Vulnerable and Stabilized States After Cerebral Ischemic Events
Implications of Kinetic Modeling in the SOCRATES, POINT, and THALES Trials
Abstract
Background and Objectives
Trials of acute secondary prevention after minor stroke or transient ischemic attack (TIA), such as SOCRATES, POINT, and THALES, demonstrate a high initial rate of recurrence after ischemic events that drop quickly to a lower rate, suggesting a transient vulnerable clinical state, which may call for different treatments than the subsequent stabilized state. A kinetic model incorporating vulnerable and stabilized states provides estimates of the distinct kinetic rates reflecting the temporal features of underlying stroke mechanisms. We aimed to compare these kinetic rates between treatments and across trials, asking whether these features point to common pathophysiologic processes underlying stroke recurrence, and inform the targeting and timing of enhanced antiplatelet therapy in recurrent stroke prevention.
Methods
Kaplan-Meier recurrence-free survival curves in the SOCRATES, POINT, and THALES trials were estimated for each treatment group and fitted by nonlinear regression to the 2-state kinetic model, producing estimates of kinetic parameters, with standard errors estimated using the nonparametric bootstrap with repetitive resampling.
Results
For each trial, the 2-state kinetic model fit the survival curves better than did the null (single-state) kinetic model or the Weibull model (p < 0.05). Recurrence rates in the vulnerable state (k1) were 100-fold higher than in the stabilized state (k2). Transition rates from the vulnerable to stabilized state (k0) were still more rapid. Kinetic parameters were consistent across the trials, without significant differences between the trials. Enhanced antiplatelet regimens produced significant reductions in k1 (aspirin alone: 0.030 ± 0.004 d−1; active treatment: 0.016 ± 0.003 d−1; p < 0.01) but did not affect k0 or k2, suggesting that active treatment only affected risk in the vulnerable state. Modeling based on these kinetic parameters suggests that most of the benefit of active treatment occurred within 3 days.
Discussion
Across multiple trials of acute secondary prevention after minor stroke or TIA, recurrence of stroke is well-described by a 2-state kinetic model postulating vulnerable and stabilized states, with similar kinetic parameters across trials. Enhanced antiplatelet regimens only affected the recurrence rates in the vulnerable state, over a brief period. This analysis suggests that 2 distinct states follow acute cerebral ischemic events, subject to differential impact of immediate or delayed therapies.
Introduction
After an initial minor stroke or transient ischemic attack (TIA), the risk of recurrence of stroke is the highest initially. Trials of various antiplatelet regimens for acute secondary prevention initiated within 12–24 hours after the index event, such as the Clopidogrel in High-Risk Patients with Nondisabling Cerebrovascular Events (CHANCE),1 Acute Stroke or Transient Ischemic Attack Treated with Aspirin or Ticagrelor and Patient Outcomes (SOCRATES),2 Platelet-Oriented Inhibition in New TIA and Minor Ischemic Stroke Trial (POINT),3 and Acute Stroke or Transient Ischemic Attack Treated with Ticagrelor and ASA for Prevention of Stroke and Death (THALES),4 have been consistent in showing that the early recurrences within the first week exceed those in the following months. This suggests the possibility of a shift over time between different clinical states, subject to distinct pathophysiologic risks and treatment opportunities, with distinct rates of recurrence. Such temporal features can be modeled by kinetic analysis, with assumptions of state populations related by stochastic transitions determined by non-negative time-independent kinetic rates. The data from these trials that recorded the timing of outcomes starting nearly immediately after stroke or TIA provide an opportunity to precisely examine the kinetics of clinical events after an initial ischemic event and to discover whether tested treatment regimens have distinct effects on the different clinical states.
We developed a 2-state kinetic model of stroke recurrence, postulating an initial vulnerable state having a higher rate of stroke recurrence, rapidly transitioning to a stabilized state with a lower rate of stroke recurrence. These assumptions predict a mathematical form of the event-free survival curve that demonstrated excellent fit to the clinical data of the POINT trial.5 We hypothesized that this 2-state kinetic model would better fit the survival data for each of these recent trials of acute secondary prevention than would a model assuming only a single clinical state after the initial minor stroke or TIA.
The kinetic rates quantified in this analysis may relate to underlying pathologic processes that confer risk of stroke, such as active thrombosis or ulceration in vulnerable arterial plaques. We hypothesized that if the kinetics of stroke recurrence after a minor stroke or TIA relate to fundamental underlying pathologic features, the kinetic parameters determined by fitting the 2-state model to clinical data ought to be similar across the control groups of acute prevention trials with similar clinical populations. Furthermore, treatments targeting distinct underlying pathophysiologic processes may differentially affect the early and late phases of stroke recurrence after an index event. Awareness of such temporal changes in pathophysiology and responsiveness to enhanced antiplatelet therapy regimens over aspirin monotherapy may lead to optimized secondary prevention strategies. We hypothesized that the enhanced antiplatelet regimens tested in these trials might selectively affect the stroke recurrence rates in the vulnerable state only. Testing these hypotheses, we aimed to understand the basis for the temporal features of stroke recurrence after an initial ischemic event, providing insights into the impact of antiplatelet therapy on the pathophysiologic events underlying recurrence, and into the effective timing of such treatment.
