To diagnose renal artery stenosis efficiently, angiography should be used selectively. Most physicians rely on captopril renal scintigraphy as a selection criterion, but the diagnostic accuracy of this test is low (sensitivity, 65% to 77%; specificity, 90%) [7, 8]. As an alternative, clinical characteristics can be used to select hypertensive patients for angiography [9]. Patients with normal renal function whose blood pressure can be controlled with one or two drugs can be excluded from angiography [9, 10]. In the remaining patients (those with drug-resistant hypertension), such clinical characteristics as atherosclerotic vascular disease, smoking history, and presence of an abdominal bruit can be used to estimate a patient's probability of renal artery stenosis [11-14]. This estimate can then be used in selection for angiography.

We analyzed the clinical characteristics of 477 patients with drug-resistant hypertension or an increase in serum creatinine concentration during therapy with angiotensin-converting enzyme (ACE) inhibitors who participated in the Dutch Renal Artery Stenosis Intervention Cooperative (DRASTIC) study [9]. We developed a clinical prediction rule for quantifying the probability of renal artery stenosis [15] and demonstrated the potential consequences of this rule for clinical practice by applying it to our patients.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

The DRASTIC study is a prospective cohort study conducted at 26 departments of internal medicine with an interest in hypertension throughout the Netherlands [9]. The diagnostic phase of the study was designed to find an optimal strategy for diagnosing renal artery stenosis. In the DRASTIC study, 1133 hypertensive patients 18 to 75 years of age with preserved renal function (serum creatinine concentration 200 µmol/L [2.26 mg/dL]) were enrolled. These patients were referred for analysis of hypertension by general practitioners (55%) or hospital specialists (45%), in most cases because their hypertension was difficult to treat with antihypertensive drugs. Sixty percent of patients were from four hospitals. After giving written informed consent, patients were randomly assigned to one of two standard protocols with antihypertensive drugs: amlodipine, 10 mg, plus atenolol, 50 mg, in patients older than 40 years of age or enalapril, 20 mg, plus hydrochlorothiazide, 25 mg, in patients older than 40 years of age. Blood pressure was measured with a standard sphygmomanometer at three consecutive visits at least 1 week apart. Measurements were taken three times per visit after a 5-minute rest with the patient in the sitting position. Patients were selected for diagnostic workup if they had drug-resistant hypertension, defined as a mean diastolic blood pressure per visit of 95 mm Hg or more while receiving the standard drug regimen during all three visits or prescription of an additional drug regardless of blood pressure response. Patients were also selected if the serum creatinine concentration increased 20 µmol/L (0.23 mg/dL) or more during therapy with ACE inhibitors. In these patients, intra-arterial digital subtraction angiography and other, noninvasive tests were performed. In accordance with the study protocol, patients who responded well to standard treatment were not evaluated further. The diagnostic phase of the study was followed by a therapeutic phase in which patients with atherosclerotic stenosis were randomly assigned to receive medication or renal angioplasty.

Definitions

After performing a literature study, we selected 12 clinical characteristics indicative of renovascular disease (predictors) [10, 11, 16-26]: age, sex, ethnicity (black or other), signs and symptoms of atherosclerotic vascular disease (femoral or carotid bruit, angina pectoris, claudication, myocardial infarction, cerebrovascular accident, or vascular surgery), recent onset of hypertension (within the past 2 years), family history of hypertension (parents, siblings, or children with hypertension), smoking history (ever or never), obesity (body mass index 25 kg/m²), abdominal bruit, advanced hypertensive retinopathy (fundus grade III or IV), serum creatinine concentration, and hypercholesterolemia (serum cholesterol level > 6.5 mmol/L [251.35 mg/dL] or use of cholesterol-lowering agents). These characteristics were used to predict the presence of renal artery stenosis. A patient was considered to have renal artery stenosis when the angiogram showed at least one stenosis of 50% or more in a renal artery according to the local-radiologist.

Model Development

Data are presented as a proportion or as the mean ±SD. The univariable association between clinical characteristics and presence of renal artery stenosis was studied by computing the value and 95% CI of the odds ratio. In a multivariable analysis, clinical characteristics were combined as predictor variables in a logistic regression model predicting the presence of renal artery stenosis (outcome) [27]. For each patient in the multivariable analysis, the probability of renal artery stenosis was calculated from the regression model (predicted probability). The reliability, discriminative ability, and validity of the model were assessed. The Appendix gives details on model development and evaluation.

