Classical Regression and Predictive Modeling.

BACKGROUND: With the advent of personalized and stratified medicine, there has been much discussion about predictive modeling and the role of classical regression in modern medical research. We describe and distinguish the goals in these 2 frameworks for analysis. METHODS: The assumptions underlying and utility of classical regression are reviewed for continuous and binary outcomes. The tenets of predictive modeling are then discussed and contrasted. Principles are illustrated by simulation and through application of methods to a neurosurgical study. RESULTS: Classical regression can be used for insights into causal mechanisms if careful thought is given to the role of variables of interest and potential confounders. In predictive modeling, interest lies more in accuracy of predictions and so alternative metrics are used to judge adequacy of models and methods; methods which average predictions over several contending models can improve predictive performance but these do not admit a single risk score. CONCLUSIONS: Both classical regression and predictive modeling have important roles in modern medical research. Understanding the distinction between the 2 frameworks for analysis is important to place them in their appropriate context and interpreting findings from published studies appropriately.

Commentary on Post, et al. Ultra-early tranexamic acid after subarachnoid hemorrhage: A randomized controlled trial. Lancet 2021.

BACKGROUND: Tranexamic acid (TA) administration in aneurysmal subarachnoid hemorrhage (SAH) within the first 24 hours may reduce the incidence of early aneurysmal rebleeding. However, this is also the potential for an increased risk of delayed cerebral ischemia if TA is administered for more than 72 hours following the initial aneurysmal rupture. METHODS: In the ultra-early tranexamic acid after subarachnoid hemorrhage randomized controlled trial by Post et al., patients were randomized to receive TA within the first 24 hours, or until start of aneurysm treatment. These results were compared to a matched control group. RESULTS: Ultra-early administration (≤24 h) of TA reduced the incidence of rebleeding, and did not alter the incidence of delayed cerebral ischemia and/or extracranial thrombosis. Further, no significant differences were noted between the TA group and control arm in the incidence of good (modified Rankin scores 0-3) clinical outcomes at 6 months. CONCLUSION: Ultra-early administration of TA (≤24 h) resulted in a lower rate of recurrent hemorrhage, without increasing the incidence of delayed cerebral ischemia in SAH patients.

Review of Toyoda K, et al. Dual antiplatelet therapy using cilostazol for secondary prevention in patients with high-risk ischemic stroke in Japan. Lancet Neurol. 2019 Jun;18(6):539-548.

BACKGROUND: Prior meta-analyses showed that treatment with cilostazol, with or without aspirin, significantly reduced the incidence of recurrent ischemic stroke, occurrence of hemorrhagic stroke, and frequency of other serious vascular adverse events. METHODS: This review highlights the value of the randomized controlled trial (RCT) by Toyoda et al. entitled, "Dual antiplatelet therapy using cilostazol for secondary prevention in patients with high-risk ischemic stroke in Japan: a multicenter, open-label, randomized controlled trial." Here, dual therapy consisting of cilostazol and another antiplatelet agent was used to prevent secondary ischemic stroke in high-risk Japanese patients. RESULTS: Patients on dual therapy consisting of cilostazol/aspirin or cilostazol/clopidogrel had significantly lower frequencies of recurrent stroke. However, there were significant differences in the incidence of attendant hemorrhagic complications utilizing mono or dual therapy. CONCLUSION: This RCT demonstrated the safety of dual therapy, consisting of cilostazol/aspirin or cilostazol/ clopidogrel, in preventing secondary ischemic stroke in a high-risk Japanese population. Further studies are required to generalize these findings to other patient populations worldwide.

Between-center and between-country differences in outcome after aneurysmal subarachnoid hemorrhage in the Subarachnoid Hemorrhage International Trialists (SAHIT) repository.