Methods
Clinical Data Sets
The design, methods, execution, and results of the SOCRATES, POINT, and THALES trials have been reported previously.2-4 These trials were selected as they shared a similar focus, on acute secondary prevention in patients with TIA or minor ischemic stroke, and enrolled subjects rapidly, within 12 hours (for POINT) or 24 hours (for SOCRATES and THALES) of the index (initial) stroke event. In each of these trials, subjects in the control group were treated with aspirin alone, while those in the active treatment groups received enhanced antiplatelet regimens (ticagrelor alone in SOCRATES, aspirin plus clopidogrel in POINT, and aspirin plus ticagrelor in THALES).
The POINT trial data set was provided by the National Institute of Neurological Disease and Stroke. Access to the data sets of the SOCRATES and THALES trials was provided through a data use agreement with AstraZeneca, Inc., and made available on a secure server through Vivli, Inc.
The shared data sets included 4,881 subjects from the POINT trial, 12,018 subjects from the SOCRATES trial, and 9,565 subjects from the THALES trial. Data were not shared for subjects who had not given or had withdrawn consent (113 subjects in SOCRATES and 20 subjects in THALES) nor for subjects from countries not allowing data reuse based on the consent wording (1,182 subjects in SOCRATES and 1,489 subjects in THALES). Survival tables were prepared from the shared data files using SAS Studio version 9.4. From the shared data sets, 101 subjects from the SOCRATES trial and 44 subjects from the THALES trial were not assigned to a treatment group and were excluded from the analysis. Intention-to-treat populations, definitions of outcome events, designation of event times, and data censoring methods were defined according to the trial protocols. The present analysis used treatment assignments by the intention-to-treat principle and examined ischemic stroke recurrence as the outcome event.
Kinetic Modeling of Clinical Events
Kinetic modeling can be applied in the analysis of event recurrence in clinical trials by considering subjects to be in interconnected states, with transitions between states occurring stochastically with non-negative time-invariant kinetic rates.6 A simple null hypothesis for stroke recurrence prevention trials would assume that all subjects are in a single vulnerable state after minor stroke or TIA, characterized by a single rate constant k for stroke recurrence, hypothesizing no difference in the incidence rate of recurrent stroke across time (Figure 1). In this case, the survival function at time t, denoted as Su(t), would follow a single exponential decay:
Stroke recurrence after initial stroke or TIA may be conceived as (A) a simple kinetic model with a single state and a single kinetic rate characterizing event occurrences (null model) or as (B) a 2-state model, in which subjects are initially in a vulnerable state V, susceptible to stroke outcomes by a rapid process with rate k1 but also can transition at rate k0 to a stabilized state S, no longer vulnerable to the rapid rate of stroke outcomes. In both states, events can occur by a slower process with rate k2. The Weibull model is not shown because it is not a kinetic model that can be represented by a kinetic diagram.
The 2-state kinetic model postulates that patients are either in a vulnerable state V, with a risk for outcome events at a rate k1, or in a stabilized state S, with a lower rate k2 for outcome events, and they can transition from state V to state S with a rate k0, with the fraction of subjects initially (at time t = 0) in state V designated as V0 (Figure 1). This model can be shown5 to yield a biexponential function for event-free survival over time given by:
In this model, the survival function Su(t) is fully determined by 3 parameters: K = k0 + k1, k1’ = k1 × V0, and k2. The parameters k0 and k1 can be independently estimated from survival data only with an assumption for the value of V0. Note that the 2-state model (2) can be viewed as a mixture of 2 exponential distributions, with the mixing proportions determined by the event rates for each state and the transition rate between states. Moreover, the null model (1) is a particular case of model 2 obtained when k2 = k and k1 = 0.
An alternative comparison can be made using a more flexible parametric model that is frequently used in survival analyses, the Weibull model, with 2 free parameters λ and k:
The Weibull model is not a kinetic model and does not provide direct clinical interpretation of the parameter values.
We assessed the appropriateness of the full model (2) vs the null model (1) or the Weibull model (3) in data sets of the POINT, SOCRATES, and THALES trials and tested for significant differences in the kinetic parameters k0, k1, and k2 across the trials. For simplicity, only results under the assumption of V0 = 1 are reported (i.e., after index event, all patients are in the vulnerable state). Similar results were obtained using the values of V0 = 0.5 or V0 = 0.1 or using K, k1’, and k2 as free parameters.
Statistical Methods
Using the statistical software R, version 4.2.0, Kaplan-Meier survival curves were estimated from survival data for each of the 3 studies, and compared with predictions of the simple exponential (null) model vs the 2-state kinetic model, derived as described previously.5 Estimates of model kinetic parameters were obtained using nonlinear regression fit of survival probabilities evaluated at the event times using the nls function, and their standard errors were estimated using the nonparametric bootstrap (with resampling).7,8 A large number (B = 9,999) of simulated bootstrap samples were generated (sampling, with replacement, the pairs of survival times, and censoring indicators), and for each bootstrap sample, the estimated Kaplan-Meier recurrence-free survival probabilities evaluated at the event times were then fit by nonlinear regression to the null and the 2-state models, thus generating the bootstrap versions of the kinetic parameters, with standard errors calculated as the standard deviations of the bootstrap versions of parameter estimates. Self-starting initial values for the exponential and biexponential model were developed by the authors, with no convergence issues.
