A prospective cohort study of risk prediction in simultaneous pancreas and kidney transplantation

Abstract

Introduction

Current risk prediction scoring systems in pancreas transplantation are limited to organ factors and are specific to predicting graft outcome. They do not consider recipient factors or inform regarding recipient morbidity. The aim of this study was to assess the utility of commonly used general surgical risk prediction models (P-POSSUM [Portsmouth Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity], MODS [multiple organ dysfunction score], Charlson co-morbidity index, revised cardiac risk index, ASA [American Society of Anesthesiologists] grade and Waterlow score), and to correlate them with total length of hospital stay (LOS) and critical care unit (CCU) LOS, important surrogate markers of patient outcome.

Methods

All risk prediction scores were calculated prospectively for all simultaneous pancreas and kidney (SPK) transplant recipients from November 2011 to October 2013, and correlated with outcome measures.

Results

Overall, 57 SPK transplant recipients were analysed. The mean age was 42.0 years (standard deviation [SD]: 7.60 years), 27 (52%) were male and the mean body mass index was 25.43kg/m² (SD: 3.11kg/m²). The mean pancreas and kidney cold ischaemic times were 703 minutes (SD: 182 minutes) and 850 minutes (SD: 192 minutes) respectively. The median total LOS and mean CCU LOS was 17 days (range: 8–79 days) and 7 days (SD: 4.04 days) respectively. When correlated with risk prediction scores, Waterlow score was the only significant predictor of total LOS and CCU LOS (p<0.001 [Spearman’s correlation] and p=0.001 [Pearson’s correlation] respectively).

Conclusions

Preoperative risk prediction plays an important part in planning perioperative care. To date, no validated risk prediction scoring system exists for SPK transplantation. This prospective study indicates that Waterlow score identifies high risk individuals and has value in the prediction of outcome following SPK transplantation.

Risk and outcome prediction scores are required prior to major surgery to aid in the counselling and consent processes, and in planning intra and postoperative care. The search for a reproducible and accurate model has been ongoing for over 30 years.¹ Several scoring systems are established and validated with variable sensitivity in the context of general surgery (predominantly abdominal and vascular), to help stratify and quantify an individual’s risk in the perioperative period. These include POSSUM (Physiological and Operative Severity Score for the enUmeration of Mortality and morbidity),² MODS (multiple organ dysfunction score),³ Charlson co-morbidity index,⁴ revised cardiac risk index⁵ and ASA (American Society of Anesthesiologists) grade.⁶ They are all readily available in clinical practice and are used in decision making prior to undertaking major general surgery.

Despite the number of validated risk prediction scoring mechanisms available in major general surgery, there is a relative paucity of quantitative tools in transplantation. An ideal transplant specific scoring system would require the consideration of recipient, donor and organ factors, and aim to predict organ survival and/or patient morbidity/mortality postoperatively. However, the large number of confounding factors and outcomes renders existing systems relatively insensitive. Scoring systems therefore concentrate on specific confounders, with the aim of predicting a specific primary outcome. In liver transplantation, the MELD (model for end-stage liver disease) is used to score the severity of disease in the recipient⁷ whereas the Eurotransplant liver donor risk index accounts for donor factors.⁸

In pancreas transplantation (PT), despite the significant long-term multisystem benefits of euglycaemia^9–11 and (in over 90% of cases) simultaneous independence from renal replacement therapy, immediate risks following transplantation remain significant and unpredictable.¹² They include vascular complications,¹³ graft pancreatitis,¹⁴ enteric complications (leaks, fistulas and intra-abdominal collections)¹⁵ and infective complications, which may occur in up to 75% of recipients.¹⁶ Recipients can remain in a morbid state following PT for prolonged periods. For this reason, risk prediction scores have been developed (P-PASS [preprocurement pancreas allocation suitability score]¹⁷ and PDRI [pancreas donor risk index])¹⁸ but these are limited to donor factors and specific to predicting graft outcome. They are unable to inform the clinician or patient about an individual’s postoperative morbidity and recovery time.

