Prognostic factors in pontine haemorrhage: A systematic review

Short abstract

Background

Pontine haemorrhage comprises approximately 10% of intracerebral haemorrhages. There is a common presumption that pontine haemorrhage is inherently associated with poor outcome.

Purpose

The aim of the review was to identify chief predictors of prognosis in (pontine haemorrhage) through systematic review of published literature.

Methods

A query of PubMed/MEDLINE was conducted in search of studies in English language since, 1980 focusing specifically on outcome in pontine haemorrhage. References for each publication were reviewed for additional studies not detected by the PubMed/MEDLINE probe. Surgical outcome studies were excluded from the review.

Findings

The query identified 7867 titles, after removal of duplicates and irrelevant studies, 20 titles were included in the review. In a total of 1437 pontine haemorrhage patients included in the 20 studies, the overall rate for early all-cause mortality was 48.1%. Level of consciousness on admission and haemorrhage size were the most consistent predictors of mortality in patients with pontine haemorrhage. Haemorrhage localisation within the pons was also a prognostic factor, but not consistently. Age and intraventricular extension were not found to be powerful prognostic predictors.

Discussion/Conclusion

Based on this review, level of consciousness on admission and haemorrhage size were the most influential prognostic factors in pontine haemorrhage, whereas age, haemorrhage localisation, and intraventricular haemorrhage did not consistently predict prognosis.

Introduction

Pontine haemorrhage (PH) comprises approximately 10% of intracerebral haemorrhages (ICH).¹ With an estimated mortality rate ranging widely from 30% to 90%, PH is the most pernicious form of ICH.^2–4 There is an inherent presumption of poor outcome associated with PH to the extent that some therapeutic clinical trials in ICH have excluded this category of patients.^5–7 Preconceived notion regarding PH prognosis, however, could usher into self-fulfilling prophecies. It is therefore important to identify reliable prognostic factors that would allow realistic expectation of outcome and prevent early care limitations.

Over the years, a host of predictors for outcome in PH have been proposed, yet the utility and reliability of many have not been substantiated. This is a review of some of the important factors that have been shown to influence outcomes in patients with PH.

Methods

Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) was utilised as guidance for this systematic review.⁸ A query of PubMed/MEDLINE was conducted between 3 September 2017 and 8 September 2017, in which medical subject headings ‘pons + haemorrhage,’ ‘pontine haemorrhage’ and ‘brainstem haemorrhage’ and ‘infratentorial haemorrhage’ were used. Studies in English that focused on outcomes in PH from 1980 onward were isolated and gathered. References for each publication were reviewed for additional studies not detected by the PubMed/MEDLINE probe. Surgical outcome studies were excluded from the review. Positive predictive values (PPV), negative predictive values (NPV) and positive likelihood ratios (PLR) were calculated using the standard formulas.

The query identified 7867 citations. A total of 2462 titles were found to be relevant, from which four non-English language studies were excluded. Thirteen additional studies were identified by reviewing the reference list of each publication. Twenty-four full-text articles were reviewed. Two post-surgical PH studies were excluded. Two studies included the same cohort of patients, one of which (the one with the least number of patients) was excluded. Twenty-one studies were used for this review. Cohort size per study ranged from 10 to 281 patients.^9,10 The oldest cohort was from 1983 and the most recent was from 2017; the majority were retrospective. The PRISMA flowchart of this query is displayed in Figure 1.⁸ Of the 21 outcome studies in PH, one was excluded from the overall analysis because of high number of PH caused by cavernous and arteriovenous malformations.¹¹ The results of this study were, however, reported in the review.

Figure 1.

