Risk Factors for Syncope Recurrence After a Positive Tilt-Table⇓ Test in Patients With Syncope
Background Recent work with head-up tilt-table testing has suggested that many patients with syncope may have recurrent neurally mediated episodes of bradycardia, hypotension, or both. The purpose of this study was to determine how to identify patients at high risk of a recurrence of neuromediated syncope after a positive isoproterenol/tilt-table test.
Methods and Results A cohort of 101 drug-free patients in a university hospital outpatient clinic with syncope and a positive isoproterenol/tilt-table test underwent baseline assessment of demographic variables, symptomatic burden, and hemodynamic and clinical responses to tilt testing. The primary outcome measure was the time to the first recurrent syncopal spell. The actuarial probabilities of remaining syncope free after 1 and 2 years were 72% and 60%, respectively. Multivariate proportional hazards analysis demonstrated that the most powerful predictor of a recurrence of syncope was the logarithm of the number of preceding syncopal spells (P<.001). Other predictive variables included the duration of syncopal symptoms, tilt-test symptomatic outcome, and trough heart rate. The probability of a recurrence of syncope also varied with the logarithm of the frequency of preceding spells (P=.008). The median frequency of pretest spells was 0.3/month; after the tilt test, the median frequency dropped ≈90% to 0.03 per month.
Conclusions The risk of a recurrence of syncope after a positive tilt-table test can be predicted with simple pretest and intratest variables.
Recurrent syncope is a common clinical problem and a significant diagnostic and therapeutic challenge. Recent work with head-up tilt-table testing has suggested that many patients with syncope have recurrent episodes associated with variable combinations of bradycardia and hypotension.1 2 3 4 This has been variably described as vasovagal syncope, bradycardia/hypotension syndrome, or neuromediated syncope. Attention is focused increasingly on how best to treat patients with recurrent neuromediated syncope. To counsel patients and to plan clinical trials, we need to know which patients are at high risk of recurrent syncope after tilt-table testing and, conversely, which patients are at such low risk of recurrent syncope that they could be managed conservatively without drug therapy.
At this time, there are no data that can be used to identify high- or low-risk patients and little data documenting the clinical course of drug-free patients after tilt-table testing. The purpose of the present study was to determine how to identify patients at high risk of recurrent neuromediated syncope after a positive isoproterenol/tilt-table test. Specifically, we used univariate and multivariate analyses to determine which pretest clinical variables and intratest tilt-table variables were significant risk factors for recurrent neuromediated syncope.
The global patient population consisted of 338 sequentially consenting patients referred for assessment of syncope who underwent tilt-table testing if they had had (1) two or more syncopal episodes, (2) one syncopal episode and four or more presyncopal episodes, or (3) a single episode of syncope causing serious injury. Patients with structural heart disease, documented ventricular tachycardia, or bifascicular block also underwent ambulatory ECG and programmed electrical stimulation with a previously described protocol.5 Tilt-table tests and electrophysiological studies were performed on separate days. No patients underwent tilt-table testing while taking β-adrenoceptor–blocking agents,6 disopyramide, or drugs with anticholinergic activity.
Patients underwent tilt-table testing in a quiet room after they had fasted for 4 to 8 hours. They were comfortably restrained on an electric tilt table. Instrumentation consisted of a peripheral intravenous cannula and automatic and manual sphygmomanometric blood pressure cuffs, but intra-arterial cannulation was not performed. Our previously published protocol7 8 was used. The limitations of this approach with regard to the generalizability of our findings are reviewed in “Discussion.” The test ended after frank syncope during an infusion of 5 μg/min isoproterenol or to a total of 10 minutes in the head-up position with either presyncope or no symptoms. Heart rate, blood pressure, and symptoms were recorded each minute. Based on our previous work,7 tilt tests were considered to be positive if they ended in syncope or in presyncope and a drop in rate–systolic pressure product to ≤9000 mm Hg/min. We stress that tilt tests ended only with syncope or at the end of a 10-minute head-up tilt. Because presyncope was not a cause for stopping the test before the end of a 10-minute period, the two positive test outcomes (syncope and presyncope) can be assessed as separate prognostic risk factors for clinical outcome.