Since the null (exponential) model is nested in the full model (biexponential model), the comparative goodness-of-fit of the full model was assessed using the Wald test, with the covariance matrix estimated using the nonparametric bootstrap method, as described above. To test whether the 2-state kinetic model (biexponential model) provides a significantly better fit than the Weibull model (not nested in the full model), the estimated R2 measures of goodness of fit for each model were compared using a Wald test, with the variances for the differences in R2 being estimated by the bootstrap. To test for differences in the kinetic parameters between the control and active treatment groups within trials, for comparisons of the effect of active treatment on their values across the trials, and for comparisons of the values of kinetic parameters between pairs of trials, the 2-sided z-tests were used, with standard errors again estimated using the nonparametric bootstrap. The significance level was set at the nominal level of 0.05 for all comparisons.
Data Harmonization for Cross-Trial Comparisons
For comparisons of the derived kinetic rates across trials, we sought to minimize any potential variation stemming from differences in the trial design, such as study end points or active treatments. The first consideration was that, for POINT, time-to-event data were recorded as fractions of days, while in SOCRATES and THALES, time-to-event data were recorded as whole days. An adjusted POINT data set was created using time-to-event data rounded to the nearest whole day. Reanalysis using this adjusted data set showed only small changes in event-free survival curves and in the estimates of kinetic parameters (Supplemental Data). Second, in POINT, subjects were randomized within 12 hours of index event, whereas in SOCRATES and THALES, the time-to-randomization (TTR) could be up to 24 hours. Finally, in SOCRATES and THALES, for the recorded event times, there was by convention a day added to the difference between the date of the event and the date of randomization. Thus, the adjusted SOCRATES and THALES data sets were constructed separating subjects with TTR of 0–12 hours from those with TTR of 12–24 hours and with subtraction of the added day from the recorded time-to-event. Reanalysis and estimation of kinetic parameters of the 2-state model using these adjusted data sets showed that qualitatively similar survival curves and model estimates were derived (eTable 1, links.lww.com/WNL/D184, eFigure 1, links.lww.com/WNL/D183).
Model-Based Prediction
Model-based prediction of event-free survival, based on measured estimates of kinetic parameters, was conducted in Microsoft Excel 2016. For estimates of effects of altered timing of treatment, transitions in a kinetic parameter at times T were modeled with event-free survival for times t < T calculated according to (1) and for times T and greater, according to (1) with the altered kinetic parameters and with substitution of V0’ = 
for V0 and t’ = t – T for t.
for V0 and t’ = t – T for t.Standard Protocol Approvals, Registrations, and Patient Consents
This study was submitted to the University of Chicago Institutional Review Board and was determined to be exempt from further review. It was not registered as a clinical trial, and a formal protocol was not prepared.
Data Availability
All the present results, including the R markdown files with reproducible analyses as well as additional details of the implementation of the methods using the R software, are available from the authors on request.
Results
In each of the SOCRATES, POINT, and THALES trials, in both control and active treatment groups, ischemic stroke was the predominant outcome event, and most events occurred within the first 7 days after trial enrollment (Table 1). Fewer events occurred over the remainder of follow-up in the trials, which extended for 90 days in POINT and SOCRATES and 30 days in THALES. This temporal course suggests a change over time of event risks that might be explained by a transition between underlying clinical states, consistent with our hypothesized 2-state model.
| Trial | N | Entry time window | Trial duration | Primary outcome | Treatment groups | N | Primary outcome Events - total | Primary outcome Events – 7 d | IS outcome events - total | IS outcome events - 7 d |
|---|---|---|---|---|---|---|---|---|---|---|
| SOCRATES | 11,917 | 24 h | 90 d | Stroke (IS or HS), MI, or death | Control: aspirin 100 mg daily | 5,960 | 408 | 246 | 360 | 239 |
| Active: ticagrelor 90 mg bid | 5,957 | 376 | 215 | 325 | 198 | |||||
| POINT | 4,881 | 12 h | 90 d | IS, MI, or vascular death | Control: aspirin 81 mg daily | 2,449 | 160 | 111 | 155 | 104 |
| Active: aspirin 81 mg plus clopidogrel 75 mg daily | 2,432 | 121 | 70 | 112 | 67 | |||||
| THALES | 9,521 | 24 h | 30 d | Stroke (IS or HS) | Control: aspirin 75–100 mg daily | 4,749 | 258 | 197 | 243 | 194 |
| Active: aspirin plus ticagrelor 90 mg bid | 4,772 | 206 | 148 | 185 | 142 |
Sample sizes for shared subject data, entry time window, trial duration, and outcome events stratified by treatment are shown. Primary outcomes included ischemic strokes (IS) in each trial, hemorrhagic stroke (HS) in SOCRATES and THALES, and myocardial infarction (MI) in SOCRATES and POINT. Early (less than 7 d) and total outcomes are shown.