OBJECTIVE: Differences in clinical outcomes between centers and countries may reflect variation in patient characteristics, diagnostic and therapeutic policies, or quality of care. The purpose of this study was to investigate the presence and magnitude of between-center and between-country differences in outcome after aneurysmal subarachnoid hemorrhage (aSAH). METHODS: The authors analyzed data from 5972 aSAH patients enrolled in randomized clinical trials of 3 different treatments from the Subarachnoid Hemorrhage International Trialists (SAHIT) repository, including data from 179 centers and 20 countries. They used random effects logistic regression adjusted for patient characteristics and timing of aneurysm treatment to estimate between-center and between-country differences in unfavorable outcome, defined as a Glasgow Outcome Scale score of 1-3 (severe disability, vegetative state, or death) or modified Rankin Scale score of 4-6 (moderately severe disability, severe disability, or death) at 3 months. Between-center and between-country differences were quantified with the median odds ratio (MOR), which can be interpreted as the ratio of odds of unfavorable outcome between a typical high-risk and a typical low-risk center or country. RESULTS: The proportion of patients with unfavorable outcome was 27% (n = 1599). The authors found substantial between-center differences (MOR 1.26, 95% CI 1.16-1.52), which could not be explained by patient characteristics and timing of aneurysm treatment (adjusted MOR 1.21, 95% CI 1.11-1.44). They observed no between-country differences (adjusted MOR 1.13, 95% CI 1.00-1.40). CONCLUSIONS: Clinical outcomes after aSAH differ between centers. These differences could not be explained by patient characteristics or timing of aneurysm treatment. Further research is needed to confirm the presence of differences in outcome after aSAH between hospitals in more recent data and to investigate potential causes.

Antithrombotic agents and traumatic brain injury in the elderly population: hemorrhage patterns and outcomes.

OBJECTIVE: Among the elderly, use of antithrombotics (ATs), antiplatelets (APs; aspirin, clopidogrel), and/or anticoagulants (ACs; warfarin, direct oral ACs [DOACs; dabigatran, rivaroxaban, apixaban]) to prevent thromboembolic events must be carefully weighed against the risk of intracranial hemorrhage (ICH) with trauma. The goal of this study was to assess the risk of sustaining a traumatic brain injury (TBI), ICH, and poorer outcomes in relation to AT use among all patients 65 years or older presenting to a single institution with head trauma. METHODS: Data were collected from all head trauma patients 65 years or older presenting to the authors' supraregional tertiary trauma center over a 24-month period and included age, sex, injury mechanism, medical history, international normalized ratio, Glasgow Coma Scale (GCS) score, ICH presence and type, hospital admission, reversal therapy, surgery, discharge destination, Extended Glasgow Outcome Scale (GOSE) score at discharge, and mortality. RESULTS: A total of 1365 head trauma patients 65 years or older were included; 724 were on AT therapy (413 on APs, 151 on ACs, 59 on DOACs, 48 on 2 APs, 38 on AP+AC, and 15 on AP+DOAC) and 641 were not. Among all head trauma patients, the risk of sustaining a TBI was associated with AP use after adjusting for covariates. Of the 731 TBI patients, those using ATs had higher rates of ICH (p <0.0001), functional dependency at discharge (GOSE score ≤ 4; p < 0.0001), and mortality (p < 0.0001). Elevated rates of ICH progression on follow-up CT scanning were observed in patients in the warfarin monotherapy (OR 5.30, p < 0.0001) and warfarin + AP (OR 6.15, p = 0.0011). Risk of mortality was not associated with single antiplatelet use but was notably high with 2 APs (OR 4.66, p = 0.0056), warfarin (OR 5.18, p = 0.0003), and DOAC use (OR 5.09, p = 0.0149). CONCLUSIONS: Elderly trauma patients on ATs, especially combination therapy, are at elevated risk of ICH and poor outcomes compared with those not on AT therapy. While both AP and warfarin use alone and in combination were associated with significantly elevated odds of sustaining an ICH among TBI patients, only warfarin use was a predictor of hemorrhage progression on follow-up scans. The use of a single AP was not associated with mortality; however, the combination of both aspirin and clopidogrel was. Warfarin and DOAC users had comparable mortality rates; however, DOAC users had lower rates of ICH progression, and fewer survivors were functionally dependent at discharge than were warfarin users. DOACs are an overall safer alternative to warfarin for patients at high risk of falls.

Determining blood flow direction from short neurovascular surgical microscope videos.

Neurovascular surgery aims to repair diseased or damaged blood vessels in the brain or spine. There are numerous procedures that fall under this category, and in all of them, the direction of blood flow through these vessels is crucial information. Current methods to determine this information intraoperatively include static pre-operative images combined with augmented reality, Doppler ultrasound, and injectable fluorescent dyes. Each of these systems has inherent limitations. This study includes the proposal and preliminary validation of a technique to identify the direction of blood flow through vessels using only video segments of a few seconds acquired from routinely used surgical microscopes. The video is enhanced to reveal subtle colour fluctuations related to blood pulsation, and these rhythmic signals are further analysed in Fourier space to reveal the direction of blood flow. The proposed method was validated using a novel physical phantom and retrospective analysis of surgical videos and demonstrated high accuracy in identifying the direction of blood flow.