Retrospective studies investigating recipient factors have consistently found that high body mass index (BMI) and increasing age predict poorer outcome following PT, with higher rates of morbidity in these groups.^19–22 No cumulative score is validated that considers a multisystem approach in assessing recipient morbidity.

The Waterlow score was introduced in the 1980s as a nursing tool to stratify risk for patients with respect to developing decubitus ulcers (pressure sores).²³ It uses a multisystem approach and scores patients based on a number of variables (Table 1) before categorising patients into ‘at risk’, ‘high risk’ or ‘very high risk’ for developing pressure ulcers. It is used routinely by medical staff in UK hospitals for this purpose. Recently, evidence has emerged that a high preoperative Waterlow score (>20) correlates with increased inpatient mortality and 30-day morbidity in a general surgical setting.²⁴ As Waterlow score incorporates a multisystem approach in risk assessment and PT recipients suffer with a multisystem disease, it seems logical that the scoring system may have sensitivity in risk prediction following PT.

Table 1.

Contents of Waterlow pressure ulcer scoring system with potential scores for each category

Scoring category	Range of potential scores
Body mass index	0–3
Skin integrity	0–3
Sex	1–2
Age	1–5
Nutritional status	0–4
Continence	0–3
Mobility	0–5
Special risks and significant morbidity	0–24
Neurological deficit	0–18
Type and length of surgery	0–13
Medications	0–4

The aim of the study was therefore to prospectively assess the utility of commonly used general surgical risk prediction models (P-POSSUM [Portsmouth POSSUM], MODS, Charlson index, revised cardiac risk index, ASA grade and Waterlow score), and to correlate these with total length of hospital stay (LOS) and critical care unit (CCU) LOS following simultaneous pancreas and kidney (SPK) transplantation.

Methods

The study was registered with and received approval from the research and development division of the Central Manchester University Hospitals NHS Foundation Trust. All SPKT recipients in a single centre over a two-year period (November 2011 – November 2013) were included for initial analysis. Early graft loss (within 72 hours) was excluded as the recovery rate of these patients differs from that of SPK transplant recipients. Patient mortality was excluded owing to the risk of potential confounders influencing analysis with small cohorts and a lack of data regarding hospital discharge.

All risk scores (P-POSSUM, MODS, Charlson index, revised cardiac risk index, ASA grade and Waterlow score) were calculated at the time of admission to hospital, prior to undergoing SPK transplantation. The total number of recipient diabetic complications (history of cardiovascular disease, peripheral vascular disease, previous cerebrovascular incident, retinopathy, autonomic neuropathy, peripheral neuropathy, diabetic nephropathy and hypertension) were also noted and correlated with the outcome measures.

For those risk assessment scores requiring physiological data, the values recorded on the ward at admission were used for the calculation of that score. Donor data were collected using the Electronic Offering System online forms. Organ recipient and outcome data were collected prospectively. P-POSSUM and Charlson scores were obtained using online calculators (http://www.riskprediction.org.uk/pp-index.php and http://www.biomedcentral.com/content/supplementary/1471–2407–4-94-S1.xls respectively). All scores were correlated with a patient’s total LOS and CCU LOS. LOS was calculated from date of admission to readiness to discharge from either hospital or the CCU.

Traditionally, the Waterlow score stratifies patients into ‘at risk’ (>9), ‘high risk’ (>14) or ‘very high risk’ (>19) of developing decubitus ulcers. However, in the case of SPK transplant recipients, an individual’s absolute Waterlow score has been correlated with total LOS and CCU LOS. Potentially confounding factors (recipient age, BMI, dialysis status and time on transplant register; length of surgery; pancreas and kidney cold ischaemic times [CITs]; and donor age, BMI and donor after brainstem death or donor after cardiac death status) were also correlated with the outcome measures to assess their effect on LOS.

