Helen E. Robson

Survival analysis after gastrostomy: a single-centre, observational study comparing radiological and endoscopic insertion.

Objectives Percutaneous endoscopic gastrostomy (PEG) using the pull through technique is the most widely used insertion method. An alternative is a per-oral image-guided gastrostomy (PIG), which may be advantageous in high-risk patients. As there are no large studies comparing PEG against PIG, we wished to analyse mortality after endoscopic or radiological gastrostomy insertion. Methods Patients referred for a gastrostomy are prospectively included in a database along with demographic, biochemical and outcome data. Analysis of gastrostomy insertions from February 2004 to 2007 was performed with reference to method of insertion and 30-day outcome. Patients were subgrouped into cognitive impairment, dysphagic stroke, oropharygeal cancer, neurological cancer and other. Results One hundred and seventy PIGs and 233 PEGs were inserted (mean age 62 years, 268 males). There were no differences in age between the PIG and the PEG group. The PIG 30-day mortality was 26 of 170 (15.3%) and the PEG 30-day mortality was 25 of 233 (10.7%) (P=0.17). One-year mortality was 92 of 170 (54.1%) for PIG and 131 of 233 (56.7%) for PEG (P=0.60). Subgroup analysis revealed higher 30-day mortality in patients with nasopharyngeal cancer undergoing PIG; 14 of 106 (13.2%) versus one of 69 (1.4%) (P=0.005). However, PIG patients were older than PEG patients (mean age 64 vs. 59.7 years, P=0.019) and had more comorbidities (21.1% in the PEG group and 37.7% in the PIG group). Conclusion Overall PIG and PEG seem to have similar 30-day and 1-year mortality rates. Our data suggest that clinicians may opt for either approach depending on technical considerations and local availability.

Mortality Among Patients Who Receive or Defer Gastrostomies

BACKGROUND & AIMSnThere are few data on outcomes and mortality of patients who have received gastrostomies. We assessed 30-day and 1-year mortalities of patients in the United Kingdom who were referred to hospitals for gastrostomies and of patients who deferred this intervention.nnnMETHODSnWe collected data from 1327 patients referred to 2 hospitals in Sheffield, United Kingdom, forxa0gastrostomies from February 2004 through May 2010. Data were analyzed to determinexa030-day and 1-year mortalities. Predicted mortality by using the validated Sheffield Gastrostomy Scoring System (SGSS) was then compared with actual mortality by using area under the receiver operator curves to determine levels of agreement in patients referred for gastrostomy.nnnRESULTSnThree hundred four patients (23%) did not undergo gastrostomy after multidisciplinary team discussion, which was based on physicians recommendations. This group had 35.5% mortality at 30 days and 74.3% at 1 year, whereas mortality among patients who underwent gastrostomy (nxa0= 1027) was 11.2% at 30 days and 41.1% at 1 year (P < .0001, compared with patients who deferred the procedure). The area under the receiver operator curves for the SGSS demonstrated acceptable agreement between predicted and actual mortality in patients who underwent or were deferred gastrostomy.nnnCONCLUSIONSnOn the basis of data from 1327 patients, those who undergo gastrostomy have significantly lower mortality than those who defer the procedure. Without applying the SGSS, clinicians are able to select patients most likely to benefit from gastrostomy. The SGSS could provide objective support to clinicians involved in making ethically contentious or potentially litigious decisions.

Albumin level and patient age predict outcomes in patients referred for gastrostomy insertion: internal and external validation of a gastrostomy score and comparison with artificial neural networks

BACKGROUNDnSignificant mortality after gastrostomy insertion remains and some risk factors have been identified, but no predictive scoring system exists.nnnOBJECTIVEnTo identify risk factors for mortality, formulate a predictive scoring system, and validate the score. Comparison to an artificial neural network (ANN).nnnDESIGNnEndoscopic database analysis.nnnSETTINGnSix hospitals (2 teaching hospitals) in the South Yorkshire region, United Kingdom.nnnPATIENTSnThis study involved all patients referred for gastrostomy insertion.nnnINTERVENTIONnGeneration of clinical scores to predict 30-day mortality in patients undergoing gastrostomy insertion.nnnMAIN OUTCOME MEASUREMENTSnRisk factors for 30-day mortality. Internal and external validation of the score. Comparison with an ANN.nnnRESULTSnUnivariate analysis showed that 30-day mortality was associated with age, albumin levels, and cardiac and neurological comorbidities. Multivariate analysis showed that only age and albumin levels were independent. Modeling provided scores of 0, 1, 2, and 3 corresponding to 30-day mortalities of 0% (0-2.1), 7% (2.9-13.9), 21.3% (13.5-30.9), and 37.3% (24.1-51.9), respectively. Application of the scoring system at the other teaching hospital and the 4 district general hospitals gave 30-day mortality rates that were not significantly different from those predicted. Receiver operating characteristic curves for the score and the ANN were comparable.nnnLIMITATIONSnNonrandomized study. Score not used as a decision-making tool.nnnCONCLUSIONnThe gastrostomy score provides an estimate of 30-day mortality for patients (and their relatives) when gastrostomy insertion is being discussed. This score requires evaluation as a decision-making tool in clinical practice. ANN analysis results were similar to the outcomes from the clinical score.