“Syncope” was defined as a transient state of unconsciousness characterized by spontaneous recovery or by recovery during the supine position. “Presyncope” was defined as a state of lightheadedness usually associated with one or more symptoms of decreased vision, the sensation of hearing voices distantly, slow response times to verbal stimuli, nausea, vomiting, or partial loss of postural tone and which substantially reproduced the clinical presyncope of the patient. “Duration of symptoms” refers to the number of months elapsed between the first historical syncopal spell and the diagnostic tilt-table test.
After tilt-table testing and a brief rest period, all patients underwent counseling regarding the diagnosis, probable pathophysiology, lack of mortality, and uncertain symptomatic prognosis of neuromediated syncope. All patients were reassured, counseled on recognizing their presyncopal prodromal symptoms, and urged to take specific appropriate postural maneuvers when presyncopal. Throughout this study, we recognized the lack of level I to IV studies9 demonstrating drug efficacy. The lack of certainty about the efficacy of any form of pharmacological therapy was discussed, and no patients were urged to accept empiric drug therapy. All patients who were started on drug therapy did so at their own request. Throughout the study, we used β-adrenergic–blocking agents (when not contraindicated) for initial attempts at treatment. Thus, although patients were started on drug treatment at their request, β-blockers were selected as the particular pharmacological therapy by the investigators. The differences between the treated and untreated populations are described in “Results.”
All patients were asked to notify the syncope clinic of their first recurrence of syncope, and all patients were also contacted every 6 months by telephone. We defined the following end points. The primary end point was the first recurrent syncopal spell in patients who did not receive pharmacological therapy from the day of the tilt test and who had a subsequent syncopal spell. Presyncope was not an end point (see “Discussion”). Patients who did not have a recurrent syncopal spell and had not received drug treatment were censored on the date of last contact with the clinic. Patients who crossed over from no treatment to treatment for any reason and had had no recurrent syncope by that time were censored on the date of crossover.
We first determined whether a variable was normally distributed or skewed. The skewed distributions were analyzed after natural logarithmic transformation. Mean (±SD) and median values were calculated for continuous variables, and frequencies were measured for categorical variables. Differences between groups were examined for statistical significance with a two-sample t test for continuous variables (using a logarithmic transformation on skewed distributions as appropriate) and by Fisher’s exact test for categorical variables. P values are two-tailed without adjustment for multiple P values. Kaplan-Meier estimates of syncope-free survival were used to illustrate the overall syncope-free survival rate and to assess the prognostic effect of variables.
Proportional Hazards Model
Because most of the variables were continuous, the association of each variable with the probability of remaining syncope-free was assessed first with a univariate Cox proportional hazards regression model,10 which was then used for multivariate analysis. All variables were entered into the multivariate model regardless of their significance in the univariate analysis. The statistical significance of each variable in the hazards models was evaluated with the Z statistic for the regression coefficient. Variables significant at P<.10 were retained in the proportional hazards model. All two-way interactions between the variables were assessed for statistical significance. A Martingale residual analysis11 was used to assess the most appropriate functional form for the continuous variables, and Schoenfeld residuals12 were used to test the assumption of proportional hazards. The instantaneous relative risks for dichotomous variables were calculated from the exponential of the estimated regression coefficient β of the variable, with all other variables in the model remaining constant. The instantaneous relative risk for a continuous variable for an increase in n units of that variable is exp(β)n.