POINT = Platelet-Oriented Inhibition in New TIA and Minor Ischemic Stroke Trial; SOCRATES = Acute Stroke or Transient Ischemic Attack Treated with Aspirin or Ticagrelor and Patient Outcomes; THALES = Acute Stroke or Transient Ischemic Attack Treated with Ticagrelor and ASA for Prevention of Stroke and Death.
In each treatment group and in each trial, the biexponential survivor function predicted by the 2-state model fit the Kaplan-Meier stroke-free survival curves quite closely, far better than did the single-exponential survivor function predicted by the null (1-state) model and better than the Weibull model (Figure 2). Statistical comparisons showed significantly better fit to the clinical data by the 2-state model than by the 1-state exponential model in both control and active treatment groups in each trial (p < 0.001 for all comparisons) and better fit than by the Weibull model (p < 0.05 for all comparisons, Table 2, eTable 2, links.lww.com/WNL/D185).
Kaplan-Meier estimates of event-free survival functions for occurrence of ischemic stroke in control (red) and active treatment (green) groups in POINT, SOCRATES, and THALES trials, with fitted survival functions provided by the null (1-state) model, the Weibull model, and the full (2-state) model. POINT = Platelet-Oriented Inhibition in New TIA and Minor Ischemic Stroke Trial; SOCRATES = Acute Stroke or Transient Ischemic Attack Treated with Aspirin or Ticagrelor and Patient Outcomes; THALES = Acute Stroke or Transient Ischemic Attack Treated with Ticagrelor and ASA for Prevention of Stroke and Death.
| Model | One-state kinetic model | Two-state kinetic model | Two-state model vs 1-state model | Two-state model vs Weibull model | ||
|---|---|---|---|---|---|---|
| kNull (d−1) | k0 (d−1) | k1 (d−1) | k2 (d−1) | |||
| SOCRATES | ||||||
| Control (n = 5,960) | 9.6 × 10−4 | 0.28 ± 0.02 | 0.013 ± 0.001 | 2.0 × 10−4 ± 0.2 × 10−4 | p < 0.001 | p < 0.001 |
| Active (n = 5,957) | 8.8 × 10−4 | 0.26 ± 0.02 | 0.010 ± 0.001 | 2.1 × 10−4 ± 0.2 × 10−4 | p < 0.001 | p = 0.001 |
| POINT | ||||||
| Control (n = 2,449) | 1.3 × 10−3 | 0.62 ± 0.10 | 0.030 ± 0.005 | 2.9 × 10−4 ± 0.5 × 10−4 | p < 0.001 | p = 0.010 |
| Active (n = 2,432) | 7.7 × 10−4 | 0.48 ± 0.10 | 0.015 ± 0.003 | 2.1 × 10−4 ± 0.4 × 10−4 | p < 0.001 | p = 0.02 |
| THALES | ||||||
| Control (n = 4,749) | 2.2 × 10−3 | 0.31 ± 0.02 | 0.014 ± 0.001 | 2.7 × 10−4 ± 0.7 × 10−4 | p < 0.001 | p < 0.001 |
| Active (n = 4,772) | 2.0 × 10−3 | 0.30 ± 0.03 | 0.010 ± 0.001 | 2.1 × 10−4 ± 0.7 × 10−4 | p < 0.001 | p < 0.001 |
POINT = Platelet-Oriented Inhibition in New TIA and Minor Ischemic Stroke Trial; SOCRATES = Acute Stroke or Transient Ischemic Attack Treated with Aspirin or Ticagrelor and Patient Outcomes; THALES = Acute Stroke or Transient Ischemic Attack Treated with Ticagrelor and ASA for Prevention of Stroke and Death.
For comparisons of kinetic rates across trials, data harmonization was required, as described in Methods, to account for the differences in time-to-randomization in SOCRATES and THALES and for the differences in definitions of time-to-event across the trials. Generally, these adjustments did not produce discernable differences in the fitted curves nor in the kinetic rates. Subtraction of an added day for time-to-event data in SOCRATES and THALES, allowing for greater resolution of the timing of very early events, produced somewhat greater values for the estimates of the transition rate parameter k0 (eTable 1, links.lww.com/WNL/D184). Of note, in SOCRATES, there was a distinct separation in event-free survival between the control and active treatment groups only in the subset of subjects randomized within 12 hours of index event, with no suggestion of any treatment effect in those randomized at 12–24 hours. However, this distinction by time-to-randomization was not evident in the THALES data.
