SGEM#325: Thin Ice – Subgroup Analysis of the THAWS Trial - The Skeptics Guide to Emergency Medicine

Subscribe: RSS

Date: March 31st, 2021

Guest Skeptic: Prof Daniel Fatovich is an emergency physician and clinical researcher based at Royal Perth Hospital, Western Australia. He is Head of the Centre for Clinical Research in Emergency Medicine, Harry Perkins Institute of Medical Research; Professor of Emergency Medicine, University of Western Australia; and Director of Research for Royal Perth Hospital.

Reference: Toyoda et al. MRI-guided thrombolysis (0.6 mg/kg) was beneficial for unknown onset stroke above a certain core size. THAWS RCT Substudy. Stroke 2021

Case: A 74-year-old man presents to the emergency department after waking up with left sided weakness. He was last seen well when going to bed at 10pm the night before. He has a history of hypertension and dyslipidemia. His medications include an angiotensin-converting enzyme inhibitor and a statin. The NIHSS score is 7. The MRI shows an occlusion of the right MCA-M2, the DWI-ASPECT is 9, and lesion volume is 3.5ml.

Background: We have talked about stroke management a number of times recently including SGEM#297 on the reanalysis of ECASS-3 by Alper et al 2020. The SGEM bottom line was that the “reanalysis of the original ECASS-3 data does not support the potential benefit of tPA given between 3-4.5 hours after onset of stroke symptoms and confirms the known potential harm”.

There have been 13 foundational trials looking at thrombolysis for acute ischemic stroke. Of the 13, eleven failed to show benefit for their primary outcome and four were stopped early due to harm or futility. Only two RCTs claimed benefit for their primary outcome. Those were ECASS-3 in 2008 and the NINDS trial from 1995. Both of those “positive” studies have been reanalyzed and question the potential efficacy while confirming the potential harm.

Dr. Jackson

We wrote an article together for the Lown Institute summarizing some of the stroke literature. The question asked was: will it take 50 or 100 years to get the right answer about tPA for acute ischemic stroke?

One aspect that we did not address was the newer trials that are using advanced imaging techniques like MRI to extend the window beyond 4.5 hours after the onset of stroke symptoms (Extend NEJM 2019 and ECASS-4: Extend 2016). Both of these trials were stopped early which can introduce additional bias towards efficacy.

The majority of patients included in the two trials extending the time window past 4.5 hours would now qualify for endovascular therapy (EVT) clot retrieval. EVT does have more robust evidence for efficacy and safety than systemic thrombolysis.

A SRMA was published by Mistry et al Stroke 2017. This included 13 studies, three randomized control trials (25% of all patients) and ten observational studies (75% of all patients). Good neurologic outcome was defined as a modified Rankin Scale (mRS) score of 0-2. The number needed to treat (NNT) was 17. However, there was no statistical difference if you only look at the higher quality RCT data and excluded the lower quality observational data.

Yang P et al. published a non-inferiority RCT in NJEM 2020 looking at this issue. The primary outcome was mRS at 90 days and found EVT alone was not non-inferior to EVT plus tPA. Two recent RCTs were published in JAMA investigating this issue. Suzuki et al failed to demonstrate non-inferiority while in contrast Zi et al found EVT alone was non-inferior to EVT plus tPA. These two EVT trials are going to be covered on a future episode of the SGEM in the near future.

The trial we are reviewing today is a sub analysis of the THAWS (Thrombolysis for Acute Wake-Up and Unclear-Onset Stroke) randomized control trial of using low dose tPA in patients with symptoms on awaking or unknown time of onset.

Clinical Question: Is MRI guided thrombolysis (0.6 mg/kg) beneficial for patients with an unknown stroke onset time?

Reference: Toyoda et al. MRI-guided thrombolysis (0.6 mg/kg) was beneficial for unknown onset stroke above a certain core size. THAWS RCT Substudy. Stroke 2021.