To enable the use of the regression model in clinical practice, a prediction rule was constructed for predicting renal artery stenosis in future patients with drug-resistant hypertension or an increase in serum creatinine concentration during therapy with ACE inhibitors. For the presence or level of each clinical characteristic in the regression model, a score was calculated on the basis of the regression coefficients (Appendix). These scores were added into a sum score. All possible sum scores and their corresponding predicted probabilities of renal artery stenosis were combined in a graph with 95% CIs of the predicted probabilities.

Role of the Funding Source

Our funding source had no role in the collection, analysis, or interpretation of the data or in the decision to submit the manuscript for publication.

Results

Top Methods Results Discussion Author & Article Info References

Statistical Analyses

Angiography was performed in 439 patients with drug-resistant hypertension and 39 patients with an increase in serum creatinine concentration during therapy with ACE inhibitors. The procedure failed in 1 patient. For the remaining 477 patients, angiography showed renal artery stenosis in 107 patients (22%), of whom 90 (84%) had atherosclerotic stenosis and 17 (16%) had fibromuscular dysplasia. Bilateral stenoses were found in 27 of 107 affected patients (25%). Renal scintigraphy was performed in 458 patients; it had a sensitivity of 72% and a specificity of 90% for the diagnosis of renal artery stenosis.

Table 1 shows the univariable distribution of the clinical characteristics for patients with renal artery stenosis and those with essential hypertension. Most clinical characteristics were indicative of renal artery stenosis (P < 0.05 or borderline significant) except sex, recent onset of hypertension, and presence of advanced hypertensive retinopathy. More young women without signs of atherosclerotic disease were found among patients with fibromuscular dysplasia than among those with atherosclerotic stenosis, but abdominal bruits occurred with the same frequency in both groups (29% and 27%, respectively).

The results of multivariable analysis are also shown in Table 1. Advanced hypertensive retinopathy was not studied any further because this clinical characteristic was missing for 43% of the patients. Data on 11 clinical characteristics of 460 patients were considered predictive of renal artery stenosis. Ethnicity and family history of hypertension were removed from the regression model because their contribution to predicting renal artery stenosis was small. Because renal artery stenosis is believed to be more prevalent in young women and old men, interaction between age and sex was tested; this interaction was not statistically significant (P = 0.09). We included an interaction term between age and smoking because this was the only biologically plausible interaction term that was statistically significant (P = 0.01). This interaction term accounts for the fact that the predictive value of increasing age was stronger for patients who never smoked than for current and former smokers. Finally, the type of standard treatment did not provide additional diagnostic information when it was included in the regression model (P > 0.2). The multivariable odds ratios in Table 1 reflect the predictive effect of the individual clinical characteristics while correcting for the other predictors in the multivariable model. For example, the multivariable odds ratio for atherosclerotic vascular disease was lower than the univariable odds ratio because the model also accounted for the effects of age and smoking history.

Model Performance

Figure 1 shows the agreement between the predicted and the observed probabilities. For 204 patients (44%), the predicted probability of stenosis was 0% to 10%. The predicted probabilities of stenosis obtained from the model agreed well with the observed frequency of stenosis (goodness-of-fit test, P > 0.2). The model discriminated well between patients with renal artery stenosis (predicted probability, 49% ± 29%) and patients with essential hypertension (predicted probability, 15% ± 16%); the area under the receiver-operating characteristic (ROC) curve was 0.84 (95% CI, 0.79 to 0.89). Among patients with stenosis, the discriminative ability of the regression model was better for those with atherosclerotic stenosis (predicted probability, 52% ± 29%) than for those with fibromuscular dysplasia (predicted probability, 34% ± 26%).

View larger version (20K): [in this window] [in a new window]   Figure 1. Agreement between the observed probability of stenosis and the probability of stenosis as predicted by the regression model in 460 patients with drug-resistant hypertension or an increase in serum creatinine concentration during therapy with angiotensin-converting enzyme inhibitors.