External validation of a prognostic scoring system for percutaneous endoscopic gastrostomy (PEG)

Introduction There is a significant mortality associated with PEG. The Sheffield Gastrostomy Score (SGS) was devised to try and improve outcomes following this procedure by predicting 30-day mortality. After prospectively collecting demographic, biochemical and outcome data from a cohort of patients having a new PEG inserted (Hospital A, n = 403), multivariate analysis identified that age and albumin were determinants of 30-day mortality (p < 0.001). Age and albumin were attributed scores in a scoring system, the SGS, with age scoring 0 or 1 (<65/≤65 years) and albumin scoring 0, 1 or 2 (>35, 25–34 and <25 g/L). Composite scores in the SGS of 0 to 3 corresponded to 30-day mortalities of 0%, 7%, 21.3% and 37.3%, respectively. The SGS was then validated on a separate cohort of patients from a second hospital in Sheffield, from within the same trust (Internal validation: Hospital B, n = 153) with comparable 30-day mortality figures. Thus our aim was to externally validate the SGS (outside of Sheffield) and determine its potential application to a wider population. Methods Four local district general hospitals were used to externally validate the SGS. Demographic, biochemical and outcome data was obtained from all patients undergoing PEG insertion between June 2006 and June 2009. χ2 analysis was used to compare actual 30-day mortality rates in the external cohort to the rate predicted by the SGS. Results A total of 679 new PEGs were inserted during this period in these four hospitals. Complete records were obtained from 88% (597/679) and analysed (median age 71, 322 male). The 30-day mortality for this external validation cohort was 13.6%. The main indications for PEG insertion were stroke (33.7%), oropharyngeal malignancy (32.3%) and neurodegenerative diseases (17.4%). Univariate analysis predictors of 30-day mortality in this external cohort were: increasing age, low serum albumin and stroke. There were no significant differences between the observed and the predicted mortality (p ≤ 0.05) in the external validation cohort using the SGS. Conclusion Our results suggest that the SGS can accurately stratify patients into different risk categories, with higher scores correlating with significantly lower survival. We feel that the SGS could have an important role in those patients where the benefits and risks of gastrostomy are unclear. Table 1 OC-033 PEG score in internal and external validation cohort Score Predicted 30 day mortality Hospital A Internal validation cohort Hospital B Observed mortality (p value) External validation cohort Observed mortality (p value) 0 0% 0% (1) 3.2% (0.57) 1 7% 7.7% (1) 10.1% (0.42) 2 21.30% 15.2% (0.41) 20.3% (0.88) 3 37.30% 29.3% (0.51) 31.6% (0.66)

Outcomes following gastrostomy: radiologically-inserted v. percutaneous endoscopic gastrostomy

Gastrostomy insertion has been demonstrated to be of benefit in selected patients. Percutaneous endoscopic gastrostomy (PEG) using the pull-through technique is the most widely used insertion method, but it is recognised to have important complications, particularly in patients with respiratory risk factors. An alternative is a radiologically-inserted gastrostomy (RIG). It has been suggested that RIG may be advantageous in patients who are potentially at ‘high risk’ from respiratory complications. However, there are no large studies comparing PEG v. RIG. All patients referred for a gastrostomy are prospectively included in a database along with demographic, biochemical and outcome data. Analysis of gastrostomy insertions over the period February 2004–February 2007 was performed with reference to method of insertion and outcome at 30 d. Selection for method of insertion is left to the discretion of the referring clinician. Patients were allocated to the following subgroups: cognitive impairment (n 5); dysphagic stroke (n 36); nasopharygeal cancer (n 175); neurological (n 116); other (n 71). Over the study period 170 RIG and 233 PEG were inserted (mean age 62 years, 268 males). There were no differences in age between the RIG group and PEG group and case mix was comparable except in the nasopharyngeal cancer group (proportionally more RIG). The RIG 30 d mortality was twenty-six of 170 (15.3%) and the PEG 30 d mortality was twenty-five of 233 (10.7%; P= 0.17). Mortality at 1 year was ninety-two of 170 (54.1%) for RIG and 131 of 233 (56.7%) for PEG (P= 0.60). Within subgroups the only significant difference in 30 d mortality was in those patients with nasopharyngeal cancer: fourteen of 106 (13.2%) for RIG and one of 69 (1.4%) for PEG (P = 0.005). However, patients referred for RIG were significantly older than those referred for PEG (mean age (years) 59.7 v. 64; P= 0.019) and had a higher prevalence of significant comorbidities (21.1% in the PEG group and 37.7% in the RIG group). Overall, RIG and PEG appear to have similar 30 d mortality rates. In patients with nasopharyngeal cancer there was a higher mortality in those referred for RIG; however, pre-selection by the referring clinician as a result of perceived risk of endoscopic insertion may have biased the outcome. A randomized trial comparing both methods in this subgroup is needed.

