Percutaneous endoscopic gastrostomy: 30-day mortality trends and risk factorsSimon EJ Janes, C SG Price, S Khan
Department of surgery, New Cross Hospital, Wolverhampton, England WV1 0QP, United Kingdom
BACKGROUND AND AIMS: 30-day Percutaneous endoscopic gastrostomy (PEG) mortality of 8% (1992). Recent concerns suggest that mortality may have increased, prompting a comparison of current practice with that reported earlier. MATERIALS AND METHODS: Data regarding PEG insertion with relation to case mix, complications, 30-day mortality and associated risk factors, in 2002, in a British University Hospital was compared with that in 1992. Logistic regression analysis was used to determine factors independently predictive of 30-day mortality. RESULTS: In 2002, 112 patients (70% males, mean age 67.5 years; 1992: 63.6 years) underwent PEG. The 30-day mortality increased significantly from 8% (1992) to 22% (2002), P= 0.03. During this time, PEG insertion rate increased ten-fold, however, procedure-related mortality decreased from 2% to nil. In terms of percentage, the indications for PEG in 1992 and 2002 respectively were: cerebrovascular disease (33/25), head and neck tumours (16/24), motor neuron disease (27/11, P= 0.01). The proportion of PEGs for non-evidence-based indications increased from 16% in 1992 to 31% in 2002, P= 0.048. The number of PEGs placed radiologically increased (0/17, P= 0.02). Radiological patients received less antibiotic prophylaxis (P< 0.001) and had more PEG site infections than standard placement, P= 0.04. Multivariate analysis identified nil by mouth 7 days or 11.4 (CI 3.2-41.7), albumin 30 g/L or 12 (2.2-66.7) and >1 cardiac factor or 5.1 (1.02-25.6) as independent predictors of 30-day mortality. CONCLUSIONS: The ten-fold rise in the PEG insertion rate has been accompanied by a three-fold rise in 30-day mortality. This may reflect a lowered threshold of PEG insertion. The risk factors identified may help decision-making in cases where the risk-benefit relationship is not clear-cut.
Keywords: Percutaneous endoscopic gastrostomy, mortality, risk factors
Since its initial description over two decades ago, percutaneous endoscopic gastrostomy (PEG) has become the most widely employed method for long-term enteral nutrition. It is generally better-tolerated and causes less reflux or aspiration than nasogastric tube feeding. The benefit of PEG, in terms of morbidity and mortality, is well-established in patients with cerebrovascular disease, oropharyngeal malignancy and motor neuron disease (MND). However, the demands for PEG insertion have increased to include illnesses where the long-term benefit of PEG feeding is less certain. Consequently, in Western countries the number of patients on home enteral nutrition has doubled in the past five years.
The growing number of gastrostomies may reflect a change in patient selection or alternations in the clinicians' attitudes to enteral feeding, as the prevalence of disease conditions warranting PEG has not changed two-fold during the same period. Furthermore, recent evidence indicates 30-day mortality has, in fact, increased,, paralleling the change in case mix.
The aim of the study was to compare the case mix and mortality with a previously published audit from the our institution. We also determined factors predictive of 30-day mortality, in order to help identify patients at high-risk of mortality.
The data of subjects who underwent PEG in a teaching hospital in the UK during a 12-month period beginning April 2001 was analysed after its retrieval using a hospital coding database and the International Classification of diseases 10th Edition (ICD-10) codes G448, G341 and G342. The follow-up data for a period of six months was also available. Referrals for gastrostomy were made from geriatricians, neurologists, Ear, Nose and Throat (ENT) and maxillo-facial surgeons, general physicians, the adult Intensive Care Unit (ICU), and neurosurgeons.
An endoscopist or radiologist confirmed suitability for gastrostomy and informed consent was obtained. Patients without the capacity to give informed consent had a declaration signed by the responsible consultant stating that the procedure was in the patient's best interest, in accordance with the United Kingdom Law.
The team responsible for the patient's care rather than the endoscopist decided the use of antibiotic prophylaxis. Routine investigations before the procedure included: full blood count, clotting time, urea, electrolytes, and liver function tests. Xylocaine throat spray was used for anaesthesia and intravenous midazolam administered for sedation. Nine or 15 gauge FrekaTM gastrostomy tubes were inserted via the pull technique, or under fluroscopic guidance by a radiologist. Patients who underwent fluoroscopic guided PEG insertion by a radiologist formed the 'Radiological Group'. Patients received intravenous fluids for 24 hours, and unless there were complications, water was passed down the PEG six hours after insertion. Enteral feeding generally began the following day. Patients undergoing laparoscopic, open or CT-guided gastrostomy were excluded from the analysis.