Comparisons of rates across the POINT, SOCRATES, and THALES trials, using these adjusted data sets and focusing on control groups, assigned to treatment with aspirin monotherapy in each trial, allowed for comparisons among subjects treated in a similar fashion. Comparison of kinetic rates k0 (the rate of transition from the vulnerable state to the stabilized state), k1 (the rate of recurrence in the vulnerable state), and k2 (the rate of recurrence in the stabilized state), generated under the assumption of V0 = 1 (all subjects starting in the vulnerable state), showed similar values across trials, with small and nonsignificant differences (Figure 3). Estimated values for k0 were between 0.5 d−1 and 1 d−1, corresponding to a predicted half-life of approximately 1–2 days for the fraction of patients remaining in the vulnerable state, while predicted values for k1 were between 0.026 d−1 and 0.037 d−1, corresponding to half-lives of several weeks. The recurrence rates in the stabilized state, given by k2, were found to be 100-fold lower, with values between 2.4 × 10−4 d−1 and 4.7 × 10−4 d−1, yielding half-lives of 4–8 years (Table 3).
Comparisons of estimates (with bootstrap standard errors) of k0, k1, and k2 (estimates ± SE), assuming V0 = 1, in control and active treatment groups across the 3 trials. Rates are plotted on a logarithmic scale for comparison. *Across trials, k1 was significantly less in active treatment groups than in control groups (p < 0.05). POINT = Platelet-Oriented Inhibition in New TIA and Minor Ischemic Stroke Trial; SOCRATES = Acute Stroke or Transient Ischemic Attack Treated with Aspirin or Ticagrelor and Patient Outcomes; THALES = Acute Stroke or Transient Ischemic Attack Treated with Ticagrelor and ASA for Prevention of Stroke and Death.
| k0 (d−1) | k1 (d−1) | k2 (d−1) | |
|---|---|---|---|
| SOCRATES | |||
| Control (5,960) | 0.50 ± 0.09 | 0.026 ± 0.005 | 2.5 × 10−4 ± 0.4 × 10−4 |
| Active (5,957) | 0.36 ± 0.08 | 0.014 ± 0.003 | 2.4 × 10−4 ± 0.5 × 10−4 |
| p = 0.22 | p = 0.027 | p = 0.93 | |
| POINT | |||
| Control (2,449) | 0.53 ± 0.11 | 0.026 ± 0.006 | 2.4 × 10−4 ± 0.4 × 10−4 |
| Active (2,432) | 0.48 ± 0.12 | 0.015 ± 0.004 | 2.0 × 10−4 ± 0.3 × 10−4 |
| p = 0.78 | p = 0.12 | p = 0.37 | |
| THALES | |||
| Control (4,749) | 0.98 ± 0.24 | 0.037 ± 0.010 | 4.7 × 10−4 ± 1.10 × 10−4 |
| Active (4,772) | 0.70 ± 0.19 | 0.019 ± 0.006 | 4.3 × 10−4 ± 1.4 × 10−4 |
| p = 0.37 | p = 0.11 | p = 0.81 | |
| Mean values across trials | |||
| Control | 0.67 ± 0.09 | 0.030 ± 0.004 | 3.2 × 10−4 ± 0.4 × 10−4 |
| Active | 0.51 ± 0.08 | 0.016 ± 0.002 | 2.9 × 10−4 ± 0.5 × 10−4 |
| Comparison between treatment groups across all trials | p = 0.28 | p = 0.004 | p = 0.45 |
| Comparisons of rates in control groups between trials | |||
| POINT vs SOCRATES | p = 0.84 | p = 0.99 | p = 0.94 |
| POINT vs THALES | p = 0.09 | p = 0.34 | p = 0.05 |
| SOCRATES vs THALES | p = 0.06 | p = 0.31 | p = 0.06 |
Estimates and bootstrap estimates of standard errors (SE) are shown for each trial.
POINT = Platelet-Oriented Inhibition in New TIA and Minor Ischemic Stroke Trial; SOCRATES = Acute Stroke or Transient Ischemic Attack Treated with Aspirin or Ticagrelor and Patient Outcomes; THALES = Acute Stroke or Transient Ischemic Attack Treated with Ticagrelor and ASA for Prevention of Stroke and Death.
The kinetic parameters were compared between the control and active treatment groups in each trial, and for k0 and k2, no significant differences were found. By contrast, there were significant or near-significant decreases in the value of k1 in active treatment vs control groups in each trial, and a significant decrease in the average k1 across the trials (Table 3; control k1 = 0.030 ± 0.004 vs active treatment k1 = 0.016 ± 0.003, p < 0.01). This result points to an effect of the active treatments with enhanced antiplatelet therapy only on the rates of stroke recurrence in the patients in the vulnerable state, without a significant effect on the rates in the patients in the stabilized state, nor on the rates of transition from vulnerable to stabilized states.
This model provides for prediction of the effects of delays in the initiation of treatment or changes in the duration of treatments that may alter one kinetic parameter or another. For example, if enhanced antiplatelet therapy indeed only affects k1, decreasing its value from approximately 0.03 d−1 to 0.016 d−1, with k0 and k2 estimated as 0.6 d−1 and 0.0003 d−1, as suggested by the present results, effects of the active antiplatelet therapy are quite transient. Using these estimated values and effect sizes and under the assumption that all the patients start in the vulnerable state (i.e., V0 = 1), immediate and continuous enhanced antiplatelet treatment improves fractional event-free survival at 90 days from 0.927 to 0.948. Delaying the initiation by 1, 2, or 3 days diminishes this to approximately 53%, 28%, and 15%, respectively, of the full treatment benefit (Figure 4). On the other hand, limiting the duration of antiplatelet treatment to 1, 3, or 7 days provides approximately 46%, 84%, and 98%, respectively, of the effects of treatment for the full 90 days.