Population: Patients with stroke symptoms on awaking or with unknown time of onset (greater than 4.5 hours since last known well and less than 4.5 hours of symptom recognition). Substudy of THAWS published 2020 (n = 131).
Intervention: IV alteplase 0.6 mg/kg (10% bolus followed by 90% infusion over 60 minutes)
Comparison: Standard care, not placebo controlled. Standard care was the use of one to three antithrombotic drugs, including oral aspirin (160–300 mg/day), oral clopidogrel (75 mg/day), intravenous argatroban, or intravenous unfractionated heparin, but excluding the combination of argatroban and heparin, according to decisions of the attending physician. (Argatroban is an anticoagulant that is a small molecule direct thrombin inhibitor). In this SUBSTUDY (n= 126), patients were dichotomized by ischemic core size or NIHSS.
Outcome:
- Primary Outcome: Good neurologic function using modified Rankin Scale (mRS) score of 0-1 at 90 days
- Secondary Outcomes: Category (ordinal) shift in mRS at 90 days; mRS 0-2 at 90 days; category shift NIHSS score at 24 hours and 7 days; imaging outcomes: recanalization on MRA at 22-36 hours; infarct volume on FLAIR at 7 days minus infarct volume on DWI at baseline. Safety: sICH at 22-36 hours and major extracranial bleeding and death.

Authors’ Conclusions: SUBSTUDY: “Patients developing unknown onset stroke with DWI-ASPECTS 5 to 8 showed favorable outcomes more commonly after low-dose thrombolysis than after standard treatment.”

Note: “ASPECTS” stands for the Alberta Stroke Program Early CT Score. It is used to determine MCA stroke severity using available CT data. To compute ASPECTS, 1 point is subtracted from 10 for any evidence of early ischaemic change for each of the defined regions. Normal CT = 10 points.

Using the traditional cutoff (less than 8 vs equal to or greater than 8) as a rough estimate for predicting independence may help inform decisions. ASPECTS suggests that early CT changes in stroke may be a harbinger of poor outcomes.

As a reminder, the authors’ conclusions from the original THAWS RCT were: “No difference in favorable outcome was seen between alteplase and control groups among patients with ischemic stroke with unknown time of onset. The safety of alteplase at 0.6 mg/kg was comparable to that of standard treatment. Early study termination precludes any definitive conclusions.”

Quality Checklist for Randomized Clinical Trials:

The study population included or focused on those in the emergency department. Yes
The patients were adequately randomized. Yes
The randomization process was concealed. No
The patients were analyzed in the groups to which they were randomized. Yes
The study patients were recruited consecutively (i.e. no selection bias). Unsure
The patients in both groups were similar with respect to prognostic factors. No
All participants (patients, clinicians, outcome assessors) were unaware of group allocation. No
All groups were treated equally except for the intervention. No
Follow-up was complete (i.e. at least 80% for both groups). Yes
All patient-important outcomes were considered. Yes
The treatment effect was large enough and precise enough to be clinically significant. No

Results: The original THAWS trial enrolled 131 patients. It was to have 300 patients but was stopped early due to results from the WAKE-UP trial. The mean age was 74 years, 69% had hypertension, 35% dyslipidemia, 37% atrial fibrillation, 20% diabetic and 17% had a history of ischemic stroke/TIA.

THAWS Subgroup Analysis: Some subgroups had better neurologic outcomes with tPA compared to control while others did not.

Primary Outcome: Good neurologic outcome (mRS 0-1) at 90 days
- DWI-ASPECTS 5 to 8 (RR 4.75 [95% CI, 1.33–30.2]) * (statistically significant)
- DWI-ASPECTS 9 to 10 (RR 68 [95% CI, 0.45–1.02])
- Core volume >4 ml (RR 6.15 [95% CI, 0.87–43.64])
- Core volume ≤6.4 ml (RR 81 [95% CI, 0.57–1.17])

THAWS: No statistical difference in good neurologic outcome at 90 days.