A standardised proforma was used to record patient age, gender, Glasgow Coma Score (GCS) on admission, number of medications on admission, main presenting complaint, past medical history including detailed cardiovascular history (myocardial infarction, ischaemic heart disease, congestive cardiac failure, atrial fibrillation, hypertension), Oxford Community Stroke Project Classification (for stroke patients), referring speciality, consecutive days nil by mouth from admission until peg placement, total days nasogastric feeding prior to PEG, time from admission until PEG, history of pneumonia or aspiration pre-peg, catheterisation or Urinary Tract Infection (UTI) pre-PEG, type of antibiotic prophylaxis (if used), routine blood tests, date of procedure, and pre-procedure pulse, blood pressure and oxygen saturation. Complications and post-procedure infections were noted if they were documented in the medical or nursing notes.
The following conditions were considered evidence-based PEG referrals: cerebrovascular disease, ENT tumours, MND and multiple sclerosis (MS).,,, Risk factors for 30-day mortality were analysed statistically using odds ratios (OR). Categorical data were compared using the chi-square test or Fishers exact test, and the students' t test was used to compare the means of normally distributed continuous variables. OR are expressed with 95% confidence intervals (CI), P< 0.05 was considered significant throughout. Logistic regression analysis was used to determine factors independently predictive of 30-day mortality. Kaplan-Meier survival curves were plotted to determine factors affecting survival and the log rank test was applied to compare differences in survival between groups. Factors significantly affecting survival were entered into a stepwise Cox regression model to determine independent predictors of survival. Data was analysed using the Statistical Package for the Social Sciences version 10.
The study population in 2002 consisted of 112 patients with mean age of 67.5 years (1992: 63.6 years). Seventy per cent of the subjects were male. A significant cardiovascular history was present in 44 (39%) patients (range 1-4 factors) [Table - 1]. Sixty-three per cent received nasogastric feeding before PEG, and 26 (23%) were nil by mouth for 7 consecutive days before gastrostomy. Of these patients, the majority had suffered either a substantial CVA (with GCS <8) or were in the ICU for treatment of sepsis. The mean number of medications received on admission was 3.9 per patient (SD 2.7) with as many as 47 (42%) patients taking 5 or more drugs. Over 60% of the patients were referred from the geriatric or neurological services [Table - 2].
Antibiotic prophylaxis was given to 58 (52%) patients. A vast majority received cefuroxime (48 patients, 42%). However there was no evidence that this reduced the infection rates (OR 0.6, 95% CI 0.3 -1.4).
Six patients (5%) died within the first week. Thirty-day mortality was 22% (25 patients, 20 endoscopic PEG, five radiological), 90-day mortality 30% and 180-day mortality 39%. The mean time from admission until PEG was similar for those who died within 30 days (17.6) and those who lived for over 30 days (20.6), P>0.5.
Post-procedure complications occurred in 37 (33%) patients. PEG site infections occurred in 13 (12%) subjects with three patients having methicillin-resistant Staphylococcus aureus infections. Twenty-two patients developed chest infections (20%), 11 (10%) of which followed aspiration. Five patients (4%) developed diarrhoea, two (2%) of which were due to Clostridium difficile.
Univariate analysis showed that patients presenting with a GCS <10, receiving at least 5 medications on admission, age over 75 years or more than one cardiac risk factor had an increased risk of death within 30 days [Table - 3]. Patients with the following risk factors identified between admission and PEG also had an increased risk of 30-day mortality: history of pneumonia or aspiration, serum albumin concentration 30 g/L, urea 9 mmol/L, withholding of oral feeds for at least 7 days and haemoglobin concentration lower than 10 g/dL [Table - 3]. Of five patients with pneumonia as main diagnosis on admission, three (60%) died within 30 days, indicating a clinically important but not statistically significant trend, [OR 5.8 (0.6-71.8)]. Use of antibiotic prophylaxis or male gender did not affect mortality [Table - 3].
When these factors were entered into a multivariate model the independent predictors of 30-day mortality were: withholding of oral feeding for at least 7 days [OR 11.4 (3.2-41.7)], serum albumin 30 g/L OR 12 (2.2-66.7) and >1 cardiac factor, OR 5.1 (1.02-25.6) [Table - 4].