One goal of the present study was to develop a model that would enable clinicians to predict the clinical outcome of patients based on pretest and intratest variables. Smooth continuous estimates of modeled survivor function were necessary for construction of this predictive model. The survivor function first was calculated from the proportional hazards model.13 The survivor function is a step function, similar to a Kaplan-Meier estimate. Smoothed survivor functions were calculated by fitting the Weibull distribution S(t)=exp[−(ρt)λ] to the survivor step function. The Weibull distribution was selected for several reasons. First, it is closely related to the exponential Poisson distribution, which assumes the randomness of outcomes in a population and has proved to be useful in outcome studies with supraventricular arrhythmias. Second, it fit the data better than other equations (see below). Third, it allows a continuous estimate of changes in relative risk of outcomes over time. Note that a Weibull distribution with λ=1 is a conventional exponential distribution. The survivor function in Fig 1⇓ is well fit by a Weibull distribution with λ=0.566 with a 95% confidence interval of 0.543 to 0.588, indicating that a simple exponential distribution would not adequately fit the data.
Smooth estimated survivor functions for individuals with nonzero covariates were then calculated with one coefficient from the proportional hazards model. The proportional hazards regression model with pretest variables was used to produce a model that can predict the risk of a recurrence of syncope within the first 12 or 24 months after a tilt-table test. The baseline hazard function was estimated with a Weibull distribution, and the combination of values of the two clinical variables (number of spells and duration of symptoms) were calculated to solve the proportional hazards equation for each specified risk.
Of the 338 patients in the global population, 196 had a positive tilt-table test. Of these, 52 received empiric therapy with β-adrenoceptor blockers and 10 were treated with other medications. An additional 33 were lost to follow-up. The 101 patients who did not receive drug therapy form the basis of this report. The baseline clinical characteristics and tilt-test variables of the 43 patients lost to follow-up or treated with other medications closely resemble the characteristics of the 153 patients who either did not receive pharmacological therapy (101) or received β-blockers (52), although these 153 patients had an insignificantly lower geometric mean number of syncopal spells (7.1; confidence interval, 5.3 to 9.4) than did the excluded patients (12.5; confidence interval, 7.3 to 21.4) (P=.074).
The baseline clinical characteristics and tilt-test variables of the 52 patients treated with β-blockers closely resemble the characteristics of the 101 study patients, although the patients who elected to be treated were older (44±21 versus 36±19 years; P=.03) and had fainted more frequently. The patients who subsequently received β-blockers had had a geometric mean frequency of 0.55 spells per month (confidence interval, 0.30 to 0.92), whereas untreated patients had had a geometric mean frequency of 0.31 spells per month (confidence interval, 0.23 to 0.41) (P=.04).
The clinical and tilt-table test variables of the study population are shown in Table 1⇑. The population consisted of 43 men and 58 women with a mean age of 36±19 years (range, 11 to 89 years). They had had a geometric mean of 6 syncopal spells over a geometric mean of 21 months with a geometric mean frequency of 0.3 syncopal spells per month. During the tilt test, 49 patients developed syncope and 52 patients developed presyncope. Table 2⇓ shows the distribution of patients according to their historical number of syncopal spells or duration of symptoms. Although many patients (46 of 101) had ≤3 spells, there was a substantial population (33 of 101) with >10 spells. Similarly, although many patients (33 of 101) had had symptoms for ≤10 months, there were many others (51 of 101) with symptoms lasting >30 months.
Of the 101 patients in this study, 28 had a syncopal spell at some time after a positive tilt-table test, whereas 73 did not have recurrent syncope. The probability of remaining free of syncope is shown in Fig 1A⇑. The step function indicates the Kaplan-Meier estimate of the survivor function (with 95% confidence intervals). The smooth curve indicates the survivor function of the Weibull distribution that provided the closest fit to the data. There is a steady decline in the probability of remaining syncope free with time. This probability after 1, 2, and 3 years was 0.72, 0.62, and 0.51, respectively.
The populations of patients with and without a recurrence of syncope are compared in a univariate analysis in Table 1⇑. Patients with a recurrence of syncope were younger, were more likely to be women, had more syncopal spells and a longer duration of symptoms, and had a significantly higher peak heart rate during tilt-table testing. There were no significant differences between these two populations in any of the other tilt-table test variables.