The preferred reporting items for systematic reviews and meta-analyses flowchart.⁸ *The results of one study was reported in the review but not used in the analyses.¹¹

Results

Mortality and functional outcome

Based on 20 studies consisting of 1437 patients, the overall rate of early (within 30 days of PH or shortly thereafter) death of all causes was 48.1% (Table 1).^{1,2,4,9,10,12–26} This figure reflects results of studies spanning four decades. The individual mortality rates reported by these studies ranged from 33.7% to 77.5%.^14,26 Mortality rates for the oldest and the most recent studies were 71.7% and 39.8%, respectively.^12,26 Twelve studies reported on outcomes in survivors, totalling 951 patients.^{1,2,10,12–15,17,19–21,26} The rate of good outcome at various points of follow-up was 26.3% (Table 1). The point at which functional outcome was assessed after PH varied widely among the studies, ranging from 30 days to eight years. There was also significant subjectivity and heterogeneity among these studies with regard to the definition of good outcome, which ranged from ability to perform activities of daily living to quantification of disability using Glasgow Outcome or modified Rankin scales (mRS). Two studies used mRS ≤3 and one used mRS ≤2 to indicate good outcome (Table 1).^10,17,26 The combined 90-day rate of mRS ≤3 was 39.8% and the rate of mRS ≤2 in a follow-up period of two to eight years was 15.4%.

Table 1.

Studies included in assessment of early mortality in patients with pontine haemorrhage.

	Author	Year	Study type	N	N died	Duration to death	N good outcome	Average follow-up
1	Nakijama12	1983	Retrospective	60	43	14 days	11 (subjective)	NR
2	Kushner and Bressman9	1985	Retrospective	10	6	In-hospital	NR	NR
3	Masiyama et al.13	1985	Retrospective	26	12	23 days	5 (subjective)	6–42 months
4	Weisberg14	1986	Retrospective	40	31	In-hospital	4 (subjective)	NR
5	Chung and Park4	1992	NR	61	37	7 days	NR	NR
6	Wijdicks and St Louis15	1997	Retrospective	38	21	‘Within days of admission’	8 (subjective)	3–12 months
7	Fong16	1999	Retrospective	39	23	10 days	NR	NR
8	Murata et al.2	1999	Retrospective	80	38	‘Within few days’	28 (GOS ≥4)	NR
9	Dziewas et al.17	2003	Retrospective	39	27	Average of 16 days	6 (mRS, 0–2)	2–8 years
10	Wessels et al.1	2004	Retrospective	29	9	5 ± 3 days	12 (GOS ≥4)	NR
11	Balci et al.18	2005	Retrospective	32	18	Average 8 days	NR	NR
12	Jung et al.19	2007	Retrospective	35	13	In-hospital	12 (subjective)	Average 13.9 months
13	Shin et al.20	2007	Retrospective	35	15	In-hospital	13 (GOS ≥3)	3 months
14	Jang et al.10	2011	Retrospective	281	110	30 days	27 (mRS, 0–3)	90 days
15	Matsukawa et al.22	2014	Retrospective	118	66	Median 51 days	NR	NR
16	Nishizaki et al.21	2012	Retrospective	19	9	In-hospital	0 (GOS ≥4)	NR
17	Ye et al.24	2015	Prospectivea	76	33	30 days	NR	NR
18	Meguro et al.23	2015	Retrospective	101	59	30 days	NR	NR
19	Morotti et al.25	2016	Retrospective	49	30	30 days	NR	NR
20	Huang et al.26	2017	Retrospective	171	68	30 days	74 (mRS, 0–3)	90 days
			Prospectiveb	98	33	30 days	50 (mRS, 0–3)	90 days
	Total	−	−	1437	691	−	250	−

NR: not reported; GOS: Glasgow Outcome Scale; mRS: modified Rankin Scale.

^aBased on post hoc analysis of prospectively collected data.

^bExternal validation cohort for the study by Huang et al.

Clinical factors

Aetiology

A study of 44 consecutive patients with PH indicated that PH due to cavernous malformations had a more benign course and was associated with more favourable long-term outcome than primary PH due to hypertension.¹¹ In addition, patients with primary PH tended to have more severe clinical presentation, based on Glasgow Coma Scale (GCS) scores.¹¹ This study may have been biased by an unusually high number of PH cases with cavernous malformations (45.5%). Furthermore, it is unclear how the authors confidently diagnosed cavernous malformations using magnetic resonance imaging.

