducted the current study to determine how well CRS performed in actual clinical practice.

All patients at Yale–New Haven Hospital, New Haven, Conn, who underwent CRS and renal arteriography within a 6-month period (January 1, 1991, through December 31, 1995), were included. Patients were ineligible if they had undergone a revascularization procedure prior to the study period. If patients underwent subsequent CRS or renal arteriography more than once during the study period, only results from the first tests were used in data analysis.

Medical records, CRS reports and films, and arteriography reports were reviewed for all eligible patients. The CRS films were independently reviewed by a nuclear radiologist (H.D.) who was blinded to patient information, prior readings, and results of artiography. Arteriograms were reviewed if the original report did not characterize the percentage of stenosis for each renal artery; interpretation was performed by a vascular radiologist who was blinded to the prior reading and CRS results.

Using the same institutional protocol as that of an earlier study¹⁴ for performing and reporting of CRS, positive scintigraphic evidence of RAS required a time-to-peak activity of more than 11 minutes on either the precaptopril or the postcaptopril scan or a calculated glomerular filtration ratio of greater than 1.5 between the 2 kidneys on the postcaptopril scan.

Criterion of positive results of renal arteriography in patients with atherosclerotic disease included stenosis of more than 75% or stenosis of more than 50% with poststenotic dilitation. Fibromuscular lesions were graded as mild, moderate, or severe, and moderate or severe lesions were considered positive for hemodynamically significant RAS, according to criteria previously applied at our institution.¹⁴

Data abstracted from the medical records included age at time of study, sex, ethnicity, history of diabetes mellitus, history of elevated total cholesterol levels or of cholesterol-lowering medication use, physical examination findings of vascular disease, history of vascular disease, history of tobacco use, and most recent serum creatinine values (in relation to the date of the CRS). Physical examination findings of vascular disease required the absence of a femoral pulse or the presence of a femoral, carotid, flank, or abdominal bruit. History of vascular disease was defined as documented myocardial infarction, cerebrovascular disease, or peripheral vascular disease.

Data were analyzed for each kidney and each patient as the unit of observation. Sensitivity, specificity, and positive and negative predictive values (and corresponding 95% confidence intervals [CIs]) were calculated for all patients combined, as well as separately for patients with and without evidence of vascular disease. The results of arteriography were considered the "gold standard" measurements.

Ninety patients were identified from hospital databases as having undergone both CRS and renal arteriography during the study period. Among these patients, 1 medical record was not located; the original captopril scan was missing for another patient; and 2 patients underwent subsequent studies. Accordingly, a total of 86 patients (96%), representing 169 kidneys, were included in the analysis (3 of the 86 had previously undergone a nephrectomy).

Sixty patients (70%) had evidence of vascular disease at the time of study, and 26 did not. Patients with vascular disease were older, and more likely to be white, than patients without vascular disease (data not shown).

Table 1 shows the prevalence, sensitivity, specificity, and positive and negative predictive values of CRS and renal arteriography for all patients.No significant difference was found in the mean interval of time between CRS and renal arteriography among patients whose scanning results were false-negative and true-positive: 35.5 ± 54.8 (mean ± SD) days vs 47.6 ± 53.5 days, respectively.

Table 2 shows the prevalence, sensitivity, specificity, and positive and negative predictive values of CRS and renal arteriography for patients who had evidence of vascular disease at the time of the study. Table 3 shows the prevalence, sensitivity, specificity, and positive and negative predictive values of CRS and renal arteriography for patients who did not have evidence of vascular disease at the time of the study.

Although complete follow-up data on all patients were not available (and were not a major focus of the study), the medical records for a subset of patients did document whether angioplasty or surgical revascularization was performed in patients with arteriogram-documented RAS. Of the patients with false-negative CRS results (15 of 19 kidneys), 14 received revascularization. Of the patients with true-positive CRS results, 34 received revascularization (40 of 54 kidneys). Follow-up data on blood pressure were not available.