The variables listed in Table 1⇑ were subjected to a univariate analysis of their prognostic value in predicting the hazard of recurrent syncope after a positive tilt-table test. The variables in Table 3⇓ are those that reached a statistical difference of P<.1 and those that became significant in a subsequent multivariate proportional hazards analysis. The pretest variable that predicted a recurrence of syncope most powerfully was the logarithm of the number of syncopal spells preceding the tilt test (P<.001). Other pretest variables that had prognostic significance were the logarithm of age, the logarithm of the duration of symptoms, and the frequency of syncopal spells. Of the tilt-table test variables listed in Table 1⇑, only peak heart rate was a significant univariate predictor of a recurrence of syncope (P=.019). Two other variables, a polynomial transformation of trough heart rate and syncope as an outcome, were not significant univariate predictors of outcome but were identified subsequently as significant predictors in a multivariate proportional hazards analysis (Table 4⇓).
Kaplan-Meier estimates of the probability of remaining free of syncope for four different groups of subjects are presented in Fig 1B⇑. Patients with <6 syncopal spells over a duration of <24 months have a 1-year event-free probability of 0.94, whereas patients with ≥6 syncopal spells over a duration ≥24 months have a probability of remaining free of syncope after 1 year of only 0.54. Patients with <6 syncopal spells over a period ≥24 months and patients with ≥6 syncopal spells over a duration of <24 months have intermediate probabilities of remaining free of syncope at 1 year of 0.66 and 0.71, respectively. Thus, the simple pretest clinical characteristics of the number of syncopal spells and the duration of symptoms are important predictors of the probability of a recurrence of syncope.
The multivariate proportional hazards analysis was performed in two stages. The first model included only the pretest variables, whereas the second model considered both pretest and intratest variables. In the first analysis (Table 4⇑, top), the variable log (spells and number) was a powerful predictor of a recurrence of syncope (P<.001). In addition, there was a significant interaction (P=.008) of this variable with log (duration of symptoms). This indicates that the prognostic effect of each of these two variables must be considered simultaneously. We have illustrated this interactive effect in Fig 2A⇓ and 2B⇓. This figure depicts the probability of remaining free of syncope for a modeled average patient with various combinations of numbers of pretest syncopal spells and durations of symptoms. A patient with a duration of symptoms of 6 months has a much higher likelihood of a recurrence of syncope if he or she had had 17 syncopal spells preceding the tilt-table test compared with a patient who had had only 2 preceding syncopal spells. The probabilities of remaining free of syncope after 2 years in these two modeled patients were 0.26 (17 spells) and 0.82 (2 spells), respectively. In contrast, a patient with symptoms lasting 84 months had 2-year probabilities of remaining free of syncope of 0.46 and 0.66 if he or she had had 17 and 2 preceding spells, respectively. Interestingly, a patient with 7 syncopal spells had a probability of remaining free of syncope that was independent of the duration of symptoms. Both patients had probabilities of remaining free of syncope after 2 years of 0.54 (Fig 2A⇓ and 2B⇓).
The significant interaction of the number of syncopal spells and duration of symptoms as predictors of a recurrence of syncope indicates that the relation between these two variables is not as straightforward as the simple frequency of symptoms. This interaction is summarized in Fig 2C⇑. The risk of a recurrence of syncope for patients who had 7 syncopal episodes is independent of the duration of symptoms. With this homogeneous group as baseline, we estimated the instantaneous relative risk as a function of the total number of previous syncopal episodes for five populations that have histories of symptoms of 1, 6, 27, 84, and 240 months. These were the 5th, 25th, 50th, 75th, and 95th percentiles of durations. In Fig 2C⇑, we illustrate the risk for an individual with a given number of episodes (horizontal axis) in a certain length of time (lines) relative to an individual who had 7 episodes in the same length of time. The risk of a recurrence of syncope increases linearly with the logarithm of the total number of preceding syncopal episodes. The significant interaction indicates that although the risk of a recurrence of syncope always increases with the number of spells, the rate of increase is much greater if the symptoms occurred over a short period of time. For patients with <7 episodes, the relative risk is always <1, indicating that individuals with <7 episodes are at less risk than are patients with ≥7 episodes within the same period of time. In addition, this relative risk decreases as the duration of symptoms decreases. Conversely, for patients who have >7 episodes, the relative risk increases rapidly as the duration of symptoms decreases.