The following factors independently affected survival during follow-up (Cox regression analysis): withholding feeds for at least 7 days, [hazard ratio (HR) 2.3 (1.4-3.4, P< 0.000)], serum albumin concentration 30 g/L [HR 2.8 (1.4-5.6, P= 0.004)] and history of pneumonia [HR 2.1 (1.13.8, P= 0.017)]. Kaplan-Meier curves for these factors are represented in [Figure - 1], [Figure - 2], [Figure - 3] respectively.
The mean age was similar in the radiological group (67.3 years, SD 13.6) and the standard group (67.5 years, SD 14.4). Although proportionately more patients died after radiological PEG placement (26%), the trend was not significant, OR 1.3 (0.3-4.4), P>0.5. The radiological group had more patients with head and neck tumours (47%), or MND (32%) than the standard group (19% and 6%), P= 0.02 and P= 0.003 respectively (Fisher's exact test). Only one (5%) patient in the radiological group had cerebrovascular disease as a main diagnosis, compared to 33 (35%) in the standard placement group, P= 0.009, 2 6.8, df 1.
Radiological PEG patients received less prior antibiotic prophylaxis when compared to standard placement patients (16% and 59% respectively, P<0.001, 2 11.9, df 1) and had more PEG site infections (five vs. eight patients, P= 0.04, Fisher's exact test). Despite this, rates of post-procedure complications were almost identical (32% radiological group, 33% standard group). Home discharge was achieved more often in the radiological group (68% vs. 38%, P=0.01, 2 6.1, df 1).
Forty-nine patients (44%) achieved home discharge. Eighteen patients (16%) were discharged to nursing homes and 23 (21%) died in the hospital during the index admission. Thirteen patients (12%) had their PEG removed after 6 months.
Comparison between 1988-92 and the current project
The 30-day mortality increased from 8% in 1992 to 22% in 2002, P= 0.03, 2 4.6, df 1. [Table - 5] summarises the indications for PEG in 1992 and 2002. The number of PEGs for non-evidence-based conditions rose from 16% in 1992 to 31% in 2002, P= 0.048, 2 3.9, df 1 [Figure - 4], [Table - 5]. A similar proportion had cerebrovascular disease (29% 1992, 33% current study). However, less patients had MND in the current study (11% vs. 27%, P= 0.01, 2 6.0, df 1). Home discharge rates were lower in the present study than in 1988-92 (43% vs. 69%) P= 0.002, 2 9.6 df 1. Patients in the present study tended to be older, mean age 67.5 vs. 63.6 years, P=0.08. There were no radiological PEGs in the original study compared to 19 in the current study.
During the last ten years, PEG placement has increased ten-fold whereas the procedure-related mortality has decreased from 2% to nil Thus although technical skill improved, the threshold for PEG insertion and the ability to select patients who would survive 30 days has decreased significantly. Patients in the recent cohort had more coexisting medical problems before PEG placement, which had a substantial impact on mortality.
Ten years ago, more than 80% of PEGs were placed in patients with CVD, MND, ENT tumours or MS. This proportion fell to 69% in the current series, due to an increase in PEG placement for acute medical conditions where the long-term benefits of PEG are unproven. Patients admitted for pneumonia illustrate this point: of the five patients with pneumonia as primary diagnosis, three died within 30 days. No patient in the previous study had pneumonia as the main reason for admission. Furthermore, of the six patients referred from the ICU in the current series, only two survived 30 days whereas 10 years previously there were no ICU patients. Comparison of the two studies is partly limited because the original study was prospective and the current study was retrospective. However, this should not have affected mortality rates, as our follow-up was complete except for one set of case notes.
To date no study has demonstrated a survival benefit in patients with pneumonia undergoing PEG. In the largest study to date Grant et al retrospectively reviewed 81,000 American medicare beneficiaries who underwent PEG. The most common primary diagnoses were CVD and neoplasia, with a 30-day mortality of 25%. They found that 30-day mortality was highest among those with non-aspiration pneumonia. Others have found that aspiration pneumonia was a risk factor for 30-day mortality and it is suggested that patients with pneumonia gain no improvement in nutritional or functional status when treated with PEG. Our data support these conclusions and shows an adverse effect on overall survival in addition to 30-day mortality.