Transplant protocol

The criteria used by individual transplant units for acceptance on to the waiting list for pancreatic transplantation are governed by national guidelines produced by NHS Blood and Transplant. Patients were allocated organs from the waiting list based on their blood group, human leucocyte antigen matching and waiting time. Pancreas implantation was undertaken as described previously.²²

Statistical analysis

Statistical analysis was carried out with SPSS^® version 20 (IBM, New York, US), using Pearson’s correlation coefficient and Spearman’s rank correlation coefficient to compare continuous data (normal and skewed distribution respectively), and Kruskal–Wallis one-way analysis of variance to compare categorical data. A p-value of <0.05 was deemed statistically significant.

Results

Recipient demographics

All SPK transplant recipients between November 2011 and October 2013 (n=57) were included in the study. (All were primary transplants.) There were five exclusions: four recipients had on-table pancreatectomy because of on-table thrombosis (n=3) and uncontrollable haemorrhage (n=1), and one patient died prior to discharge owing to sepsis with a preceding transplant pancreatectomy. Of the remaining 52 recipients, the mean age was 42.00 years (standard deviation [SD]: 7.60 years), there were 27 male recipients (52.0%) and the mean BMI was 25.43kg/m² (SD: 3.11kg/m²). The mean duration of diabetes was 26.69 years (SD: 7.43 years) and the mean time on the transplant register was 22.28 months (SD: 13.03 months). Fifty recipients (96.2%) were white British, one was African and one was Indian in origin; 15 patients (28.8%) had not had dialysis, 17 (32.7%) were on peritoneal dialysis at the time of SPK transplantation and 20 (38.5%) were on haemodialysis at the time of transplantation.

Donor demographics

Of the donors, 34 (65.4%) were male and 41 (78.8%) were donors after brainstem death with a mean age of 34.02 years (SD: 12.52 years) and a mean BMI of 24.21kg/m² (SD: 3.17kg/m²). Table 1 shows the correlation of donor factors with the outcome measures.

Operative demographics

The mean length of procedure was 350 minutes (SD: 72.51 minutes), with mean pancreas and kidney cold ischaemic times (CIT) of 703 minutes (SD: 182.16 minutes) and 849 minutes (SD: 192.03 minutes) respectively. Table 2 shows the correlation of operative factors with outcome measures.

Table 2.

Correlation of donor, recipient and operative confounding factors with total LOS and CCU LOS in simultaneous pancreas and kidney transplantation

Confounding factor	Mean (SD)	p-value for total LOS	p-value for CCU LOS
Length of procedure (mins)	349.5 (72.5)	0.712*	0.424**
Pancreas CIT (mins)	703.0 (182.2)	0.293*	0.248**
Kidney CIT (mins)	849.9 (192.0)	0.357*	0.266**
Recipient BMI (kg/m2)	25.43 (3.11)	0.002*	0.044**
Recipient age (yrs)	42.0 (7.6)	0.670*	0.633**
Duration of diabetes (yrs)	26.7 (7.4)	0.428*	0.045**
Time on transplant register (mths)	22.3 (13.0)	0.712*	0.222**
Dialysis mode	No dialysis: 15 (28.8%) HD: 20 (38.5%) PD: 17 (32.7%)	0.289***	0.964***
Donor age (yrs)	34.0 (12.5)	0.870*	0.066**
Donor BMI (kg/m2)	24.21 (3.17)	0.085*	0.426**
Donor stauts	DBD: 41 (78.8%) DCD: 11 (21.2%)	0.209***	0.705***

Clinical outcomes

The median total LOS was 17 days (range: 8–79 days) and the mean CCU LOS was 7 days (SD: 4.04 days).

Risk assessment scores

In the entire cohort, the ASA grade was 3. While all the risk assessment scores identified and correlated with some accuracy to clinical outcome, the Waterlow score was the only significant predictor of total LOS and CCU LOS (Spearman’s r=0.472, p<0.001, and Pearson’s r=0.469, p=0.001, respectively) (Fig 1). The results of all the risk assessment scores and their correlation with total LOS and CCU LOS are presented in Table 3. These findings are not adjusted for any of the potential confounders. With the exception of recipient BMI, however, these were not found to be associated significantly with total LOS or CCU LOS (Table 2).

Figure 1.

Correlation of Waterlow score with total length of stay and critical care unit length of stay

Table 3.