Augmented reality guidance in cerebrovascular surgery using microscopic video enhancement.

Cerebrovascular surgery treats vessel abnormalities in the brain and spinal cord, including arteriovenous malformations (AVMs) and aneurysms. These procedures often involve clipping the vessels feeding blood to these abnormalities, making accurate classification of blood vessel types (feeding versus draining) important during surgery. Previous work to guide the intraoperative identification of the vessels included augmented reality (AR) using pre-operative images, injected dyes, and Doppler ultrasound, but each with their drawbacks. The authors propose and demonstrate a novel technique to help differentiate vessels by enhancing short videos of a few seconds from the surgical microscope using motion magnification and spectral analysis, and constructing AR views that fuse the analysis results as intuitive colourmaps and the surgical microscopic view. They demonstrated the proposed technique retrospectively with two real cerebrovascular surgical cases: one AVM and one aneurysm. The results showed that the proposed technique can help characterise different vessel types (feeding and draining the abnormality), which agree with those identified by the operating surgeon.

Intra-operative Video Characterization of Carotid Artery Pulsation Patterns in Case Series with Post-endarterectomy Hypertension and Hyperperfusion Syndrome.

Cerebral hyperperfusion syndrome (CHS) is a complication that can occur after carotid endarterectomy (CEA), the treatment of choice to decrease the subsequent risk of fatal or disabling stroke for patients with symptomatic severe stenosis of the carotid artery. Because of its rarity and complexity, the mechanism of the condition is still unclear, making its prevention via prediction and monitoring challenging. This is especially true during surgery, when multiple factors can induce physiological changes, including blood pressure and baroreceptor functions, which are crucial factors for post-CEA hypertension and CHS. Thus, with intra-operative videos taken by surgical microscopes, we employed a new video processing technique to magnify ordinarily invisible carotid artery pulsation patterns as rhythmic color fluctuations. We applied the technique for three CEA cases, two of which developed CHS with post-CEA hypertension. For those with CHS, abnormal pulsation patterns were detected at the site of the baroreceptors. The results suggested that intra-operative baroreceptor dysfunction can potentially be linked with post-operative hypertension, as well as the occurrence of CHS. Guided by the preliminary discovery, further investigation may help establish the introduced technique as a simple and contactless technique to help predict post-CEA hypertension and CHS in order to facilitate the management and understanding of the condition and improve the care of CEA.

Clinical outcome prediction in aneurysmal subarachnoid hemorrhage - Alterations in brain-body interface.

BACKGROUND: Brain-body associations are essential in influencing outcome in patients with ruptured brain aneurysms. Thus far, there is scarce literature on such important relationships. METHODS: The multicenter Tirilazad database (3551 patients) was used to create this clinical outcome prediction model in order to elucidate significant brain-body associations. Traditional binary logistic regression models were used. RESULTS: Binary logistic regression main effects model included four statistically significant single prognostic variables, namely, neurological grade, age, stroke, and time to surgery. Logistic regression models demonstrated the significance of hypertension and liver disease in development of brain swelling, as well as the negative consequences of seizures in patients with a history of myocardial infarction and post-admission fever worsening neurological outcome. CONCLUSIONS: Using the aforementioned results generated from binary logistic regression models, we can identify potential patients who are in the high risk group of neurological deterioration. Specific therapies can be tailored to prevent these detriments, including treatment of hypertension, seizures, early detection and treatment of myocardial infarction, and prevention of hepatic encephalopathy.

Aneurysmal subarachnoid hemorrhage prognostic decision-making algorithm using classification and regression tree analysis.