Predicted effects of delayed start of enhanced antiplatelet therapy (A) or limited duration of enhanced antiplatelet therapy (B), using estimates for kinetic parameters taken from the present results (k0 = 0.6 d−1, k1 = 0.03 d−1 in control group and 0.016 d−1 in active treatment group, and k2 = 0.0003 d−1, with V0 = 1).
Discussion
Recent trials of acute secondary prevention with antiplatelet agents, including SOCRATES, POINT, and THALES, by virtue of rapid randomization of subjects after index events, have illuminated the very high early rate of stroke recurrence after initial minor stroke or TIA. The clinical reasons for the enhanced early risk have not been fully explained but can be ascribed to features, such as unstable intravascular thrombus, inflamed activated plaque, or critical arterial stenosis.9 These cited trials share similarities in the study design and show striking similarities in qualitative temporal features of event-free survival curves, with rapid early rates of recurring events slowing to a much lower long-term rate. Our study demonstrates that in each trial, these features support a model consisting of a short-lived vulnerable state rapidly transitioning to a stabilized state, with kinetic rates characterizing these transitions that are quantitatively similar across trials. The 2-state kinetic model produces excellent fits of clinical survival data for each trial, superior to the fits produced by either the null kinetic (exponential) model with a single underlying state or the Weibull model. The Weibull model, while more flexible than the 1-state model, does not allow for a physiologic interpretation of the parameters and does not provide as adequate a fit as does the 2-state model. The rejection of the null model provides strong evidence against the assumption of one clinical state after index stroke or TIA that is uniform over time. An early transient vulnerable state during the first week after ischemic stroke has previously been inferred from other observations, such as by detection of recurrent ischemic lesions by diffusion-weighted imaging.9 Heightened awareness of this transient vulnerable state can inform future randomized clinical trial design in secondary stroke prevention, when planning exposure duration and timing of outcome assessments. Furthermore, it may guide treatment optimization in the clinical setting.
The kinetic parameters were similar in the aspirin-treated control groups across trials after differences in trial designs were corrected, suggesting that common underlying pathophysiologic states were represented in the different trial populations. Although availability of data did not allow for the CHANCE1 and CHANCE 210 trials to be analyzed in parallel, inspection of published survival curves in these trials suggests that they also follow a temporal course suggestive of 2 clinical states, with qualitatively similar kinetics.
As hypothesized, the kinetic rates of stroke recurrence for the vulnerable state (k1) in the active treatment groups with enhanced antiplatelet regimens were significantly lower than the rates in the control groups, while the kinetic rates k0 and k2 were not found to be significantly different between the active treatment and control groups. This finding suggests that the intensity of antiplatelet therapy primarily affects the recurrence rate in the vulnerable state and does not affect the recurrence rates in the stabilized state nor the rates of transition from the vulnerable to stabilized states. There are likely to be other mechanisms that also affect these rates, including the type of stroke, age, and comorbidities, and this model may be refined to test the effects of these risk factors on the kinetic parameters. It is intriguing that in SOCRATES, although the overall effect of treatment on the primary trial end point did not achieve significance, the present analysis suggests that treatment assignment to the rapidly acting antiplatelet agent ticagrelor had a potentially significant effect only in the subset of subjects randomized within 12 hours of the index stroke or TIA, supporting an effect of the active treatment when administered quickly, within 12 hours from the index event.
A clinical implication of the present analysis is that the mechanisms underlying short-term recurrence of stroke, reflected by k1, are not necessarily the same as those for longer-term recurrence, reflected by k2, and therapeutic strategies effectively targeting acute secondary prevention in the days after minor stroke or TIA may not be the same as those that will be optimal in long-term secondary prevention. Already many centers, when applying the CHANCE or POINT trial results to practice, follow implications of results of previous time course analyses11,12 and recommend aspirin and clopidogrel for only 1–3 weeks after minor stroke or TIA rather than for the 90-day period used in the POINT protocol. Even protective effects of aspirin monotherapy in comparison with control, in a large analysis of pooled data from multiple trials, predominately occurred in the initial 6 weeks after minor stroke or TIA.13 Different antithrombotic regimens may be more appropriate for long-term secondary stroke recurrence prevention. Trials comparing dual antiplatelet strategies against antiplatelet monotherapy for long-term secondary prevention have generally failed to demonstrate convincing benefit,14,15 suggesting counterbalancing risks from the combination of clopidogrel and aspirin as compared with monotherapy.16,17 Positive trial results favoring dual antiplatelet therapy over the long-term are largely restricted to support for addition of cilostazol to aspirin or clopidogrel.18 Further trials, incorporating an arm with antiplatelet monotherapy, are needed to identify the best antiplatelet strategies to reduce the long-term risks after a minor stroke event, reflected in the k2 rate.