Differentiating between primary PH and that due to a cavernous malformation is not always simple in the acute phase. Some features that may assist in delineation are absence of history of hypertension, presence of cavernous malformations elsewhere in the brain on standard magnetic resonance imaging, especially if there are ≥5 lesions present suggesting familial cavernous angiomatosis; prior haemorrhage in the same location, or presence of developmental venous anomalies, which are found in 30% of individuals with cavernous malformations.^27,28 Gradient echo magnetic resonance imaging may also reveal microhaemorrhages in periventricular areas suggesting prior hypertensive ICH.²⁹

Level of consciousness

Coma on admission is arguably the most decisive prognosticating factor in patients with ICH.³⁰ The 20 outcome studies in PH used either ‘coma on admission,’ or various GCS score cutoffs – ranging from <4 to ≤9 – for prediction of 30-day mortality.^1,22–24,26 Irrespective of an arbitrary GCS score threshold, the plurality of studies looking at variables influencing outcomes in PH have consistently identified depressed level of consciousness as the single, most reliable predictor of death and disability.^{1,2,10,12,13,15–20,22–24} Eight studies, wherein multivariate regression analysis was conducted using either admission coma or GCS score as variables (a total of 968 patients) identified these factors as independent predictors of early mortality (Table 2).^{10,17,19,22–26}

Table 2.

Studies that conducted multivariate logistic regression to determine independent predictors of early death.

	Age	Admission GCS	Admission coma	Haemorrhage size	Haemorrhage location	Parenchymal extension	Intraventricular Hemorrhage	Hydrocephalus	Hyperthermia	Hyperglycaemia	Blood pressure	Pupillary reflex	Respiration
Dziewas et al.17	−	−	•	•	•	−	NS	−	−	−	NS	NS	−
Jung et al.19	NS	•	−	•	NS	NS	NS	NS	−	−	−	−	−
Jang et al.10	−	−	•	NS	NS	−	NS	•	−	−	•	•	•
Matsukawa et al.22	−	•	•	•	−	•	NS	−	NS	−	NS	−	−
Ye et al.24	NS	−	•	•	•	−	−	−	−	−	NS	−	−
Meguro et al.23	NS	•	•	NS	NS	NS	NS	−	NS	•	NS	•	−
Morotti et al.25	•	•	−	•	−	−	NS	−	−	−	−	−	−
Huang et al.26	−	•	•	•	−	NS	NS	NS	−	NS	NS		NS

Black dot indicates independent predictor of mortality (significant); dash indicates that the variable was not used in the multivariate analysis; NS, variable included in multivariate analysis, but the result was not significant.

Combing the data from studies of PH patients that stratified outcomes strictly based on “coma on admission” and not according to absolute GCS thresholds yielded 585 patients.^{1,2,10,12,13,15,17,18} The PPV and NPV, and PLR of coma on admission for early death were 77.2% and 82.9%, and 3.98 (95% confidence interval (CI): 3.18–5.00), respectively.

Grading scales

The ICH score is a validated prognostication model in ICH, and its ability to predict 30-day mortality also applies to PH.^31–33 However, ICH score may not be fully applicable in PH due to the small size of the pons vis-à-vis its severe clinical manifestations in case of haemorrhage, even in small volumes.²⁶ Subsequently, other outcome prediction scales have been proposed specifically for PH. Based on a recent study of 171 primary PH patients, a grading scale has been developed and validated for predicting 30-day mortality and 90-day functional outcome.²⁶ This scale used two independent factors: PH volume and GCS score (Table 1).²⁶ Point assignments according to this scale were as follows: GCS score 3 to 4 (2 points), 5 to 7 (1 point) and 8 to 15 (0 point); PH volume >10 mL (2 points), 5 to 10 mL (1 point) and <5 mL (0 point). Estimated 30-day mortality rates for patients with scores of 0, 1, 2, 3 and 4 were 2.7%, 31.6%, 42.7%, 81.8% and 100%, respectively (Table 3).²⁶ The area under the receiver operating characteristic curve showed that that this scale was more discriminative than the ICH Score in predicting 30-day mortality and 90-day good outcome.²⁶ For prediction of death in 30 days, the sensitivity of this grading system was 90.9% and the specificity was 63.1%.²⁶ For 90-day good outcome, the sensitivity and specificity were 87.8% and 77.6%, respectively.²⁶

Table 3.