BACKGROUND: Classification and regression tree analysis involves the creation of a decision tree by recursive partitioning of a dataset into more homogeneous subgroups. Thus far, there is scarce literature on using this technique to create clinical prediction tools for aneurysmal subarachnoid hemorrhage (SAH). METHODS: The classification and regression tree analysis technique was applied to the multicenter Tirilazad database (3551 patients) in order to create the decision-making algorithm. In order to elucidate prognostic subgroups in aneurysmal SAH, neurologic, systemic, and demographic factors were taken into account. The dependent variable used for analysis was the dichotomized Glasgow Outcome Score at 3 months. RESULTS: Classification and regression tree analysis revealed seven prognostic subgroups. Neurological grade, occurrence of post-admission stroke, occurrence of post-admission fever, and age represented the explanatory nodes of this decision tree. Split sample validation revealed classification accuracy of 79% for the training dataset and 77% for the testing dataset. In addition, the occurrence of fever at 1-week post-aneurysmal SAH is associated with increased odds of post-admission stroke (odds ratio: 1.83, 95% confidence interval: 1.56-2.45, P < 0.01). CONCLUSIONS: A clinically useful classification tree was generated, which serves as a prediction tool to guide bedside prognostication and clinical treatment decision making. This prognostic decision-making algorithm also shed light on the complex interactions between a number of risk factors in determining outcome after aneurysmal SAH.

Systematic review of clinical prediction tools and prognostic factors in aneurysmal subarachnoid hemorrhage.

BACKGROUND: Clinical prediction tools assist in clinical outcome prediction. They quantify the relative contributions of certain variables and condense information that identifies important indicators or predictors to a targeted condition. This systematic review synthesizes and critically appraises the methodologic quality of studies that derive both clinical predictors and clinical predictor tools used to determine outcome prognosis in patients suffering from aneurysmal subarachnoid hemorrhage (SAH). METHODS: This systematic review included prospective and retrospective cohort studies, and randomized controlled trials (RCTs) investigating prognostic factors and clinical prediction tools associated with determining the neurologic outcome in adult patients with aneurysmal SAH. RESULTS: Twenty-two studies were included in this systemic review. Independent, confounding, and outcome variables were studied. Methodologic quality of individual studies was also analyzed. Included were 3 studies analyzing databases from RCTs, 8 prospective cohort studies, and 11 retrospective cohort studies. The most frequently retained significant clinical prognostic factors for long-term neurologic outcome prediction include age, neurological grade, blood clot thickness, and aneurysm size. CONCLUSIONS: Systematic reviews for clinical prognostic factors and clinical prediction tools in aneurysmal SAH face a number of methodological challenges. These include within and between study patient heterogeneity, regional variations in treatment protocols, patient referral biases, and differences in treatment, and prognosis viewpoints across different cultures.

Pathophysiologic mechanisms of brain-body associations in ruptured brain aneurysms: A systematic review.

BACKGROUND: Patients with ruptured brain aneurysms and aneurysmal subarachnoid hemorrhage suffer neurological damage from primary injury of the aneurysm rupture itself, as well as a number of secondary injurious processes that can further worsen the affected individual's neurological state. In addition, other body systems can be affected in a number of brain-body associations. METHODS: This systematic review synthesizes prospective and retrospective cohort studies that investigate brain-body associations in patients with ruptured brain aneurysms. The methodologic quality of these studies will be appraised. RESULTS: Six cohort studies were included in this systemic review. The methodologic quality of each study was assessed. They had representative patient populations, clear selection criteria and clear descriptions of study designs. Reproducible study protocols with ethics board approval were present. Clinical results were described in sufficient detail and were applicable to aneurysmal subarachnoid hemorrhage patients in clinical practice. There were few withdrawals from the study. Limitations included small sample sizes and between-study differences in diagnostic tests and clinical outcome endpoints. Several pathophysiologic mechanisms of brain-body associations in ruptured brain aneurysms were clarified through this systematic review. Sympathetic activation of the cardiovascular system in aneurysmal subarachnoid hemorrhage not only triggers the release of atrial and brain natriuretic peptides it can also lead to increased pulmonary venous pressures and permeability causing hydrostatic pulmonary edema. Natriuretic states can herald the onset or worsening of clinical vasospasm as the renin-angiotensin-aldosterone system is activated in a delayed manner. CONCLUSIONS: This systematic review synthesizes the most current evidence of underlying mechanisms of brain related associations with body systems in aneurysmal subarachnoid hemorrhage. Results gained from these studies are clinically useful and shed light on how ruptured brain aneurysms affect the cardiopulmonary system. Subsequent neuro-cardio-endocrine responses then interact with other body systems as part of the secondary responses to primary injury.

Endovascular Therapy for Ruptured Cerebral Aneurysms in the Elderly: Poor Accessibility of the Guiding Catheter and Use of Local Anesthesia as the Predictors of Procedure-Related Rupture.