The discriminative ability of the prediction rule differed among the four hospitals that included most of the patients. For these hospitals, the area under the ROC curve varied from 0.68 to 0.92. This corresponds with the finding that the associations between stenosis and clinical characteristics of patients from these hospitals were not equally strong or were contradictory. For example, atherosclerotic vascular disease was not predictive of stenosis in one hospital and was even more prevalent in patients with essential hypertension in another hospital. This inconsistency may be explained in part by small sample sizes: The numbers of patients included by these four hospitals were 44, 56, 77, and 151.

Using the Model in Clinical Practice

In the prediction rule for renal artery stenosis, a score was assigned to the level or presence of each clinical characteristic in the regression model (Table 2). These scores were added into a sum score that, through the logistic formula, corresponded with a predicted probability of renal artery stenosis. In Figure 2, the predicted probabilities and their 95% CIs can be derived from the sum scores in a graphical manner. For instance, the sum score for a 46-year-old male patient who smoked in the past; has no signs or symptoms of atherosclerotic vascular disease; received a diagnosis of hypertension 1 year ago; has a body mass index of 23 kg/m (²), no abdominal bruit, a serum creatinine concentration of 112 µmol/L (91.27 mg/dL), and a serum cholesterol level of 5.4 mmol/L (208.82 mg/dL); and does not take cholesterol-lowering drugs is 11 (4.5 + 0 + 0 + 1 + 2 + 0 + 3.5 + 0). The scores for age and creatinine concentration were obtained by linear interpolation. Figure 2 shows that the predicted probability of renal artery stenosis for this patient is 25% (CI, 13% to 43%). The probability can also be calculated by using the formula given in the Appendix.

View larger version (14K): [in this window] [in a new window]   Figure 2. Predicted probability of renal artery stenosis in patients with drug-resistant hypertension as a function of the sum score. The sum score was derived from the prediction rule (Table 2) Thin lines represent 95% CIs.

The probability of stenosis according to the prediction rule can be used to select patients for renal angiography. If angiography is performed only in patients with a probability of stenosis above a certain cut-off level, the number of angiograms performed in the total group of patients will be reduced. Table 3 shows the results of using different cut-off levels for the predicted probability of stenosis. The first row in Table 3 gives the scenario of performing angiography in every patient and therefore identifying all patients with stenosis (sensitivity, 100%). If angiography is performed only in patients whose predicted probability of stenosis is, for example, 10% or more, the number of patients undergoing angiography will be reduced to 61%. However, 1 of every 10 stenoses will be missed (sensitivity, 90%). With increasing cut-off levels, the number of patients undergoing angiography is reduced more and more; as a consequence, however, the number of missed stenoses increases. When a probability of 30% was chosen as the cut-off level, the diagnostic accuracy of the prediction rule (sensitivity, 68%; specificity, 87%) approximated that of renal scintigraphy (sensitivity, 72%; specificity, 90%) in our patient population.

Discussion

Top Methods Results Discussion Author & Article Info References

We developed a clinical prediction rule to predict the presence of renal artery stenosis from the clinical characteristics of 477 patients with drug-resistant hypertension or an increase in serum creatinine concentration during therapy with ACE inhibitors who participated in a prospective study on diagnosis and treatment of renal artery stenosis (the DRASTIC study) [9]. By attributing a score to the presence or absence of nine clinical characteristics, a sum score was obtained that corresponded to a probability of renal artery stenosis. The prediction rule proved to be reliable and discriminated well between patients with renal artery stenosis and those with essential hypertension. By applying the prediction rule in clinical practice to select patients for renal angiography, the number of angiograms obtained may have been reduced considerably in a cost-effective manner.

Clinical characteristics have been mentioned before as a means of identifying patients with renal artery stenosis [16, 20-23]. Several studies have described the relative frequency of characteristics in patients with renal artery stenosis and those with essential hypertension, such as age, duration of hypertension, atherosclerosis, cigarette smoking, and presence of an abdominal bruit. Some of these clinical characteristics are interrelated, such as those suggestive of atherosclerotic vascular disease. In our multivariable model, we assessed the independent associations between clinical characteristics and the presence of renal artery stenosis. Moreover, our simple prediction rule enables the clinician to quantify the probability of stenosis for any specific patient. Unlike other studies describing schemes for selecting patients suspected of having renal artery stenosis on the basis of their clinical characteristics [10, 11], our study provides quantitative insight into the potential consequences of applying our selection criteria.