The hypothesized vulnerable state producing similar rates of early stroke recurrence across trials may result in some cases from active inflammation or ulcerated plaques in supplying vessels, perhaps with superimposed platelet-rich thrombus, in patients with underlying large-artery atherosclerosis. There is suggestion that enhanced antiplatelet therapies in these trials have the greatest efficacy in this subgroup.19 The very short half-life of this vulnerable state implies a surprisingly rapid resolution of this vulnerability. Treatments aiming to further shorten the duration of the vulnerable state by suppression of plaque inflammation or promotion of endothelial recovery would need to be delivered immediately and to act very rapidly to effectively reduce the duration of vulnerability.
A significant limitation of the present quantitative analysis is the inability to simultaneously determine V0, k0, and k1 from the survival data alone. If some alternative means could provide an independent estimate of V0 (the fraction of subjects initially in the vulnerable state), definitive estimates of k0, k1, and k2 would be possible. Still, even without direct estimation of V0, the model allows for unambiguous determination of the recurrence rate in the stabilized state k2 and the ratio of k1 between the active and control groups and shows that active treatment significantly decreases k1. Another limitation is the shorter duration of clinical observation in the THALES trial, limiting the precision in the estimation of k2 for THALES, and possibly indirectly affecting the estimations of the other rates.
In conclusion, kinetic analysis of acute secondary stroke prevention trials shows that after minor stroke or TIA, patients experience a vulnerable state that transitions rapidly to a more stabilized state, with distinct pathobiologies and treatment responses likely associated with these distinct states. More intensive antiplatelet therapy regimens primarily protect against recurrence of stroke in the vulnerable state and are quite transient in effect. In the stabilized state, there is no detectable benefit of enhanced antiplatelet therapy and risks of bleeding likely outweigh any potential benefit. The results support a practice of very rapid introduction of enhanced antiplatelet therapy after initial minor stroke or TIA and also support limiting its duration to the short term. Other treatment strategies may be needed to alter risks of recurrence over the long-term, in the stabilized state. Awareness of the kinetic features of this temporal shift may inform design of future randomized clinical trials when determining the duration of interventions and the timing of outcome assessments.
Glossary
- CHANCE
- Clopidogrel in High-Risk Patients with Nondisabling Cerebrovascular Events
- POINT
- Platelet-Oriented Inhibition in New TIA and Minor Ischemic Stroke Trial
- SOCRATES
- Acute Stroke or Transient Ischemic Attack Treated with Aspirin or Ticagrelor and Patient Outcomes
- TIA
- transient ischemic attack
- THALES
- Acute Stroke or Transient Ischemic Attack Treated with Ticagrelor and ASA for Prevention of Stroke and Death
- TTR
- time to randomization
Acknowledgment
The authors are grateful to James E. Siegler, MD, for providing helpful comments regarding a late draft of this manuscript. The authors thank the National Institutes of Neurological Disorders and Stroke for provision of the POINT trial data set and the sponsors of the SOCRATES and THALES trials for approval of access to these trial data sets. This manuscript is based on research using data from data contributor AstraZeneca that has been made available through Vivli, Inc. Vivli has not contributed to or approved and is not in any way responsible for the contents of this publication.
Appendix Authors
| Name | Location | Contribution |
|---|---|---|
| James R. Brorson, MD | Department of Neurology, The University of Chicago, IL | Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data |
| Mihai Giurcanu, PhD | Department of Public Health Sciences, The University of Chicago, IL | Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data |
| Shyam Prabhakaran, MD | Department of Neurology, The University of Chicago, IL | Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data |
| S. Claiborne Johnston, MD, PhD | CMO and cofounder Harbor Health, Austin, TX | Major role in the acquisition of data; analysis or interpretation of data |
Footnote
Editorial, page 981
References
1.
Wang Y, Wang Y, Zhao X, et al.; CHANCE Investigators. Clopidogrel with aspirin in acute minor stroke or transient ischemic attack. N Engl J Med. 2013;369(1):11-19.
2.
Johnston SC, Amarenco P, Albers GW, et al. Ticagrelor versus aspirin in acute ischemic stroke or transient ischemic attack. N Engl J Med. 2016;375(1):35-43.
3.
Johnston SC, Easton JD, Farrant M, et al.; Clinical Research Collaboration, Neurological Emergencies Treatment Trials Network, and the POINT Investigators. Clopidogrel and aspirin in acute ischemic stroke and high-risk TIA. N Engl J Med. 2018;379(3):215-225.
4.
Johnston SC, Amarenco P, Denison H, et al. Ticagrelor and aspirin or aspirin alone in acute ischemic stroke or TIA. N Engl J Med. 2020;383(3):207-217.
5.
Brorson JR, Bulwa ZB. Two-state kinetic model of rates of stroke recurrence in the POINT study population. Stroke. 2021;52(4):1446-1449.
6.
Diamond GA, Kaul S. Forbidden fruit: on the analysis of recurrent events in randomized clinical trials. Am J Cardiol. 2013;111(10):1530-1536.
7.