BACKGROUND: Endovascular therapy is favored for ruptured intracranial aneurysms in the elderly. However, poor accessibility to the aneurysm through the parent artery and use of local anesthesia in this age group may predispose to intraprocedural complications. OBJECTIVE: To evaluate whether age-related poor access to the ruptured target aneurysm and use of local anesthesia are associated with increased incidence of procedure-related rupture during endovascular embolization. METHODS: A total of 117 patients with 117 ruptured aneurysms underwent endovascular embolization at a single institution. Correlation of increasing age with poor accessibility of the guiding catheter was analyzed. In addition, the distance from the aneurysm to the guiding catheter was investigated to identify an association with incidence of procedure-related rupture. Correlation of local anesthesia with procedure-related rupture was also evaluated in the multivariable analysis. RESULTS: Increasing age was significantly associated with poor accessibility of the guiding catheter (P = .001, Mann-Whitney U test). Procedure-related rupture occurred in 9 of 117 aneurysms (7.7%). Longer distance between distal aneurysms and low-positioned guiding catheters carried a higher risk of procedure-related rupture than a shorter distance between proximal aneurysms and high-positioned guiding catheters (odds ratio, 19.3; 95% confidence interval, 1.84-201; P = .01, multivariable analysis). Use of local anesthesia was also a significant risk factor of procedure-related rupture by multivariable analysis. CONCLUSION: Increasing age was correlated with poor accessibility of the guiding catheter in endovascular embolization of ruptured intracranial aneurysms. Distally located aneurysms treated through a low-positioned guiding catheter and use of local anesthesia increased the risk of procedure-related rupture.

Intrasylvian hematoma caused by ruptured middle cerebral artery aneurysms predicts recovery from poor-grade subarachnoid hemorrhage.

OBJECT: Intrasylvian hematoma (ISH) is a subtype of intracranial hematoma caused by aneurysmal rupture and often presents with a poor initial neurological grade; it is not well studied. The aim of this study was to elucidate outcomes of aneurysmal subarachnoid hemorrhage (SAH) with ISH. METHODS: Data for 97 patients with poor-grade SAH (World Federation of Neurosurgical Societies Grade IV or V) were retrospectively analyzed from a single-center, prospective, observational cohort database. Ultra-early surgical clipping, removal of hematoma, external decompression for brain swelling, and prevention of vasospasm by cisternal irrigation with milrinone were combined as an aggressive treatment. Characteristics and clinical courses of SAH with ISH were identified. The authors also evaluated any correlations between poor admission-grade SAH and ISH with good functional outcome. RESULTS: Patients with poor admission-grade SAH and with ISH were more likely to have initial cerebral edema (p < 0.001, Mann-Whitney U-test), which significantly resolved overtime (p < 0.001, Mann-Whitney U-test). These patients had a better chance of functional survival (modified Rankin Scale scores of 1-3; OR 5.75; 95% CI 1.36-24.3; p = 0.017) at 6 months after hospital discharge, after adjustment for potential confounders such as younger age and better initial neurological grade by multivariable analysis. CONCLUSIONS: ISH predicted good functional recovery from poor-grade aneurysmal SAH.

Risk factors and consequences of unexpected trapping for ruptured anterior communicating artery aneurysms.

BACKGROUND: While clipping cerebral aneurysms at the neck is optimal, in some cases this is not possible and other strategies are necessary. The purpose of this study was to describe the incidence, risk factors, and outcomes for inability to clip reconstruct ruptured anterior communicating artery (ACoA) aneurysms. METHODS: Of the 70 cases of ruptured ACoA aneurysms between January 2006 and December 2013, our institutional experience revealed four cases of small ACoA aneurysms that had been considered clippable prior to operation but required trapping. When a unilateral A2 segment of anterior cerebral artery (ACA) was compromised by trapping, revascularization was performed by bypass surgery. Clinical presentation, angiographic characteristics, operative approach, intraoperative findings, and treatment outcomes were assessed. RESULTS: Very small aneurysm under 3 mm was a risk factor for unexpected trapping. The reason for unexpected trapping was laceration of the aneurysmal neck in two cases, and lack of clippaple component due to disintegration of entire aneurysmal wall at the time of rupture in the others. Aneurysms with bilateral A1 were treated with sole trapping through pterional approach in two cases. The other two cases had hypoplastic unilateral A1 segment of ACA and were treated with combination of aneurysm trapping and revascularization of A2 segment of ACA through interhemispheric approach. No patients had new cerebral infarctions of cortical ACA territory from surgery. Cognitive dysfunction was observed in three cases, but all patients became independent at 12-month follow up. CONCLUSIONS: Unexpected trapping was performed when ruptured ACoA aneurysms were unclippable. Trapping with or without bypass can result in reasonable outcomes, with acceptable risk of cognitive dysfunction.