Primary Outcome: Good neurologic outcome (mRS 0-1) at 90 days
- 47.1% tPA vs 48.3% control
- RR 97 [95% CI; 0.68–1.41] P=0.892

Secondary Outcomes:
- All p > 0.05 except recanalization of culprit artery on MRA 73.7% tPA vs 40.9% control.
- Intracranial hemorrhage (ICH) was 26% tPA vs 14% control.

ASPECT Data: The baseline DWI-ASPECTS was 10 in 38 patients, 9 in 41, 8 in 18, 7 in 11, 6 in 11, and 5 in 7, with a median of 9. So, only a small subgroup of 47 patients had an ASPECT of 5-8. Remember that the THAWS trail was planned to have 300 patients and was stopped early at 131 patients. This subgroup represents 37% of the cohort and only 16% of the population determined a priori.

They also dichotomized patients by core volume with a cutoff level based on the receiver operating characteristic (ROC) curve ROC of 6.4 ml. They observed some baseline differences between these two groups.

Core Volume Data: Patients with core volume >6.4 ml more commonly had atrial fibrillation (P=0.014), less commonly recognized symptoms at waking up (P=0.007), had higher NIHSS score (P<0.001), and had lower DWI-ASPECTS (P<0.001) than those with volume ≤6.4 ml. These baseline differences are summarized in Table 1.

With this dichotomization based on CORE volume the percentage of favourable outcome was not different between the tPA and control groups (P=0.376). In patients with volume >6.4 mL, although not statistically significant, favourable outcome was more common in the tPA group than in the control group (RR, 6.15 [95% CI, 0.87–43.64] P=0.069). Change in the NIHSS score was larger in the control group for patients with volume ≤6.4 ml and category shift in the mRS score was larger in the tPA group for those with volume >6.4 ml.

1) WAKE Up “Positive”? The authors frequently refer to the “positive” WAKEUP study (Thomalla et al. NEJM 2018). This study stopped enrolling at 503 patients, despite a target of 800. They reported a 90-day mRS 0-1 of 53% tPA vs 42% placebo (median NIHSS 6, lesion volume 2.0 mL), but deaths were 4.1% tPA vs 1.2% placebo (p 0.07). This is a clinically important difference, and it can be argued that the study was stopped before they could prove harm.

The THAWS study states their design was “similar” to WAKEUP, but there was no placebo arm in THAWS. That’s a massive difference. Also, in THAWS, they changed eligibility criteria during the trial (eg 12-hour time limit to no time limit to match WAKEUP).

It is noteworthy that in the meta-analysis by Thomalla on this subject the authors promote a positive role for tPA in these patients, and they also state that there are more deaths with tPA (Thomalla et al. Lancet 2020).

2) Dose of tPA: The dose of tPA was 0.6 mg/kg in the THAWS trial compared to the vs usual 0.9 mg/kg. The authors state: “The official dosage of alteplase for stroke thrombolysis is 0.6 mg/kg in Japan based on the results of a dose comparison trial using duteplase and the multicenter single-dose Japan Alteplase Clinical Trial. Concern for a risk of thrombolysis-associated intracranial hemorrhage (ICH) in Asians was strong motivation for the trials of low-dose thrombolysis. Several Asian physicians also used alteplase at 0.6 mg/kg from the perspective of ethnical similarity to Japanese, safety and economy.”

One of the citations used is an open label study of 0.6 mg/kg alteplase on 100 patients within 3 hours of symptom onset, (with a power calculation based upon 90% confidence intervals). Anyway, the tPA dose in WAKEUP was 0.9mg/kg vs their 0.6 mg/kg, another point of difference. Yet in THAWS, they state: “The ENCHANTED (Enhanced Control of Hypertension and Thrombolysis Stroke Study) revealed that the small difference of efficacy and overall similarity regarding the balance between efficacy and safety between 0.9 and 0.6 mg/kg of alteplase though this trial did not reach its primary end point of noninferiority.”

