Journal Club: August 2016
Intensive Blood-Pressure Lowering in Patients with Acute Cerebral Hemorrhage (NCBI Link)
Author’s Question: what is the utility of rapidly lowering BP in patients with ICH?
Method: randomly assigned eligible participants with [spontaneous] intracerebral hemorrhage (volume, <60 cm3) and a Glasgow Coma Scale (GCS) score of 5 or more to a systolic blood-pressure target of 110 to 139 mm Hg (intensive treatment) or a target of 140 to 179 mm Hg (standard treatment) in order to test the superiority of intensive reduction of systolic blood pressure to standard reduction; intravenous nicardipine to lower blood pressure was administered within 4.5 hours after symptom onset. The primary outcome was death or disability (modified Rankin scale score of 4 to 6, on a scale ranging from 0 [no symptoms] to 6 [death]) at 3 months after randomization.
Results: primary outcome of death or disability was observed in:
38.7% of the participants (186 of 481) in the intensive treatment group
37.7% (181 of 480) in the standard-treatment group
(relative risk, 1.04; 95% confidence interval, 0.85 to 1.27; analysis was adjusted for age, initial GCS score, and presence or absence of intraventricular hemorrhage).
Serious adverse events occurring within 72 hours after randomization that were considered by the site investigator to be related to treatment were reported in 1.6% of the patients in the intensive treatment group and in 1.2% of those in the standard-treatment group.
Author’s Conclusion: treatment of participants with intracerebral hemorrhage to achieve a target systolic blood pressure of 110 to 139 mm Hg did not result in a lower rate of death or disability than standard reduction to a target of 140 to 179 mm Hg.
Discussion: limitations of the study were that pts were decreased to SBP of 180 prior to treatment group. There was also concern that more than 50% of the study was performed outside the country, questioning the generalizability to our population. There was a significant amount of participants in this study to validate the results.
Bottom line: INTERACT trial showed it is safe to drop BP in pts with ICH however did not address if there is benefit. This study showed that there is no benefit and possible harm to kidney when the BP is dropped too quickly. Given these findings, BP should not be dropped quickly in pt’s with ICH due to potential harm to kidney with no significant benefit in mortality.
JC Question: Interpret figure 3. Your answer should explain how to interpret relative risk and confidence intervals.
Relative risk (RR) is the incidence of an outcome in patients with an exposure compared to incidence of an outcome in patients without the exposure. This statistic is used to report associations in tests that can confidently demonstrate causality, such as randomized controlled trials. In this figure, RR is displayed as a small box, which has a line representing the confidence interval for the relative risk running through it. In this table RR represents risk of death or disability at 3 months in patients who underwent intensive BP management compared to patients who underwent standard treatment. For patients with a GCS 3-11 and 12-14, the RR=0.9 and 1.16, respectively. Patients with a lower GCS seem to have a lower incidence of the bad outcome in the intensive treatment group since the RR is less than one. Interestingly, patients with a higher GCS seem to have higher incidence of the bad outcome since the RR is greater than one.
The confidence interval gives a range of where the “true” result is expected to exist. The authors report 95% confidence intervals, meaning they are 95% confident that the true relative risk falls somewhere in the range they describe. For patients with a GCS 3-11, the confidence interval for the RR spans 0.71 – 1.14. Since the “true” value of RR includes 1, these results do not statistically demonstrate that intensive treatment is associated with a reduction in bad outcomes.
Platelet transfusion versus standard care after acute stroke due to spontaneous cerebral haemorrhage associated with antiplatelet therapy (PATCH): a randomised, open-label, phase 3 trial (NCBI Link)
Author’s Question: whether platelet transfusion with standard care, compared with standard care alone, reduced death or dependence after intracerebral hemorrhage associated with antiplatelet therapy use.