The primary pontine haemorrhage (PPH) score developed by Huang et al.^26,a

Feature	Point
GCS score
3–4	2
5–7	1
8–15	0
Haemorrhage volume
>10 mL	2
5–10 mL	1
<5 mL	0
Score	mortality
0	2.7%
1	31.6%
2	42.7%
3	81.8%
4	100%

GCS: Glasgow Coma Scale.

^aUnit for haemorrhage volume is mL. Mortality is within 30 days.

An earlier scale incorporated three factors for predicting 30-day mortality: admission GCS score ≤6, absence of pupillary light reflex, and serum glucose ≥180 – each assigned 1 point.²³ The number of patients in that study was 101.²³ Thirty-day mortality rates were 7.7%, 33.3%, 78.9% and 100% for patients with 0, 1, 2 and 3 points, respectively.²³ With regard to this scale, PH volume was dichotomised into <20 and ≥20 mL, which may be too high of a volume threshold for the finite size of the pons. This particular scoring system also lacked external validation.

A study of 75 primary PH patients compared the performances of the Acute Physiology and Chronic Health Evaluation (APACHE) II, the Simplified Acute Physiology Score (SAPS) II, and the ICH score in predicting mortality.³⁴ Based on measurements using the Hosmer-Lemeshow goodness-of-fit test for calibration and the area under the receiver operating characteristic curve for discrimination, SAPS II had the highest calibration and APACHE II, the highest discrimination.³⁴ The results of this comparative analysis, however, must be viewed cautiously since SAPS II and APACHE II have been designed to predict in-hospital death, while the ICH Score is used to estimate mortality within 30 days.

Other factors

Despite age being a chief prognostic factor in ICH, only one study identified it as an independent predictor of 30-day mortality in PH.^25,31 Several other physiological factors have been reported in the literature as predictors of poor outcome in PH, though they have not been examined or replicated in successive investigations. History of hypertension, core temperature ≥39°C, systolic blood pressure <100 mmHg, absent motor response, absent corneal or oculocephalic reflexes, abnormal respiration, and need for mechanical ventilation have all been reported in various studies to independently correlate with death or poor outcome.^{10,15,20,22,23} Sex was not a prognostic factor in any of the studies.

Radiological predictors

Haemorrhage size

Haemorrhage size is an unswerving outcome-determining factor in ICH, and PH is no exception to this rule. Eight studies totalling 870 patients with PH demonstrated that haemorrhage size – ascertained by either volume or trans-axial diameter – is an independent predictor of mortality and outcome (Table 2).^{10,17,19,22–26} In these studies, size threshold values above which predicted death or unfavourable outcome ranged between 4 and 5 mL for volume, or 20 and 27 mm for trans-axial diameter.^{10,17,22,24,26} As previously mentioned, the pons itself constitutes a small dimensionality and therefore, the volume factor on the ICH score inadequately applies to PH. A haemorrhage volume of >30 mL in the pons is materially impossible and would certainly have catastrophic consequences. Furthermore, the ABC/2 formula used to measure ICH volume on computed tomography (CT) scan may overestimate the size of small haemorrhages in the pons.^35,36 These concerns may lead to the belief that trans-axial diameter is perhaps a more suitable technique to quantify PH size.² However, in terms of prognostication, the longitudinal dimension of a brainstem lesion cannot be undervalued. Estimation of volume using the ABC/2 equation takes both dimensions into account and is subsequently the most justifiable method to measure haemorrhage size, despite its presumed modest accuracy in PH. Three-dimensional voxel-based magnetic resonance or pixel-based CT images volumetry is a promising method that may be used in precise measurement of PH volume.^37,38