The effect of the interaction between the number of historical syncopal spells and the duration of symptoms on prognosis is depicted in Fig 3⇓. The independent variable is the probability of remaining free of syncope throughout the first year after a positive tilt-table test. Patients who have had few syncopal spells over a short period of time are at much lower risk of a syncopal recurrence than patients who have had a large number of syncopal spells over a short period of time. The effect of a long duration of symptoms is to smooth out (to some extent) the differences in these risks.
Table 4⇑ shows illustrative changes in instantaneous relative risks of a recurrence of syncope depending on pretest clinical variables. For example, a doubling in the historical number of syncopal spells increases the instantaneous relative risk of syncope by 1.23- to 2.82-fold, depending on the duration of symptoms. The 95% confidence intervals of these estimates are 1.00 to 1.52 and 1.49 to 5.38, respectively.
The prognostic significance of pretest and intratest variables in a model that incorporates both types of variables is shown in Table 5⇓ (bottom). The addition of intratest variables does not change the prognostic value of the pretest variables, as can be seen by the lack of change in the regression coefficients. In addition, syncope as a tilt-test outcome (P=.062) and a polynomial function of trough heart rate (P=.078 and P=.10 for the linear and quadratic terms, respectively) both tended to predict a recurrence of syncope. The data in Fig 4A⇓ show that patients who develop syncope on a tilt-table test have a 2-year probability of remaining free of syncope of 0.45, whereas patients who develop only presyncope on a tilt-table test have a 2-year probability of remaining fee of syncope of 0.72. The quadratic nature of the predictive value of trough heart rate is illustrated in Fig 4B⇓. This model shows that patients with trough heart rates of 60, 80, and 100 beats per minute have a 2-year probability of remaining free of syncope of 0.63, 0.37, and 0.53, respectively. Fig 4C⇓ shows that the likelihood of a recurrence of syncope is highest for patients with a trough heart rate of 75 to 85 beats per minute and that this risk decreases steadily with trough heart rates below or above this critical range.
These findings demonstrate that the risk of a recurrence of syncope after a positive isoproterenol/head-up tilt-table test can be estimated from a knowledge of two simple clinical variables—the number of pretest syncopal spells and the duration of the pretest symptoms. This estimate can be refined further based on whether patients faint or merely become presyncopal during the tilt-table test and based on the trough heart rate during symptoms. This information may prove useful in counseling patients about their risk of a recurrence of syncope, in determining which patients might be candidates for pharmacological therapy, and in determining which patients might be candidates for future interventional trials.
Until the advent of tilt-table testing, the diagnosis of idiopathic syncope frequently frustrated clinicians. Tilt-table testing has improved understanding of this syndrome, but in the absence of a gold standard, diagnostic criteria have remained a test of elusive sensitivity and contentious specificity.14 15 We elected to define a study population as patients with two diagnostic criteria: clinical syncope and symptoms that could be reproduced on isoproterenol/head-up tilt-table testing.
Choice of Clinical Outcome
Studies concerned with clinical outcome (such as this one) are critically dependent on the choice of outcome measure. In a study of neuromediated syncope, these measures include syncope, recurrences of presyncope and syncope, and objective measures of quality of life. We chose syncope as an outcome because it is easily quantifiable, is memorable, and was likely to have a high incidence due to the highly symptomatic nature of our study population. We chose to use the time to the first recurrence of syncope as a primary end point because it seemed reasonable to ask patients to defer medical therapy until that time as we anticipated that patients might then request medical therapy and as first recurrence is widely used as a primary outcome measure in arrhythmia trials. This use is based in part on the temporally random recurrence of supraventricular arrhythmias in study populations.16 17 The close fit of the recurrence data in this report to the Weibull distribution, which assumes a random process, supports this assumption. Presyncope was not chosen due to anticipated difficulties with quantifying its variable severity and duration. Recurrent presyncope is likely to be a significant problem in the lives of these patients. Although the quality of life in patients with recurrent syncope is substantially and objectively diminished,18 19 we did not use this as an end point because we made no a priori assumptions that quality of life would change after a positive tilt-table test.
