Several important, recent articles have prompted a reevaluation of screening tests for renal artery stenosis. More than 5 years ago, several relatively large, well-controlled studies evaluated the effectiveness of two tests in screening for renal artery stenosis in highly selected, high-prevalence groups: angiotensin-converting enzyme (ACE) inhibitor-induced stimulation of plasma renin activity and changes in the renal scintigram assessed by nuclear medicine techniques. Both tests probably detect a similar physiologic response to ACE inhibitors in functionally substantial renal artery stenosis, namely, a reduction in kidney glomerular filtration rate. An ACE inhibitor-induced decrease in glomerular filtration rate leads to a macula densa-stimulated increase in plasma renin activity (that is, a positive result on a plasma renin activity test after captopril therapy) and a reduction in the rate of clearance of radioactive tracer from the kidney (that is, a positive result on ACE inhibitor scintigraphy). Accordingly, it is not surprising that both tests, under ideal circumstances, yielded similar sensitivities (90% to 95%, with higher specificities for ACE inhibitor scintigraphy than for the plasma renin activity test [1]).

However, the past year has seen the publication of larger trials in which less selective groups of patients were studied. Rossi and colleagues [2] found that plasma renin activity was low or normal in one third of a large series of patients with renovascular hypertension (defined as patients with renal artery stenosis whose blood pressure decreased after correction of the stenosis). Most patients with low baseline plasma renin activity had negative results on renal vein renin tests; in addition, plasma renin activity did not increase consistently in these patients after they received captopril. Thus, if renin-based tests had been used during screening, about one third of patients who would have benefited from renal revascularization would not have undergone the procedure.

The accuracy of renal scintigraphy for diagnosing renal artery stenosis was recently reassessed in two large studies in the United States [3] and Europe [4]. The Einstein-Cornell Collaborative Hypertensive Group concluded that less than 5% of test results were false-positive; therefore, renal scintigraphy is useful for confirmation of the diagnosis. However, the accuracy of these scans for detection of renal artery stenosis was disappointing and averaged approximately 60%. This poor sensitivity was attributed to a large number of scans that were nondiagnostic because of an elevated serum creatinine level; the presence of a small, poorly functioning kidney; or bilateral renal artery stenosis. The large European study done by the pioneers of ACE inhibitor renography showed a similar sensitivity of approximately 70%. Such low sensitivities imply that these tests will miss many patients with renal artery stenosis. Clinicians must question the usefulness of these screening tests and look for new, noninvasive methods for detection of renal artery stenosis.

The article by Krijnen and coworkers [5] in this issue is therefore timely and provocative. The authors prospectively and systematically evaluated more than 400 patients with drug-resistant hypertension, all of whom underwent renal arteriography to determine whether they had renal artery stenosis (defined as a 50% narrowing of one or both renal arteries). The authors provided a statistical analysis of the association of readily available clinical criteria with renal artery stenosis and presented a regression model based on this analysis that discriminated between patients with renal artery stenosis and those with essential hypertension. They simplified the model to a clinical prediction rule that includes eight clinical criteria or routine biochemical variables that can be applied to patients at risk for renovascular hypertension. Clinicians can score these eight variables for a particular patient and derive a sum score to predict the probability of renal artery stenosis. If a 30% predicted probability of renal artery stenosis is considered to reasonably indicate renal arteriography, this clinical prediction rule would have a sensitivity of 68% and a specificity of 87%. When the authors performed renal scintigraphy on this set of patients, the scintigraphic test had a sensitivity of 72% and a specificity of 90% for detection of renal artery stenosis. Clearly, in these patients, simple clinical criteria, when carefully and quantitatively assessed, can provide as much information on the probability of detecting renal artery stenosis as renal scintigraphy, which is more complex, time-consuming, and expensive. This finding is provocative not only for those in nuclear medicine who use scintigraphy but also for practicing clinicians who must select the best set of methods to evaluate patients for renal artery stenosis.

Several limitations of this important and novel study should be considered before its carefully worded conclusions are applied to routine clinical practice. First, as with other tests for renovascular hypertension, the validity cannot be properly assessed until the true accuracy of the prediction model is determined independently and prospectively in a new group of patients who have had arteriography. By using internal bootstrapping analysis and separate analysis of data from different centers, Krijnen and colleagues have taken great pains to determine whether the data set is likely to be predictive, but this determination always requires independent and prospective assessment.