Haemorrhage localisation

Early small case series of patients with PH reported relatively higher survival rates for lateral haemorrhage localisation.^8,14,39,40 In 1992, Chung and Park classified axial CT features of primary PH into four types: basal-tegmental, bilateral tegmental, massive and small unilateral tegmental.⁴ The massive type was defined as haematomas occupying both the basis pontis and the tegmentum bilaterally, the bilateral tegmental type as those occupying the bilateral tegmentum only, the basal-tegmental type as those occupying the junction between the basis pontis and the tegmentum bilaterally, and the small unilateral tegmental type was defined as haematomas located exclusively in the unilateral tegmentum (Figure 2).⁴ Case survival rate was 94.1% for the small unilateral tegmental type, 26.19% for the basal-tegmental type, 14.3% for the bilateral tegmental type, and 7.1% for the massive type.⁴ Thus, Chung and Park concluded that small haematomas located in the unilateral tegmentum that had the best prognosis.⁴

Figure 2.

Computed tomography classification of pontine haemorrhage localisations according to Chung and Park.⁴ From top left, clockwise: massive, basal tegmental, bilateral tegmental and unilateral tegmental.

In a study of 39 consecutive patients with PH, Dziewas et al. divided trans-axial PH locations into three categories: large paramedian, unilateral basotegmental and lateral tegmental.¹⁷ The large paramedian type of PH predicted a poor prognosis, whereas the lateral tegmental type was associated with a favourable outcome.¹⁷ The transverse haematoma diameter threshold value related to outcome was found to be 20 mm.¹⁷ Another study of 29 primary PH patients reported that dorsally-located small haemorrhages (<4 mL) had significantly better outcome than those with ventrally-situated lesions.¹ Survival rates for dorsal, ventral and massive PH were 75%, 25% and 0%, respectively.¹ It is important to note that axial location of PH is also dependent on its size. The localisations described in the above-mentioned studies relate to small haemorrhages. The largest outcomes study of primary PH by Jang et al. which involved 281 patients, reported 30-day mortality rates of 66.9%, 24.7% and 1.5% for massive, ventral and dorsal localisations, respectively.¹⁰ Functional recovery rates at 90 days were not different between ventral and dorsal haemorrhages (14.8% versus 14.9%).¹⁰ However, none of the stated figures were stratified according to PH size.

Haemorrhage extension and growth

Extension of ICH into the ventricles is associated with increased risk of death and adverse outcome.^31,32 In a total of 1009 PH patients from studies that reported on intraventricular extension (IVE), the prevalence was 39.5%.^{1,2,10,15–19,22,25,26} Among seven studies that included IVE as a prognostic variable in multivariate logistic analysis, none could establish it as an independent predictor of early death in PH.^{10,17,19,22,23,25,26} One study – the largest by Jang et al. – identified the absence of IVE as an independent predictor of 90-day functional recovery (mRS score, 0–3), with an odds ratio of 3.64 (95% CI: 2.18–7.08; p = 0.001).¹⁰ It did not, however, find an association between IVE and 30-day mortality in PH.¹⁰ It is unclear why IVE is a major prognostic factor in ICH and not in PH. One theory may be the frequency of extraventricular drainage use.¹⁰ IVE in PH is more often localised to the fourth ventricle and cerebral aqueduct and subsequently, carry a higher risk of hydrocephalus. Development of hydrocephalus will in turn increase the use of extraventricular drainage. In ICH patients with IVE, extraventricular drainage is associated with reduced mortality and improved short-term outcomes.⁴¹

Extrapontine extension into the midbrain with or without thalamic involvement was shown by only one study of 118 consecutive PH patients to be independently associated with death within a median follow-up period of 51 days (hazard ratio, 2.2; 95% CI: 1.1–4.4; p = 0.033).²² Three other studies that examined the effect of parenchymal extension on survival could not confirm this finding.^19,23,26