Reduction in the Risk of a Recurrence of Syncope
One surprising finding from the present study is the apparent reduction in the risk of a recurrence of syncope after a positive tilt-table test. In this population, the median frequency of historical syncopal spells per month was 0.3; that is, half of the population had fewer than 0.3 syncopal spells per month before undergoing tilt-table testing and counseling. After this clinical encounter, half of the population did not faint during 3 years of follow-up. The cause for this apparently great reduction in the likelihood of a recurrence of syncope after a tilt-table test compared with the pretest history is unknown, but it may be due to a combination of the natural history of the syndrome, reassurance, counseling on the pathophysiology of syncope, and coaching on appropriate postural maneuvers to prevent presyncope from progressing to syncope. The data cannot rule out the possibility that some patients have a great reduction in the frequency of syncopal spells, whereas other patients have a much smaller reduction in the frequency of spells. The data do suggest that the tilt test and clinical encounter are themselves interventions. To test these hypotheses will require randomized prospective studies.
Usefulness of Risk Assessment
These data provide the first estimates of risk assessment in neuromediated syncope. Recognizing the plethora of tilt-test protocols that are now in use (see below), we suspect that most investigators will focus on historical factors such as prognostic markers. Risk assessment should prove helpful in counseling patients about their prognoses, in deciding whether an attempt at pharmacological or pacing therapy is warranted, and in designing prospective clinical trials. Future trials, to be adequately powered, should be carried out with subjects who have a sufficiently high risk of recurrent syncope.
The risk of recurrent syncope based on clinical history can be estimated with the use of the diagram in Fig 5⇓. The probabilities of syncope in the first year (top) and the first 2 years (bottom) are displayed as a function of both the number of preceding syncopal spells and the duration of preceding symptoms. The boundaries of these zones were calculated from the proportional hazards model as described in “Methods.” Four zones are displayed in which the risk of recurrent syncope is <25%, 25% to 50%, 50% to 75%, and >75%. Superimposed on these zones are single data points, each of which represents one patient.
Fig 5⇑ has three purposes. First, the risk of a recurrence of syncope for any patient after a positive tilt-table test can be estimated directly from the figure. To use this figure, one simply locates the position of any patient based on their number of syncopal spells and duration of symptoms. For example, a patient with four syncopal spells over a period of 2 months has a risk of at least one additional syncopal spell of <25% in 1 year and of 25% to 50% in 2 years. Second, it can be used to estimate inclusion criteria and sample size for future trials. For example, all patients with six or fewer syncopal spells have a risk of recurrent syncope of <50% over 2 years, whereas patients with more than six syncopal spells have a risk of syncope of >50% over 2 years. An interventional study of the former population would require a much larger study size than would one of the latter population. Third, it displays the relative proportion of patients in each category in this population of patients, who presented to a community-based tertiary care center. For example, a study restricted to patients with >75% risk of syncope in 1 year might seem desirable, but only 5 of 101 patients in our study population are in this category. Fig 5⇑ therefore illustrates not only what is desirable but also what is feasible in trial design.
The major limitations of the present study are those of generalizability. First, this report is focused on patients who had a positive outcome to the isoproterenol/head-up tilt-table test protocol used in our laboratory. Therefore, we cannot comment on the long-term outcome of patients with a positive outcome to passive prolonged head-up tilt testing,1 4 or to tilt testing using one of a variety of other provocative agents, such as nitroglycerin or a different dose of isoproterenol.15 The use of isoproterenol remains a controversial issue.14 These protocols have not been compared directly with each other in a prospective controlled trial. Although these limitations are particularly relevant when considering the generalizability of the prognostic significance of the intratest variables, they are also important because the sensitivity and specificity of the various protocols may differ.20 21
Second, we did not examine the prognostic significance of a positive outcome to tilt-table testing. In the absence of a gold standard for the diagnosis of syncope, such a study would be difficult to perform and interpret because it would require establishing a pretest diagnosis of neuromediated syncope.