Method: multicenter, open-label, masked-endpoint, randomized trial at 60 hospitals in the Netherlands, UK, and France. The study enrolled adults >18yo within 6 h of supratentorial intracerebral hemorrhage symptom onset if they had used antiplatelet therapy for at least 7 days beforehand and had a Glasgow Coma Scale score >8. Participants were randomly assigned to receive either standard care or standard care with platelet transfusion within 90 min of diagnostic brain imaging. The primary outcome was shift towards death or dependence rated on the modified Rankin Scale (mRS) at 3 months.
Results: 190 participants with 97 participants randomly assigned to platelet transfusion and 93 to standard care.
The odds of death or dependence at 3 months were higher in the platelet transfusion group than in the standard care group (adjusted common odds ratio 2·05, 95% CI 1·18–3·56; p=0·0114).
40 (42%) participants who received platelet transfusion had a serious adverse event during their hospital stay, as did 28 (29%) who received standard care. 23 (24%) participants assigned to platelet transfusion and 16 (17%) assigned to standard care died during hospital stay.
Author’s Conclusion: platelet transfusion is inferior to standard care for people taking antiplatelet therapy before intracerebral hemorrhage, and therefore cannot be recommended for this indication in clinical practice.
Discussion: This study used both intention to treat and as treated analysis resulting in very minimal change in statistical outcome. It is also concerning that the study did not clearly identify the “standard care” or include BP when factoring in results. This study did provide unbiased results.
Bottom line: Don’t give platelets to pt’s on antiplatelet therapy with ICH. Platelet transfusion is associated with adverse events and even death. The mechanism is unknown however could be related to increased thrombosis having a negative effect on collateral perfusion or the proinflammatory side effect or increase in volume associated with giving product.
JC Question: Figure 1 shows the treatment allocation and analysis profile for the patients in this study. Note that some patients randomized to each group did not receive the allocated treatment. What methods can investigators use to deal with protocol violations of this type? What are some pros and cons of each method? How did the authors of this study deal with protocol violations?
All randomized trials are subject to error by patients or providers not complying with the treatment allocation – this is called a “protocol violation.” There are three ways to manage this problem: 1. Exclude patients who did not comply, 2. Analyze the patients in the group they were originally assigned (ITT), or 3. Analyze the patients according to the actual treatment they received.
- If a therapy is beneficial, but poorly tolerated, excluding patients who did not comply with treatment may result in amplification of the apparent treatment effect. Also, dropping patients may reduce the sample size and thus result in a study with inadequate statistical power.
- ITT analysis maintains the sample size, but may under-estimate treatment effect by comparing heterogeneous groups. This type of analysis is often performed because it is conservative. Subgroup analysis can be very limited if the number of treatment group crossovers is large or subgroups are small.
- Per-treatment analysis runs the risk of introducing selection bias. In this case, providers may have opted to transfuse platelets in a sicker patient who was assigned not to receive platelets, simply because they felt the need to do something further. If sicker patients are more likely to have poor outcomes than less sick patients, then poor outcome might actually be correlated with severity of stroke rather than the transfusion.
These authors opt to perform ITT and per-treatment analysis (table 3), and allow the reader to draw their own conclusions.
HINTS Outperforms ABCD2 to Screen for Stroke in Acute Continuous Vertigo and Dizziness (NCBI Link)
Author’s Question: compare the accuracy in bedside screening for possible stroke in dizziness: a clinical decision rule (head impulse, nystagmus type, test of skew [HINTS]) and a risk stratification rule (age, blood pressure, clinical features, duration of symptoms, diabetes [ABCD2]).
Method: cross-sectional study of high-risk patients (more than one stroke risk factor) with acute vestibular syndrome (AVS; acute, persistent vertigo or dizziness with nystagmus, plus nausea or vomiting, head motion intolerance, and new gait unsteadiness) at a single academic center. All underwent neuro-otologic examination, neuroimaging (97.4% by magnetic resonance imaging [MRI]), and follow-up. ABCD2 risk scores (0–7 points), using the recommended cutoff of ≥4 for stroke, were compared to a three-component eye movement battery (HINTS). Sensitivity, specificity, and positive and negative likelihood ratios (LR+, LR–) were assessed for stroke and other central causes, and the results were stratified by age. False-negative initial neuroimaging was also assessed.