Haemorrhage expansion is an independent determinant of mortality and functional outcome after ICH.⁴² A 1 mL increase in haematoma growth is associated with a 5% higher risk of death or dependency.⁴³ Very few studies to date have specifically investigated the impact of PH expansion in on outcomes. The CT angiogram ‘spot sign’ is an independent predictor of haematoma expansion, in-hospital mortality and poor long-term outcome in supratentorial ICH.⁴⁴ A recent secondary analysis of Antihypertensive Treatment of Acute Cerebral Hemorrhage II (ATACH-II) randomised clinical trial, however, demonstrated that the predictive performance of the spot sign for ICH expansion was lower than in prior reports.⁴⁵ In a retrospective analysis of 49 consecutive patients with primary PH, the CT angiogram spot sign showed an accuracy of 57% for 30-day mortality, and did not independently predict in-hospital or 30-day mortality.²⁵ Moreover, no significant difference was observed in haemorrhage expansion rates between spot sign positive and negative patients.²⁵ Due to its size and the condensed functional machinery it contains, it is conceivable that expansion of the haemorrhage in the pons would have devastating impact on survival and outcomes.

Hydrocephalus

Among 12 studies with a total of 1063 patients, 30.3% developed hydrocephalus.^{1,2,10,15–22,25,26} Only one of the three studies that included hydrocephalus as a variable in multivariate analysis found it to be an independent predictor of mortality and functional outcome.^10,19,26 This study was by Jang et al., which estimated the odds ratios for 30-day mortality and 90-day functional recovery (absence of hydrocephalus) at 2.98 (95% CI: 1.84–5.02; p = 0.038) and 2.84 (95% CI: 1.79–5.32; p = 0.03), respectively.¹⁰

Discussion

Based on the available evidence, the overall all-cause mortality rate for PH is roughly 50%. Level of consciousness on admission and haemorrhage size are the strongest and most consistent predictors of mortality in patients with PH. Small PH with dorsal and lateral localisations may have better survival rates, compared to other localisations; certainly, more favourable prognosis than massive PH. Age, IVE and hydrocephalus were not consistently found to be strong prognostic features. No study to date has investigated the effect of haemorrhage growth specifically in PH.

Over the four decades in which these studies were conducted, the survival of PH appears to have improved. This may be partly due to advancement is the fields of neurology, neurosurgery and intensive care medicine. Improved public perception of ICH and stroke in general may have contributed to this progress. However, it is not clear if functional outcomes have improved in parallel to survival.

There were a few limitations to this review. The majority of the studies were retrospective and there was significant heterogeneity in how data were collected and interpreted. These factors precluded the ability to perform a meta-analysis. Studies spanned four decades and management of ICH has evolved over the years. Therefore, survival and outcomes were certainly affected by difference in management strategies between older and more recent studies. Studies that involved surgical evacuation and those older before, 1980 were excluded to ensure some level of consistency. Furthermore, there may have been regional influences on how PH patients were managed, especially in older series when standardised guidelines were not available. The variability in perception of ‘good outcome’ also limited the ability to determine the frequency of favourable functional outcome in patients with PH. Other factors that may have limited this review were heterogeneity in definitions and reporting of conditions such as hydrocephalus, as well as missing data in individual studies.

In conclusion, patients with PH do not seem to invariably have poor outcomes. Limiting care due to preconceived unfavourable prognosis in PH is not reasonable. Although in most randomised trials, tailored sample populations must be used to reasonably perform the study, omitting PH patients from clinical trials will prevent establishing inclusive treatment strategies for ICH. In order to better prognosticate PH and understand this insidious condition, large, prospective studies are certainly needed.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Informed consent

None required; this is a review article.

Ethical approval

None required; this is a review article.

Guarantor

RB.

Contributorship

RB is the sole author. RB researched literature, wrote the draft of the manuscript.