The prediction rule predicts the presence of anatomic renal artery stenosis in patients with preserved renal function (serum creatinine concentration 200 µmol/L [2.26 mg/dL]) who have drug-resistant hypertension or an increase in serum creatinine concentration during therapy with ACE inhibitors. The prediction rule should not be applied if other, secondary causes of hypertension are not adequately ruled out (such as parenchymal renal disease) and should not be applied to patients with impaired renal function in general. Our study group included some patients who received more medication than the standardized schemes allowed because their blood pressure was very high. Regardless of their blood pressure response to the additional drugs, these patients were considered to be resistant to the standardized regimen and underwent angiography. The prediction rule can therefore be used for patients in whom blood pressure control was achieved with more than two drugs, provided that control could not be achieved on a two-drug regimen. Before introduction on a wide scale, the model must be tested further to establish whether its predictions are valid in other settings.

Although the clinical characteristics of patients with atherosclerotic stenosis and those with fibromuscular dysplasia clearly differ somewhat, the prediction rule can be used to predict the presence of either type of renal artery stenosis. Some clinical characteristics (such as the presence of an abdominal bruit) were found to be relevant for both patient groups, but in other respects (such as signs of atherosclerotic vascular disease), patients with fibromuscular dysplasia resembled those with essential hypertension more closely than they resembled those with atherosclerotic stenosis. Thus, patients with fibromuscular dysplasia are not a distinct group of patients that can be excluded before the prediction rule is applied in clinical practice. For example, only 4 of the 17 patients with fibromuscular dysplasia in our study group were women younger than 40 years of age. We decided not to exclude patients with fibromuscular dysplasia from the analysis because the prediction rule should be applicable to all future patients who present themselves in our clinics. Although the prediction rule performed somewhat better for patients with atherosclerotic stenosis than for patients with fibromuscular dysplasia, the predicted probability in the latter group was significantly higher than that of patients with essential hypertension. Thus, the prediction rule distinguished well between both groups of patients with stenosis and patients with essential hypertension.

In this analysis, anatomic renal artery stenosis was predicted from clinical characteristics. We acknowledge that prediction of functional stenosis (that is, renovascular hypertension) would have been preferable. Unfortunately, no good definition of renovascular hypertension exists. This condition is often defined as being characterized not only by the presence of renal artery stenosis but also by the cure of the hypertension after repair of the stenosis. However, several factors may explain why relief of renal artery stenosis that has caused hypertension does not always result in cure of hypertension, such as advanced-stage hypertension (third phase of two-kidney, one-clip Goldblatt hypertension), technical failure of the intervention, or restenosis. The most important objection to the use of blood pressure response to intervention is that it is a diagnosis made a posteriori. Therefore, the most practical approach is to search for renal artery stenosis instead of renovascular hypertension.

This prediction rule is a practical and simple tool for selecting patients with drug-resistant hypertension or an increase in serum creatinine concentration during therapy with ACE inhibitors. To obtain the probability of stenosis for a specific patient, information is needed on nine clinical characteristics; this information is generally readily available in clinical practice. After prespecified scores are added to form a sum score, the corresponding probability of stenosis can be read from a graph. The usefulness of the prediction rule was shown in our data set. The prediction rule was almost as accurate as renal scintigraphy (sensitivity, 72%; specificity, 90%) in predicting renal artery stenosis if angiography was performed in patients for whom the rule predicted a probability of stenosis greater than 30%. In contrast to renal scintigraphy, however, the results of the prediction rule are immediately available and free. We therefore conclude that the prediction rule can be used as an alternative to renal scintigraphy in the selection of hypertensive patients for renal angiography, provided that the predictions prove to be valid in other settings. Embedded in the diagnostic workup of hypertensive patients who do not respond well to antihypertensive drugs, the prediction rule can help to reduce the number of negative renal angiograms without missing many patients with renal artery stenosis.

Appendix

Model Development

Deletion of cases with missing data may cause bias and increases variance [28]. For 40 patients for whom one clinical characteristic was missing, the value was therefore predicted from the other clinical characteristics by multiple regression on values of the other predictors and was subsequently imputed [28, 29]. Values for 17 patients for whom more than one value was missing were not imputed because the predicted values for these predictors would have been less reliable. These 17 patients were excluded from the multivariable analysis.