The ENCHANTED conclusion was: “This trial involving predominantly Asian patients with acute ischemic stroke did not show the noninferiority of low-dose alteplase to standard-dose alteplase with respect to death and disability at 90 days. There were significantly fewer symptomatic intracerebral hemorrhages with low-dose alteplase.” Despite its statistical failure, the ENCHANTED trial demonstrates that reduced dose tPa achieves similar clinical outcomes while decreasing death and ICH.

3) Heterogeneity: Just a reminder that there is so much heterogeneity in the stroke trials; both clinical heterogeneity and statistical heterogeneity. There are multiple different types of strokes and multiple types of statistical analytic approaches. Remember that the re-analysis of the ECASS-3 trial in 2020 by Alper et al found that we rarely consider the selective analysis reporting bias that can occur in the selection of methods for statistical analysis. So basically, people tend to use the analytic approach that gives them the answer they want to get.

4) Stopping Early: There are problems with stopping studies early. A good paper on this topic is by Guyatt et al BMJ 2012. This article highlights that studies stopped early for benefit overestimate benefits, especially if it’s a small trial. “Uncritical belief in truncated trials will be misleading.” They advocate for a high level of skepticism on the findings of trials stopped early and bemoans how these studies are often published in high profile journals with uncritical uptake by guideline panels and the media.

Interestingly, Wallach et al JAMA Intern Med 2017, evaluated how often subgroup claims are corroborated by subsequent RCT and meta-analyses. They concluded that: “Attempts to corroborate statistically significant subgroup differences are rare; when done, the initially observed subgroup differences are not reproduced.” Yusuf et al JAMA 1991 stated: “the overall trial result is usually a better guide to the direction of effect in subgroups than the apparent effect observed within a subgroup.”

5) Conflicts of Interest (COI): There were multiple financial conflicts of interest in the THAWS study, with many authors receiving payments from the manufacturer of alteplase. This is a recurring theme in the stroke literature and industry-supported studies are associated with more favorable results and conclusions than sponsorship by other sources (Lundh et al Cochrane 2012 and Hansen et al Cochrane 2019). The Institute of Medicine published guidelines on systematic reviews and meta-analyses and recommends the exclusion of “individuals with a clear financial conflict of interest” (Graham et al National Academies Press 2011).

These COIs do not make the information presented incorrect, but it does make us more skeptical.

Comment on Authors’ Conclusion Compared to SGEM Conclusion: We would re-write their conclusions to say that “A subgroup analysis of the THAWS data suggests a patients’ with DWI-ASPECTS 5 to 8 treated with low-dose thrombolysis compared to standard treatment may/may not have a more favorable”.

SGEM Bottom Line: This subgroup analysis of the THAWS data is an interesting finding but should be considered hypothesis generating and would need to be confirmed in a proper trial.

Case Resolution: The stroke service is called, and the patient is admitted to the stroke unit under their care.

Dr. Daniel Fatovich

Clinical Application: The kindest interpretation of this subgroup analysis of a small, unblinded trial that was “negative”, that was stopped early, is that it is hypothesis generating. The literature on replication of hypothesis generating subgroup analyses indicates that pursuit of this is unlikely, and if achieved, rarely reproduced.

What Do I Tell the Patient? You’ve had a stroke. It was caused by a blockage to one of the blood vessels in the brain. All strokes are bad but this one is at the milder end of the spectrum. We’re going to call the stroke team who will probably admit you to the stroke unit for therapy to help get you better.

Keener Kontest: Last weeks’ winner was Dr. Cindy Bitter from St. Louis University. Dr. Bitter is a repeat winner. She knew the main driver of hypoxia in flail chest is pulmonary contusion.

Listen to the podcast on iTunes for this weeks’ keener question. If you think you know the answer then send an email to TheSGEM@gmail.com with “keener” in the subject line. The first correct answer will receive a cool skeptical prize.