The significant reduction in the home discharge rate in the current series is further evidence of a change in patient selection, and has considerable cost implications for healthcare providers. Ten years ago PEG was perceived as cost-effective as it allowed early domiciliary discharge for dysphagic patients and had low post-procedure complication rates.
The issue of when to start enteral feeding remains controversial. Abuksis et al recently reported a 40% reduction in PEG mortality by delaying PEG insertion until 30 days after discharge, when compared with standard insertion. In a review of 161 consecutive elderly PEG patients, Raha et al found high mortality rates using PEG for elderly malnourished patients, suggesting that earlier nutritional intervention may help prolong life. A study of 55 patients with dementia and inadequate oral intake showed that PEG was an ineffective treatment for hypoalbuminaemic patients, suggesting that enteral feeding should commence before hypoalbuminaemia develops. In that study serum albumin <28 g/L was an independent predictor of six-month mortality. Our data support these findings as serum albumin 30 g/L and nil by mouth for 7 days were independent predictors of both 30-day mortality and survival.
Withholding oral feeding for at least seven days has not previously been described as a marker for 30-day mortality and probably reflects the severity of underlying disease. However, poor nutritional status may not be a true independent predictor because the underlying disease process may be associated with both malnutrition and poor outcome. Although malnourished patients fair worse, the limitation of retrospective analysis is that we are unable to conclude whether PEG prolongs survival, or is an exacerbating factor in an inevitable downhill course. Wolfsen et al followed patients with very short life expectancy, showing that PEG reduced hospital stay but did not contribute to life expectancy. There is a consensus that mortality is high when acutely ill patients undergo PEG,,,, leading some to suggest that PEG should be delayed in this group until the acute illness has resolved.,, Others advocate a trial of nasogastric feeding until recovery from the acute illness when PEG could be reconsidered.,
The dilemma we face is in trying to differentiate patients who benefit from early PEG feeding from those who would not survive for 30 days. For patients clearly in the later category, PEG is an expensive and invasive procedure that should be avoided. For those in the former, consideration of risk factors may help predict those with a poor prognosis. However, whilst poor prognostic indicators may be valid for a large patient cohort, they may not apply to separate cases and each patient should be treated individually. The balance of benefit and harm related to PEG is difficult to determine in patients with multiple co-morbidities. The risks associated with PEG should be made clear to the patient and their family, and any decision reached should be ethically justifiable.
Improving the quality of life is one long-term goal of PEG feeding although there is little data on the quality of life after PEG. Two recent studies assessing the quality of life after PEG were limited by response rates of 7-14%,, due to many patients having communication difficulties or due to loss to follow-up. Of those who responded 83% had no improvement in functional disability (Modified Rankin Scale) and only 45% thought PEG had a positive effect on their quality of life. Assessing the quality of life is beyond the scope of this study, however, it was interesting to note that one in three patients suffered post-PEG complications that would have adversely affected the quality of life.
Various risk factors for 30-day mortality have been identified, including age over 75 years,, urinary tract infection, dementia, diabetes mellitus and severe functional impairment. Grant et al found that a secondary diagnosis of congestive cardiac failure in PEG patients adversely affected 30-day mortality. Our results show that multiple cardiac pathology also strongly predicts both 30-day mortality and overall survival. These patients need careful risk-benefit evaluation before invasive procedures such as PEG. Many of the cardiac patients had five or more medications on admission, which may explain why receiving five or more medications was not a significant factor using multivariate analysis. Polypharmacy is an established predictor of hospital mortality in patients in a general geriatric ward, but has not previously been associated with PEG mortality.
Although antibiotic prophylaxis was associated with a reduction in PEG site infections in the standard group when compared to the radiological group, this study was not designed to detect such a difference, and these findings should be interpreted with caution. Nonetheless, it is disappointing that only half the patients received antibiotic prophylaxis, despite good evidence of its benefit.
Thirty-day PEG mortality has increased significantly since 1992. The ten-fold increase in PEGs reflects increased demands for PEG, encompassing conditions where the long-term benefits are uncertain. Increasing PEG placement combined with decreased home discharge rates has important cost implications. The risk factors identified may help decision-making in cases where the risk-benefit relationship is not clear-cut. Future research should identify patients who will benefit most from PEG, hence avoiding unnecessary procedures in patients with a short life expectancy.
[Figure - 1], [Figure - 2], [Figure - 3], [Figure - 4][Table - 1], [Table - 2], [Table - 3], [Table - 4], [Table - 5]