Revisiting the Repeat Head CT after Minor Head Trauma

Author: Shane Solger, MD

Edited by: Andrea Greene, MD, MPH

Just Another Minor Head Trauma?

When a healthy, middle-aged Joe with minor head trauma buys himself a CT scan in the ED, he can often be discharged from the ED after a short period of observation, barring any other concerning injuries or changes in the neurologic exam. More on imaging for minor head trauma in healthy patients is in a prior County EM post.

What about your 85-year-old Bubbe who trips over her shoes and hits the ground head first? She takes that medication that starts with an E or maybe it’s an A? The cardiologist has her taking aspirin too. She usually goes to the other hospital a few miles away, but the medication “should be in the computer”.

For an 85-year-old on blood-thinners with a ground-level fall and minor head trauma, ordering an initial non-contrast head CT is completely reasonable. Studies show significant intracranial injury (bleeds, skull fractures, etc.) rates up to 15% in patients of any age presenting to the ED with a fall after taking aspirin, clopidogrel, or warfarin [10]. Five percent of elderly patients who present to the ED after a fall will have an intracranial hemorrhage (ICH) [3].

A lower threshold to CT elderly patients is reasonable because their risk for ICH is relatively high, and their risk for clinically significant sequelae to radiation exposure is relatively low.

Second Time’s A Charm! Right?

If the initial CT is negative, and Bubbe feels great after some acetaminophen, is it safe to send her home? Do we need to repeat the head CT? How much longer should we observe her in the ED?

The American College of Surgeons’ imaging guidelines note a lack of quality prospective evidence and the presence of contradictory evidence. They suggest a liberal policy regarding repeat CT and that it may be worthwhile to obtain repeat imagining 4-8 hours after an initial negative CT. But, if a patient would be admitted for other reasons and otherwise closely monitored, a repeat CT may not be necessary [5].

The Eastern Association for the Surgery of Trauma’s 2012 guidelines suggest admission for those with a negative initial scan and a supratherapeutic INR but did not discuss performing routine repeat CT [4]. Institutional policies vary, with Vanderbilt’s policy dictating that all individuals taking a pre-injury DOAC receive an initial CT and repeat CT 12 hours after their blunt trauma [11]. The University of Texas Health Science Center at Houston’s guidelines allow observation for six hours in individuals taking a pre-injury DOAC or warfarin after negative initial CT, a stable neurological exam, an INR <3.5, and no other injuries that warranted an admission [1].

With a new meta-analysis, is there any clarity on the matter?

Analyzing the Analysis

Huang, G., Dunham, C., Chance, E. and Hileman, B., 2020. Detecting delayed intracranial hemorrhage with repeat head imaging in trauma patients on antithrombotics with no hemorrhage on the initial image: A retrospective chart review and meta-analysis. The American Journal of Surgery, 220(1), pp.55-61.

*Note: only the meta-analysis will be discussed

Clinical Question:

What is the incidence of delayed traumatic ICH in patients taking pre-injury antithrombotic agents after blunt trauma to the head ?

Outcome (Primary):

The incidence of delayed ICH on repeat CT in a patient taking any anticoagulant/antiplatelet (ACAP) agent.

Design:

The data synthesis was from studies of any design that included adults with blunt trauma, pre-injury use of ACAPs, documented absence of ICH on initial head CT, and a follow-up process to identify delayed ICH.

Included:

The search included papers in PubMed, CINAHL, and Cochrane reviews from 2005 to 2019.

Statistical Analysis:

-The authors computed a combined study proportion estimate for delayed ICH for all patients included in the review that were taking any ACAP.

-Subgroup analyses proportion estimates were computed for ASA, clopidogrel, VKAs, DOACs, single-agent therapy, and double agent therapy.

-When inter-study test for heterogeneity was <0.05, the random effects model estimates were used; in all other circumstances fixed effects model estimates were used.

Primary Results:

-24 publications were identified with 7,319 patients.

-The range of GCS was 12-15 with an average of 14.6 (eight studies did not report GCS).

-The reported ages were between 51-83 years with an average age of 77.6 (two studies did not report age).

-Ten studies included did not require a routine repeat CT.

Critical Findings:

Medication

All Patients

Routine Repeat CT

Non-Routine Repeat CT

p-value

*Primary Outcome*

Any Antithrombotic:

Studies

Total

Delayed ICH % (CI)

7,319

1.4 (0.9-1.9)

3,998

1.7 (1.1-2.4)

3,321

0.8 (0.3-1.6)

0.0008

Aspirin:

Studies

Total

Delayed ICH % (CI)

–

469

1.3 (0.2-3.5)

–

Clopidogrel:

Studies

Total

Delayed ICH % (CI)

624

1.5 (0.1-4.4)

127

2.4 (0.01-8.7)

497

0.9 (0.09-4.9)

0.1530

VKA:

Studies

Total

Delayed ICH % (CI)

3,129

1.3 (0.6-2.3)

897

1.8 (1.1-2.9)

2,232

0.9 (0.2-2.2)

0.0416

DOAC:

Studies

Total

Delayed ICH % (CI)

279

1.3 (0.2-3.4)

3.3 (0.1-17.2)

249

0.8 (0.1-2.9)

0.2901

Combination:

Studies

Total

Delayed ICH % (CI)

–

181

2.8 (1.0-6.4%)

–

Single Agent:

Studies

Total

Delayed ICH % (CI)

4,501

1.3 (0.8-2.1)

1,523

1.7 (0.9-2.7)

2,978

0.9 (0.3-1.9)

0.0274

Limitations/Critiques:

-The analyzed data was from a heterogeneous population.