Third, we used syncope, not presyncope, as the outcome measure. The reasons for this include simplicity, quantifiability, memorability, and the impact of fear of syncope on quality of life. However, part of our clinical interaction with patients included teaching the specific physical maneuvers to abort the progression of presyncope to syncope. If successful, this would reduce the risk of syncope without reducing the clinical burden of presyncope.
In contrast to concerns with the generalizability of the tilt-test protocol, the population in this study probably is broadly representative. The patients had a reasonably equal sex distribution and a broad age range and presented with a wide range of historical number of syncopal spells and duration of symptoms. Although some patients had many spells and/or long durations of symptoms, the majority had six or fewer syncopal spells and durations of symptoms of ≤21 months (Tables 1⇑ and 2⇑). Thus, the population was composed of patients with a wide range of symptomatic burdens.
Similarly, we focused on patients who elected to go home counseled but not treated with pharmacological agents and who were not lost to follow-up. This raises the possibility that selection bias might alter the findings. Although the patients who were lost to follow-up closely resembled the study patients, other unanticipated, unmeasured variables might exist. Similarly, the patients who elected to be treated were older and had had significantly more frequent syncopal spells than did the 101 untreated patients. However, in a separate analysis6 of a larger population consisting of these 101 drug-free patients and the subpopulation of the 52 patients who elected to be treated with β-blockers, we showed that the multivariate risk factors reported here all became more highly significant as risk factors for recurrent syncope in the total population. Therefore, these identified risk factors appear to be significant not only for patients who elect not to be treated but also for those who wish to be treated with β-blockers. These numerous issues, so typical in the weighing of the merit and generalizability of a retrospective study, can be addressed only in a prospective, inclusive natural history study or in the placebo arm of a large, randomized, controlled trial. This report provides the estimate of the probability of a recurrence of syncope that is necessary for the design of such a study or trial.
Finally, the large reduction in the risk of a recurrence of syncope in the population suggests that the clinical encounter that includes tilt testing is not neutral: the various nonpharmacological management strategies appear to have a salutary impact on the outcome of many patients.
In conclusion, the present study identifies significant clinical and tilt-test variables that predict recurrent syncope after a positive tilt-table test in patients with a previous history of syncope. These findings should help decide which patients might be treated pharmacologically and which patients have such a low risk of recurrent syncope that they could be treated expectantly.
This work was supported by grants from the Medical Research Council of Canada (PG11188) and the Calgary General Hospital Research and Development Committee.
- Received September 6, 1995.
- Revision received October 2, 1995.
- Accepted October 6, 1995.
- Copyright © 1996 by American Heart Association
Sheldon RS, Rose S, Flanagan P, Koshman ML, Killam S. Age-dependent effect of beta blockers on the clinical outcome of patients following a positive isoproterenol-tilt table test. Can J Cardiol. 1994;10:79C. Abstract.
Cook DJ, Guyatt GH, Laupacis A, Sachett DL. Rules of evidence and clinical recommendations on the use of antithrombotic agents. Chest. 1992;102:3055-3115.
Cox DR. Regression models and life-tables (with discussion). JR Stat [Soc B]. 1972;34:187-220.
Therneau TM, Grambsch PM, Fleming TR. Martingale-based residuals for survival models. Biometrika. 1990;77:147-160.
Schoenfeld D. Partial residuals for the proportional hazards regression model. Biometrika. 1982;69:239-241.
Oakes DR, Oakes D. Analysis of Survival Data. London, UK: Chapman & Hall; 1984.
Kapoor W, Brant N. Evaluation of syncope by upright tilt testing with isoproterenol: a nonspecific test. Ann Intern Med. 1992;116:358-363.
Kapoor WN. Methodology of upright tilt table testing. Eur J Cardiac Pacing Electrophysiol. 1992;2:242-246.
Sheldon RS. Tilt table testing in the diagnosis and treatment of syncope. Cardiovasc Rev Resp. 1994;15:8-28.