Age and serum creatinine concentration were entered into the logistic regression model as continuous variables. We studied whether transformations of these variables offered a better fit. Smoking was dichotomized as ever or never smoked; the fit of more complex classifications, such as never, past, or present smoker or number of pack-years, was also studied. Advanced hypertensive retinopathy was not included in the multivariable analysis because this characteristic was missing in a substantial number of the patients (43%). Nine clinical characteristics were selected for the regression model by backward deletion of the least significant characteristics, done by using the Akaike information criterion [30]. As a result, ethnicity and family history of hypertension were dropped from the model (P > 0.2). Interaction between clinical characteristics in predicting renal artery stenosis was studied in two ways to control for deviation from the additivity assumption [28]. First, a likelihood ratio test on all first-order interaction terms was performed (P > 0.2). Second, biologically plausible interaction terms were tested, which led to the inclusion of age x smoking in the model (P = 0.01).

Model Evaluation

The reliability of the regression model was evaluated by using the Hosmer-Lemeshow goodness-of-fit test [27]. The discriminative ability of the regression model was evaluated by the area under the ROC curve and its 95% CI [31, 32]. The ROC curve is a plot of the false-positive rate (1 minus the specificity) against the true-positive rate (sensitivity), evaluated for consecutive cut-off points of the predicted probability. The area under the ROC curve can be interpreted as the probability that the regression model will assign a higher probability of stenosis to a randomly chosen patient with renal artery stenosis than to a randomly chosen patient with essential hypertension. The area can range from 0.5 to 1 (no to optimal discriminative ability) for sensible models.

The internal validity of the regression model [28, 33] was assessed by using bootstrapping techniques, including variable selection [34]. Random bootstrap samples were drawn with replacement from the full sample (200 replications). The discriminative ability of the regression models was determined on the bootstrap samples and on the full sample, in which predictions were based on the regression models fitted on the bootstrap samples. This validation replicates the situation in which the prediction model based on our patients is applied to a group of similar patients. The area under the ROC curve was 0.84 on the full data set and 0.82 after this procedure. Next, four hospitals that included most of the patients were left out of the sample one by one, and regression models were fitted on the remaining data. The discriminative ability of these models was externally assessed on the hospital not included in the fitting procedure. This procedure replicates the situation in which the prediction model is applied in another hospital with a patient population that may be somewhat different.

Derivation of Scores in the Prediction Rule

The multivariable logistic regression model can be written as: (Equation 1)

(1)

where linear predictor LP = –7.859+ 0.059 x age + 0.033 x (75 –age) x ever smoked –0.996 x sex + 0.585 x atherosclerotic vascular disease + 0.642 x recent on set –1.027 x obesity + 1.693 x abdominal bruit + 0.502 x hypercholesterolemia + 0.032 x serum creatinine concentration.

The regression coefficients were multiplied by a shrink-age factor of 0.88, which was derived from bootstrapping procedures. Shrinkage of the regression coefficients aims to improve calibration of predictions in future patients: that is, to prevent low predictions that are too low and high predictions that are too high [28, 35]. The intercept was adjusted so that the sum of predicted probabilities equalled the number of events (106 patients with stenosis in a total of 460 patients). The shrunk formula was (Equation 2)

(2)

where LP_S = –7.033+ 0.052 x age + 0.029 x (75 –age) x ever smoker –0.877 x sex + 0.515 x atherosclerotic vascular disease + 0.565 x recent onset –0.904 x obesity + 1.490 x abdominal bruit + 0.441 x hypercholesterolemia + 0.028 x serum creatinine concentration.

This formula can be used to calculate the exact probability of stenosis. The average SE of the rounded linear predictor values was used to calculate the 95% CIs of the predicted probabilities (1/1 + e^-[LP_S ± 1.96 x SE]).

For presentation as a prediction rule, the resealed regression coefficients were multiplied by 2 and were rounded to simplify the computation for clinical practice.

Software

Descriptive analyses were performed by SPS statistical software (SPS, Inc., Chicago, Illinois). Imputation of missing values, logistic regression, and validation were carried out in the Design Library for S-plus by using the transcan, impute, lrm, and validate functions [36].

Drs. van Jaarsveld, Man in 't Veld, and Schalekamp: Department of Internal Medicine I, University Hospital Dijkzigt, Dr Molewaterplein 40, 3015 GD Rotterdam, the Netherlands.