Efron B. Computer and the theory of statistics: thinking the unthinkable. SIAM Rev. 1979;21(4):460-480.
8.
Efron B. Censored data and the bootstrap. J Am Stat Assoc. 1981;76(374):312-319.
9.
Kang D-W, Latour LL, Chalela JA, et al. Early ischemic lesion recurrence within a week after acute ischemic stroke. Ann Neurol. 2003;54(1):66-74.
10.
Wang Y, Meng X, Wang A, et al. Ticagrelor versus clopidogrel in CYP2C19 loss-of-function carriers with stroke or TIA. N Engl J Med. 2021;385(27):2520-2530.
11.
Pan Y, Jing J, Chen W, et al. Risks and benefits of clopidogrel-aspirin in minor stroke or TIA. Time course analysis of CHANCE. Neurology. 2017;88(20):1906-1911.
12.
Johnston SC, Elm JJ, Easton JD, et al. Time course for benefit and risk of clopidogrel and aspirin after acute transient ischemic attack and minor ischemic stroke. A secondary analysis from the POINT randomized trial. Circulation. 2019;140(8):658-664.
13.
Rothwell PM, Algra A, Chen Z, et al. Effects of aspirin on risk and severity of early recurrent stroke after transient ischaemic attack and ischaemic stroke: time-course analysis of randomised trials. Lancet. 2016;388(10042):365-375.
14.
Sacco RL, Diener H-C, Yusuf S, et al. Aspirin and extended-release dipyridamole versus clopidogrel for recurrent stroke. N Engl J Med. 2008;359(12):1238-1251.
15.
Benavente OR, Hart RG, McClure LA, et al.; SPS3 Investigators. Effects of clopidogrel added to aspirin in patients with recent lacunar stroke. N Engl J Med. 2012;367(9):817-825.
16.
Diener HC, Bogousslavsky J, Brass LM, et al.; MATCH Investigators. Aspirin and clopidogrel compared with clopidogrel alone after recent ischaemic stroke or transient ischaemic attack in high-risk patients (MATCH): randomised, double-blind, placebo-controlled trial. Lancet. 2004;364(9431):331-337.
17.
Bhatt DL, Fox KA, Hacke W, et al.; CHARISMA Investigators. Clopidogrel and aspirin versus aspirin alone for the prevention of atherothrombotic events. N Engl J Med. 2006;354(16):1706-1717.
18.
Toyoda K, Uchiyama S, Yamaguchi T, et al.; CSPS.com Trial Investigators. Dual antiplatelet therapy using cilostazol for secondary prevention in patients with high-risk ischaemic stroke in Japan: a multicentre, open-label, randomised controlled trial. Lancet Neurol. 2019;18(6):539-548.
19.
Amarenco P, Albers GW, Denison H, et al. Efficacy and safety of ticagrelor versus aspirin in acute stroke or transient ischaemic attack of atherosclerotic origin: a subgroup analysis of SOCRATES, a randomized, double-blind, controlled trial. Lancet Neurol. 2017;16(4):301-310.
Information & Authors
Information
Published In
Copyright
Copyright © 2023 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the American Academy of Neurology. This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License 4.0 (CC BY-NC-ND), which permits downloading and sharing the work provided it is properly cited. The work cannot be changed in any way or used commercially without permission from the journal.
Publication History
Received: April 26, 2023
Accepted: August 15, 2023
Published online: October 19, 2023
Published in issue: November 28, 2023
Disclosure
The authors report no relevant disclosures. Go to Neurology.org/N for full disclosures.
Study Funding
This study was supported in part by a gift from the estate of Bendt Bladel.
Authors
Metrics & Citations
Metrics
Citation information is sourced from Crossref Cited-by service.
Citations
Download Citations
If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Select your manager software from the list below and click Download.
Cited By
- Understanding the Relationship Between Cerebrovascular Disease and the Gut Microbiome, Stroke: Vascular and Interventional Neurology, 5, 1, (2025).https://doi.org/10.1161/SVIN.124.001272
- Stroke and high-risk TIA outcomes with reduction of treatment duration when treatment initiated in emergency rooms (SHORTER-study), International Journal of Stroke, 19, 7, (830-834), (2024).https://doi.org/10.1177/17474930241237120
- Is Recurrent Stroke More Like a Campfire or a Light Switch?, Neurology, 101, 22, (981-982), (2023)./doi/10.1212/WNL.0000000000208023
Loading...
View Options
View options
PDF and All Supplements
Download PDF and Supplementary MaterialShort Form
Short FormFull Text
View Full TextLogin options
Check if you have access through your login credentials or your institution to get full access on this article.
Personal login Institutional LoginPurchase Options
The neurology.org payment platform is currently offline. Our technical team is working as quickly as possible to restore service.
If you need immediate support or to place an order, please call or email customer service:
- 1-800-638-3030 for U.S. customers - 8:30 - 7 pm ET (M-F)
- 1-301-223-2300 for customers outside the U.S. - 8:30 - 7 pm ET (M-F)
- [email protected]
We appreciate your patience during this time and apologize for any inconvenience.