Predicting outcomes in patients deferred for gastrostomy insertion using either the Sheffield gastrostomy score or the Levine score

Gastrostomy insertion is widely accepted as the best means of providing medium and long term enteral nutrition. There is limited data on outcomes in patients referred for gastrostomy insertion that are either refused or do not survive until PEG insertion. Scoring systems such as the Levine score have been created to try and predict survival rates over 1 year for patients admitted acutely. The aim of this study was to determine the clinical characteristics and outcomes of patients in whom gastrostomy insertion was not felt to be appropriate on clinical grounds and the Levine score and the Sheffield gastrostomy score (SGS) applied. All patients referred for gastrostomy insertion in our institute are reviewed by the gastrostomy nurse specialist and prospectively included in a database. Demographics, referral indication, biochemical profile and outcome data are also included in the database. Gastrostomy referrals from October 2003 to September 2010 were analysed and actual mortality rate was compared to the expected mortality rate predicted by the SGS and Levine score. 304 patients were included in the study (median age 77 years, 175 males) all of whom were referred for consideration of gastrostomy insertion but were ultimately not inserted. The 3 top reasons for non-placement were patient too ill (29.3%), oral intake restarted (18.1%) and patient died (13.5%). Overall mortality at 30 days and 1 year was 122/304 (40.1%) and 226/304 (74.3%) respectively. The highest 30 day mortality was seen in patients with cognitive impairment 12/17 (70.6%) and followed by dysphagic stroke 53/105 (50.5%). The main predictors of 30 day mortality were age >60 (OR 7.2, 3.3–15.8, p<0.001) and albumin <25 (OR 2.5, 1.5–5.3, p = 0.01). 1 year mortality as predicted by Levine score or Sheffield score is shown in the table. The Sheffield gastrostomy score is more accurate at predicting 1-year mortality in patients deferred for gastrostomy insertion than the Levine score. This may suggest that the Sheffield PEG score could be used in the pre-assessment of patients who are being referred for PEG as a means of estimating/predicting mortality for those involved in the decision making process.

PTU-011 Predicting outcome in patients deferred for percutaneous endoscopic gastrostomy: is the Levine score or the Sheffield gastrostomy score more accurate?: Abstract PTU-011

Introduction Percutaneous endoscopic gastrostomy (PEG) insertion is widely accepted as the best means of providing medium and long-term enteral nutrition. There are limited data on outcomes in patients referred for gastrostomy insertion that are either refused or do not survive until PEG insertion. Scoring systems such as the Levine score have been created to try and predict survival rates over 1u2005year for any patients admitted acutely. We have also created a score to specifically predict mortality after gastrostomy insertion. These two scores have not been directly compared. The aim of this study was to identify which scoring system is the most useful for predicting survival after PEG. Methods All patients referred for PEG insertion in our institute are reviewed by the PEG nurse specialist and prospectively included in a database. Demographic, biochemical and outcome data are also included. PEG referrals from March 2007 to March 2009 were analysed and both the Levine score and the Sheffield gastrostomy score calculated. Mortality rates were then compared. Results During the study period 78 patients (mean age 69.5u2005years) were assessed for gastrostomy insertion but were turned down. Overall 30-day and 1u2005year mortality in this group was 33.3% and 73.1%, respectively. 1-year mortality as predicted by Levine score or Sheffield score is shown in Abstract 011. Abstract PTU-011 Expected mortality Actual mortality p Sheffield 0 15/39 (38.5%) 0/5 (0.0%) 0.14 Sheffield 1 53/100 (53.0%) 5/12 (41.7%) 0.54 Sheffield 2 68/94 (72.3%) 16/22 (72.7%) 1 Sheffield 3 40/51 (78.4%) 36/39 (92.3%) 0.08 Levine 0–1 110/799 (13.8%) 14/29 (48.3%) <0.001 Levine 2 130/719 (18.1%) 7/10 (70.0%) <0.001 Levine 3 180/563 (31.9%) 26/29 (89.7%) <0.001 Levine 4+ 299/647 (46.2%) 10/10 (100%) <0.001 Conclusion The Sheffield score is more accurate at predicting 1-year mortality in patients deferred for gastrostomy insertion. This may suggest that the Sheffield PEG score could be used in the pre-assessment of patients who are being referred for PEG as a means of estimating/predicting mortality for those involved in the decision making process. Larger comparative studies are required to validate this finding.