-They combined results of both prospective, retrospective, multicenter, and single-center analyses, whose study designs are intrinsically different.

-They included a study that was dissimilar to the others in terms of lower limits of age (51) and GCS score (12), indicating some of their pooled data had a moderate TBI

-Not all studies included age and GCS, which may indicate a broader spread of injury severity and age spectrum.

-VKA use was over-represented (14/25 studies), accounting for approximately 42% of the “any antithrombotic” group.

-Only three studies reported slices or models of CT scanners.

-⅓ of the studies did not have routine repeat CTs, introducing potential selection bias.

Problems with the STATS:

The authors inappropriately used proportion estimates to analyze both their primary outcome group and their subgroup analyses and this impacts the reliability of confidence intervals.

The total Sample (n) x Population Proportion (p) x (1-p) ≥ 10

If the data meets that specific parameter, then you can assume a normal distribution in the population, the confidence interval can be applied, and you can reliably use that Z-table from the back of your old stats book to do some math. This was appropriate only for the groups, any antithrombotic, VKA, and combination therapy. This requirement was not met by the ASA, Clopidogrel, DOAC, or single agent, so we cannot draw any meaningful conclusions about the confidence intervals that they put forth for more than half of their subgroups.

Another glaring issue is that the authors used fixed effects model estimates, which is useful when looking to compare studies that occurred under similar circumstances, AS WELL AS random effects model estimates, which would require more data to overcome intrinsic differences in study design. But they did not provide information on what was compared with which model.

In essence, the authors reported, “we used this fixed model when we thought the papers were homogenous, except when the studies were too different, which is when we used the other model to control for unforeseen variables, and we aren’t going to let you know which papers we compared with which model or how many of the papers needed the random effects model.”

Given that the populations in the different studies occurred in different countries, in different age groups, in different hospitals, in different time periods, with potential different practice patterns, different medico-legal climates, and varying institutional policies, the authors did not provide enough information to justify the use of these particular models, especially as most of them were retrospective analyses.

TLDR: Much of what occurred in the individual papers cannot be controlled for or assumed to be constant and there may have been too many independent variables to reliably say the papers could be compared with the data they had provided.

Author’s Conclusions:

For patients receiving pre-injury ACAP with blunt trauma to the head without ICH on initial CT, repeat CT should be discretionary based on neurological assessments. Given that twice as many patients receiving routine repeat CTs were found to have delayed ICH than those who obtained non-routine repeat CTs, routine imaging may detect more patients with ICH, but the ICH may not be clinically important (require intervention).

Take Aways:

Unfortunately, Huang et al.’s meta-analysis did not answer our clinical question very well. The burden of the responsibility to explain the heterogeneity of the studies lies with the authors, and they fell short of this task.

– They inappropriately compared study proportion estimates and developed confidence intervals for populations without a normal distribution.

– They included information from prospective studies as well as the retrospective studies that relied on chart review.

– They took a leap to assume that all non-routine CTs were performed due to clinical gestalt and neurological deterioration, even as ten of the studies (accounting for nearly half of the total patient population) did not include routine repeat CT in their study design

Their mish-mosh of studies were simply too dissimilar and too heavily weighted towards VKA to reliably make conclusions about ACAP therapy as a whole.

All is not lost:

While the recommendation to have all institutions completely do away with routine repeat CT in these patients isn’t supported by Huang’s paper, this is very likely a case of “The absence of (quality) evidence, does not suggest the evidence of absence.” Looking back at the American College of Surgeons’ recommendations, there are going to be cases of elderly patients on ACAP therapy with a mild TBI, a stable neurological exam, and a negative initial head CT that may still be discharged home after a period of observation and sound follow-up. In some cases, this may be a prudent plan of action.

Given that a clinically significant bleed may occur anytime during the six weeks after the initial injury, admission for observation may provide false reassurance and expose a fragile patient to the well-known dangers of the hospital. The risk of a delayed bleed is low (Huang et al report 1.4%), and the risk of a clinically significant delayed bleed requiring neurosurgery is lower; this must be juxtaposed against the serious adverse events that can occur in 14-21% of hospitalizations, among them 0.6-1.4% causing death [7].

James Li calculated the cost of two CTs and an admission to be $4,540 in 2012, which is roughly $5,150 when adjusted for inflation in 2020 [8]. Mann et al found one delayed ICH out of their 218 patients; it cost $1,122,700 to detect one bleed that did not require surgery. Huang et al found four delayed bleeds out of 237, indicating it cost $1,220,550 to detect four delayed bleeds that wouldn’t need surgery. Even with a shorter observation period, that is $571,650 to find one delayed ICH which likely will not need treatment [9].

Taking into account the cost, the risk of adverse events during hospitalization, the fact that a delayed bleed may occur up to five weeks after injury, and the low incidence of clinically important, delayed ICH after an initial negative CT, Bubbe may benefit most from a short observation period in the ED after the initial CT and then discharge home… with certain caveats.

Do you work in a center where she can reliably obtain neuro checks or is the nursing staff stretched thin? The routine CT may be necessary as a failsafe for a poorly funded institution. Is Bubbe living alone? Does she live in a facility where she can be watched? Is she homeless? Is she AAOx1 at baseline or is she a good historian that is in tune with her health?

More prospective studies are necessary before we can make blanketed recommendations for all elderly patients taking ACAP therapy who have a mild TBI and initial negative CT. Until that data is collected, one must allow for nuance in management and weigh the risks of admission against the risks of our patients’ deterioration going unnoticed.