Racial/ethnic differences in inpatient mortality and use of institutional postacute care following subarachnoid hemorrhage.

OBJECT: The goal of this study was to determine racial/ethnic differences in inpatient mortality rates and the use of institutional postacute care following subarachnoid hemorrhage (SAH) in the U.S. METHODS: A cross-sectional study of hospital discharges for SAH was conducted using the Nationwide Inpatient Sample for the years 2005-2010. Discharges with a principal diagnosis of SAH were identified and abstracted using the appropriate ICD-9-CM diagnostic code. Racial/ethnic groups were defined as white, black, Hispanic, Asian/Pacific Islander (API), and American Indian. Multinomial logistic regression analyses were performed comparing racial/ethnic groups with respect to the primary outcome of risk of in-hospital mortality and the secondary outcome of likelihood of discharge to institutional care. RESULTS: During the study period, 31,631 discharges were related to SAH. Race/ethnicity was a significant predictor of death (p = 0.003) and discharge to institutional care (p ≤ 0.001). In the adjusted analysis, compared with white patients, API patients were at higher risk of death (OR 1.34, 95% CI 1.13-1.59) and Hispanic patients were at lower risk of death (OR 0.84, 95% CI 0.72-0.97). The likelihood of discharge to institutional care was statistically similar between white, Hispanic, API, and Native American patients. Black patients were more likely to be discharged to institutional care compared with white patients (OR 1.27, 95% CI 1.14-1.40), but were similar to white patients in the risk of death. CONCLUSIONS: Significant racial/ethnic differences are present in the risk of inpatient mortality and discharge to institutional care among patients with SAH in the US. Outcome is likely to be poor among API patients and best among Hispanic patients compared with other groups.

Small bowel feeding and risk of pneumonia in adult critically ill patients: a systematic review and meta-analysis of randomized trials.

INTRODUCTION: This systematic review and meta-analysis aimed to evaluate the effect of small bowel feeding compared with gastric feeding on the frequency of pneumonia and other patient-important outcomes in critically ill patients. METHODS: We searched EMBASE, MEDLINE, clinicaltrials.gov and personal files from 1980 to Dec 2012, and conferences and proceedings from 1993 to Dec 2012 for randomized trials of adult critically ill patients in the intensive care unit (ICU) comparing small bowel feeding to gastric feeding, and evaluating risk of pneumonia, mortality, length of ICU stay, achievement of caloric requirements, duration of mechanical ventilation, vomiting, and aspiration. Independently, in duplicate, we abstracted trial characteristics, outcomes and risk of bias. RESULTS: We included 19 trials with 1394 patients. Small bowel feeding compared to gastric feeding was associated with reduced risk of pneumonia (risk ratio [RR] 0.70; 95% CI, 0.55, 0.90; P = 0.004; I2 = 0%) and ventilator-associated pneumonia (RR 0.68; 95% CI 0.53, 0.89; P = 0.005; I2 = 0%), with no difference in mortality (RR 1.08; 95% CI 0.90, 1.29; P = 0.43; I2 = 0%), length of ICU stay (WMD -0.57; 95%CI -1.79, 0.66; P = 0.37; I2 = 0%), duration of mechanical ventilation (WMD -1.01; 95%CI -3.37, 1.35; P = 0.40; I2 = 17%), gastrointestinal bleeding (RR 0.89; 95% CI 0.56, 1.42; P = 0.64; I2 = 0%), aspiration (RR 0.92; 95% CI 0.52, 1.65; P = 0.79; I2 = 0%), and vomiting (RR 0.91; 95% CI 0.53, 1.54; P = 0.72; I2 = 57%). The overall quality of evidence was low for pneumonia outcome. CONCLUSIONS: Small bowel feeding, in comparison with gastric feeding, reduces the risk of pneumonia in critically ill patients without affecting mortality, length of ICU stay or duration of mechanical ventilation. These observations are limited by variation in pneumonia definition, imprecision, risk of bias and small sample size of individual trials.