Second, the authors included patients with atherosclerosis and those with fibromuscular dysplasia, but the clinical characteristics of these two types of patients differ dramatically. Because atherosclerosis is five times more common than fibromuscular dysplasia, the clinical prediction rule is heavily biased in favor of detection of the former group. How would the prediction rule assess a woman with recent-onset hypertension, normal renal function, and no atherosclerotic or metabolic complications who developed hypertension in her late teens? The clinical prediction rule would rate her predictive probability of stenosis as very low (<5%). However, her characteristics match those of a patient with fibromuscular dysplasia, which is especially important to detect because it has a high rate of response to percutaneous transluminal renal angioplasty.

Third, the quantitative importance of the criteria will depend on the population studied. For example, renovascular hypertension is very uncommon in black persons in the United States. Because of the low percentage of black patients in this Dutch study, ethnicity does not feature in the clinical prediction rule; however, ethnicity will probably be an important component if such a rule is applied to a U.S. data set that has a more evenly mixed ethnic distribution.

Fourth, the detected variable in this study was renal artery stenosis. Renovascular hypertension is defined as hypertension that is improved or cured by reversal of stenosis. By no means do all patients with anatomic renal artery stenosis also have renovascular hypertension. Currently, no validated tests exist to determine whether renal artery stenosis is causing progressive loss of renal function (ischemic nephropathy). The clinically relevant question is usually, "Does my patient have renovascular hypertension?" not, "Does my patient have renal artery stenosis?" Further evaluation of this prediction rule is needed to determine whether the patients with renal artery stenosis who are identified are those who benefit significantly from angioplasty or surgery. Overall, less than one quarter of patients with atherosclerotic renal artery stenosis are cured of hypertension, a similar number derive no benefit from therapy, and some patients have an adverse response to the intervention. Therefore, detection of the anatomic defect of renal artery stenosis is only the first step in a comprehensive evaluation of patients for renovascular disease.

Recently, an approach similar to that of Krijnen and coworkers was used in patients with renal artery stenosis to answer a different clinical question: "If my patient with renal artery stenosis undergoes percutaneous transluminal renal angioplasty or reconstructive surgery, what is the probability of cure of hypertension?" In a retrospective analysis of 23 patients with fibromuscular dysplasia [6], half of the patients were cured of hypertension (cure was defined as normal blood pressure without antihypertensive medication 6 months after the intervention). Three clinical variables were found to be independently associated with cure: low systolic blood pressure before intervention, short duration of hypertension, and younger age. From this, a simple clinical prediction rule was developed that predicted the following probabilities of cure: less than 1% (for example, in a patient >40 years of age who had a 10-year history of hypertension and whose systolic blood pressure was >160 mm Hg) and almost 100% (for example, in a patient <30 years of age with a 6-month history of hypertension and a low preintervention systolic blood pressure). A similar analysis applied to 63 consecutive patients treated for atherosclerotic renal artery stenosis found that 21% were cured [7]. The greatest probability of cure in this group was found in men with a duration of hypertension less than 10 years and an initial diastolic blood pressure greater than 80 mm Hg. Of interest, lower diastolic blood pressure in this population predicted less chance of cure, probably because this group of patients with milder, easily controlled hypertension had functionally insignificant renal artery stenosis. Twelve of the 13 patients with atherosclerotic renal artery stenosis who were cured of hypertension were men. By applying a simple 2 x 2 Table to these three characteristics, the authors were able to develop a clinical prediction rule with a sensitivity of 92% and a specificity of 77% that predicted which patients with atherosclerotic renal artery disease would be cured of hypertension after intervention.

In this era of complex, high-technology testing, it is refreshing to find that when used in a systematic and quantitative manner, clinical findings and simple biochemical tests can perform as accurately as more complex tests in the detection of renal artery stenosis [5] or in the prediction of outcome after correction of renal artery stenosis [6, 7]. Further prospective clinical trials are needed to determine the place of these clinical prediction rules in the evaluation of patients suspected of having renovascular disease. By helping physicians to think more quantitatively about what can be expected for their patients, these clinical prediction rules should provide a more informed answer to the question of who should be evaluated and what benefit might be expected from an intervention. The study by Krijnen and colleagues should be seen as an important addition to the workup of this elusive but curable form of hypertension.