Results: total of 190 adult AVS pts were assessed (1999–2012).
Median age was 60.5 years (range = 18 to 92 years; interquartile range [IQR] = 52.0 to 70.0 years) 60.5% were men.
Final diagnoses were vestibular neuritis (34.7%), posterior fossa stroke (59.5% [105 infarctions, eight hemorrhages]), and other central causes (5.8%).
Median ABCD2 was 4.0 (range = 2 to 7; IQR = 3.0 to 4.0). ABCD2 ≥ 4 for stroke had sensitivity of 61.1%, specificity of 62.3%, LR+ of 1.62, and LR– of 0.62; sensitivity was lower for those younger than 60 years old (28.9%). HINTS stroke sensitivity was 96.5%, specificity was 84.4%, LR+ was 6.19, and LR– was 0.04 and did not vary by age.
For any central lesion, sensitivity was 96.8%, specificity was 98.5%, LR+ was 63.9, and LR– was 0.03 for HINTS, and sensitivity was 99.2%, specificity was 97.0%, LR+ was 32.7, and LR– was 0.01 for HINTS “plus” (any new hearing loss added to HINTS).
Initial MRIs were falsely negative in 15 of 105 (14.3%) infarctions; all but one was obtained before 48 hours after onset, and all were confirmed by delayed MRI.
Author’s Conclusion: HINTS substantially outperforms ABCD2 for stroke diagnosis in ED patients with AVS. It also outperforms MRI obtained within the first 2 days after symptom onset. While HINTS testing has traditionally been performed by specialists, methods for empowering emergency physicians (EPs) to leverage this approach for stroke screening in dizziness should be investigated.
Discussion: HINTS exam performed by neuro-ophthalmologist who were specially trained in HINTS exam with specialized equipment. This is less generalizable to an ED physician. Also HINTS was positive prior to stroke on MRI, however MRI used at a later time to diagnose stroke, questioning whether stroke was present at time of positive HINTS exam or developed later. ABCD2 score is based solely on history and HINTS exam uses physical exam, essentially comparing apples to oranges.
Bottom line: HINTS is a difficult exam to perform in the ED where as ABCD2 is much easier to generalize. It would be more appropriate to use the inclusion criteria for high risk pt for admission along with a multitude of differing factors. This study was performed by highly trained specialist and not ED physicians making it less applicable to our environment.
JC Question: Suppose you are unsure if your patient presenting with AVS has had a stroke, so you assign them a 50% probability of stroke. Using the positive and negative LRs in table 4, calculate the posttest probabilities for positive and negative results for ABCD2 and HINTS. Which test seems more clinically useful? Do these numbers still apply if stroke is much more common (prevalent) in Jacksonville than in Peoria?
In general, likelihood ratios give us a way of estimating the usefulness of a test in identifying a disease based on a pretest probability of the disease being present. An LR = 1 means pre- and post-test probabilities are identical. LR ≥ 10 or ≤ 0.1 are typically the most useful. LRs require a calculation, which can make them awkward to use clinically. However, they provide predictions of disease likelihood independent of disease prevalence.
You can use a nomogram or online calculator (like https://www.easycalculation.com/statistics/post-test-probability.php) to figure out the post-test probabilities of the disease. If ABCD2 is at least 4, the likelihood of stroke being present is 62%. If ABCD2 is less than 4, the probability of stroke falls to 39%. If the patient had a positive HINTS evaluation, the likelihood of stroke being present is 86%. If HINTS was negative, the post test probability of stroke is 4%. A positive HINTS increases the probability of stroke more than ABCD2 ≥ 4, and a negative HINTS more reliably rules out stroke than ABCD2 <4, so the HINTS test seems much more useful clinically.