Data from the current investigation demonstrate that the diagnostic test characteristics of CRS in a practice setting at a large referral center are not as good as reported^14,15 for patients who were enrolled in research studies at the same institution. For example, a previous publication¹⁴ reported 91% sensitivity and 94% specificity among 94 patients. In that study, 44 patients (47%) had RAS, and the positive predictive value (93%) and the negative predictive value (92%) were quite impressive. The current study found a similar prevalence of disease—43% among 86 patients—but the lower sensitivity (79%) and specificity (59%) were associated with more modest positive (58%) and negative (75%) predictive values.

One factor contributing to the observed differences in results between the studies could be that patients in practice are less likely to be monitored as closely as patients in research studies; therefore, patient preparation for CRS might not be as rigorous. Differences in conducting the test could also have a significant effect, particularly with regard to use of "contraindicated" medications (eg, angiotensin-converting enzyme inhibitors and angiotensin II receptor antagonists), as well as attention to volume status at the time of study.

Progression of underlying atherosclerosis during the interval between CRS and renal arteriography could contribute to the poor CRS performance characteristics observed in our study, but only if there was greater progression of atherosclerosis during this interval among the group of patients with false-negative results on CRS than among the group of patients with true-positive results. Although we do not have serial arteriograms to compare, the mean number of days between CRS and renal arteriography was, on average, shorter in the group with false-negative results on CRS compared with the group with true-positive results: 38 days vs 48 days, respectively, making this explanation unlikely.

The group of patients with vascular disease (Table 2) most closely resembles the subjects in the earlier study,¹⁴ yet sensitivity, specificity, and predictive values for CRS were substantially lower in the current population. Also, these diagnostic test characteristics differed for patients with and without vascular disease. This observation is consistent with spectrum bias: a problem that occurs when a diagnostic test has different sensitivity or specificity in patients with different clinical manifestations of disease.²⁰ The most striking finding for clinical practice involves the contrast of positive predictive values: 70% (95% CI, 57%-81%) for subjects with vascular disease, and 31% (95% CI, 16%-51%) for subjects without vascular disease. These discordant values, and nonoverlapping CIs, suggest that CRS performs differently in the 2 populations.

The low specificity and poor positive predictive value among the group of patients with no evidence of vascular disease and a prevalence of RAS of 17% confirmed an a priori clinical impression regarding unnecessary angiograms associated with CRS. Although the sensitivity and negative predictive value of CRS in this group were both 100%, more patients had a false-positive test result (n = 20) than had a true-positive test result (n = 9); consequently, 20 patients who did not have RAS were exposed to the risks of arteriography.

The finding that most patients with positive results on arteriography received revascularization, regardless of the CRS result, suggests that CRS was not being used to determine whether to intervene surgically. Other factors, including clinical judgment, patient preference, and renal function, are certainly involved. Unfortunately, we did not have data to determine the predictive value of CRS on blood pressure outcomes for the patients who had undergone revascularization.

The original role of CRS^9,14-16 was as a noninvasive test for detecting RAS among patients with a high clinical likelihood of the disease. These patients were judged most likely to benefit from blood pressure control after a revascularization procedure, such as bypass surgery or renal angioplasty. Our data from a practice setting suggest that CRS is a poor test to "rule in" patients with a low clinical likelihood of disease and does not perform as well as reported among patients with a high clinical likelihood of RAS.

We suggest that CRS not be recommended as an initial screening test for diagnosing RAS and renovascular hypertension, even among patients with a high clinical likelihood of disease. Rather, the appropriate use of CRS may involve patients who have known RAS (ie, positive results on magnetic resonance angiography or renal arteriography), and for whom the physician is seeking additional information to support a decision about revascularization to improve blood pressure control. Even when CRS is used for this purpose, however, strict adherence to the test protocol (including the discontinuation of selected medications and maintaining proper volume status) is essential.

Accepted for publication February 6, 2002.

Dr Concato is supported by a Career Development Award from the Department of Veterans Affairs Health Services Research and Development Service, Washington, DC.