PTU-005 Neurodegeneration as a prognostic factor post gastrostomy insertion: to PEG or not to PEG?: Abstract PTU-005

Introduction Gastrostomy insertion is well-accepted as a method for medium and long-term feeding. Although gastrostomy insertion in dementia is recognised as having a poor outcome there is a paucity of data pertaining to other neurodegenerative disease (examples: motor neurone disease, Parkinsons disease and multiple sclerosis). Methods The aim of the study was to compare the mortality of patients with neurodegenerative disease undergoing gastrostomy with other subgroups requiring the procedure. Data for all patients referred for PEG insertion at our institute are prospectively included in a database. PEG referrals from February 2004 to February 2007 were analysed. Patient demographics, indications and outcome were extracted. Group 1 acute brain injury (acute stroke, head injury, subarachnoid haemorrhage), Group 2 oropharyngeal malignancy, Group 3 neurodegenerative diseases and Group 4 other/miscellaneous (cystic fibrosis, cerebral palsy, severe pneumonia, mucopolysaccharidoses, oesophageal stricture). Results 403 gastrostomies were performed (268 males, median age 64, range 18–94). Patients with neurodegenerative diseases have a worse outcome by comparison to other disease subgroups at 1u2005year (with a mortality of 74%, p<0.002). This did not alter after adjusting for age at time of gastrostomy. Conclusion This is the first study to observe that patients with neurodegenerative disease have a higher mortality at 1u2005year after gastrostomy insertion than other subgroups referred. This may reflect the underlying disease. This observation has implications when discussing timing of gastrostomy and obtaining informed consent. Abstract PTU-005 30-day mortality 90-day mortality 1u2005year mortality All n=403 51/403 (12.7%) 81/403 (20.1%) 223/403 (55.3%) Acute brain injury n=52 8/52 (15.4%) 15/52 (28.8%) 24/52 (46.2%) Oropharyngeal cancer n=180 16/180 (8.9%) 22/180 (12.2%) 84/180 (46.7%) Neurodegenerative disease n=85 14/85 (16.5%) 17/85 (20.0%) 63/85 (74.1%)* Other n=86 13/86 (15.1%) 27/86 (31.4%) 52/86 (60.5%) * p<0.002

Outcomes Following Gastrostomy: Radiologically Inserted Vs. Percutaneous Endoscopic Gastrostomy

Gastrostomy insertion has been demonstrated to be of benefit in selected patients. Percutaneous endoscopic gastrostomy (PEG) using the pull-through technique is the most widely used insertion method, but it is recognised to have important complications, particularly in patients with respiratory risk factors. An alternative is a radiologically-inserted gastrostomy (RIG). It has been suggested that RIG may be advantageous in patients who are potentially at ‘high risk’ from respiratory complications. However, there are no large studies comparing PEG v. RIG. All patients referred for a gastrostomy are prospectively included in a database along with demographic, biochemical and outcome data. Analysis of gastrostomy insertions over the period February 2004–February 2007 was performed with reference to method of insertion and outcome at 30 d. Selection for method of insertion is left to the discretion of the referring clinician. Patients were allocated to the following subgroups: cognitive impairment (n 5); dysphagic stroke (n 36); nasopharygeal cancer (n 175); neurological (n 116); other (n 71). Over the study period 170 RIG and 233 PEG were inserted (mean age 62 years, 268 males). There were no differences in age between the RIG group and PEG group and case mix was comparable except in the nasopharyngeal cancer group (proportionally more RIG). The RIG 30 d mortality was twenty-six of 170 (15.3%) and the PEG 30 d mortality was twenty-five of 233 (10.7%; P= 0.17). Mortality at 1 year was ninety-two of 170 (54.1%) for RIG and 131 of 233 (56.7%) for PEG (P= 0.60). Within subgroups the only significant difference in 30 d mortality was in those patients with nasopharyngeal cancer: fourteen of 106 (13.2%) for RIG and one of 69 (1.4%) for PEG (P = 0.005). However, patients referred for RIG were significantly older than those referred for PEG (mean age (years) 59.7 v. 64; P= 0.019) and had a higher prevalence of significant comorbidities (21.1% in the PEG group and 37.7% in the RIG group). Overall, RIG and PEG appear to have similar 30 d mortality rates. In patients with nasopharyngeal cancer there was a higher mortality in those referred for RIG; however, pre-selection by the referring clinician as a result of perceived risk of endoscopic insertion may have biased the outcome. A randomized trial comparing both methods in this subgroup is needed.