Correlation of the studied risk assessment scores with total LOS and CCU LOS in simultaneous pancreas and kidney transplantation

Risk score (range of values)	Mean	p-value for total LOS	p-value for CCU LOS
Waterlow score (2–82)	13.87 (SD: 4.03)	<0.001*	0.001**
P-POSSUM morbidity (5.468–100%)	75.090% (SD: 11.752pp)	0.313*	0.359**
P-POSSUM mortality (0.223–100%)	9.807% (SD: 6.382pp)	0.194*	0.546**
MODS (0–24)	4† (range: 0–7)	0.864***	0.585***
Charlson co-morbidity index (2–99)	27.00 (SD: 24.76)	0.129*	0.074**
Revised cardiac risk index (0–3)	3† (range: 3–4)	0.816***	0.061***
Number of diabetic complications (0–8)	3† (range: 1–7)	0.123***	0.143***

Discussion

Risk prediction scoring systems aim to identify and accurately inform outcome for high risk patients. This information can ideally be used preoperatively in the assessment, counselling and consent processes as well as in the perioperative management to individualise a patient’s care and help reduce postoperative morbidity.

Increased total LOS and CCU LOS following surgery are surrogate markers of inpatient morbidity, and increased perioperative morbidity leads to poorer long-term outcomes.²⁵ In general surgery, a number of scoring systems exist to help predict outcome of patients undergoing interventions.^2–6

Outcome prediction following PT remains an enigma owing to the multifactorial sequela of diabetes and the nature of transplantation. Current PT scoring systems (P-PASS, PDRI) specifically analyse donor and organ factors that influence organ outcome (30-day and 1-year survival), which are surrogate measures of success. Despite this, they do not necessarily correlate accurately with patient morbidity after transplantation. To date, studies have not investigated factors in PT affecting recipient total LOS and CCU LOS or long-term survival.²⁵ In PT, studies have previously identified high recipient BMI as a predictor of poor outcome and the results of our study are consistent with these previous findings (p=0.002 and 0.044 respectively for recipient total LOS and CCU LOS).

Our study also found that the diabetic load (number of years of diabetic duration preceding transplant) correlates significantly with CCU LOS (p=0.045). However, as no adjustment was made for multiple testing, this result should be interpreted with caution. This finding may indicate lower physiological reserve in patients with higher diabetic load and exposure.

Waterlow score is strongly predictive of PT recipients’ total LOS and CCU LOS. This is in contrast to the number of diabetic complications and other commonly used general surgery risk prediction scoring systems (P-POSSUM, MODS, Charlson index, cardiac risk index and ASA grade), which have all failed to correlate with any outcome measures. These findings are independent of donor and organ factors, and are likely to be due to the benefit of the use of the multisystem approach of the Waterlow score.

The Waterlow score is used commonly by nursing staff in their preoperative assessment of surgical patients but can be easily transferred to an ambulatory or assessment environment to aid in stratification prior to transplant listing and in informing risk. It provides an objective cumulative risk score based on a combination of co-morbidities, mobilisation, nutritional factors and demographics.

In comparison, all the other scoring systems assessed in this study fail to address one or more of those variables. ASA grade is entirely subjective, and the cardiac risk index and MODS provide a very narrow range of possible scores in PT recipients, thereby providing little differentiation between individuals. Although the cardiac risk index is not normally used as an aid to predict postoperative outcomes, it was felt that in a group of patients with such extensive cardiovascular disease, it potentially had utility as a risk scoring system to assess this cohort.

P-POSSUM was used ahead of other POSSUM alternatives as it was felt that SPK transplantation is best compared with major general surgical procedures and an account of the extent of the procedure should be taken when estimating the risk. P-POSSUM and the Charlson index both take into account a patient’s co-morbid/physiological status and provide a wide breath of scores to help differentiate level of risk but they fail to account for factors such as nutritional status, which are important in risk assessment of diabetic and end-stage renal failure patients^26–28 as well as in outcome analysis following major surgery.^29,30

In addition (and most importantly), the Waterlow score identifies the highest risk recipients, allowing for individual optimisation of postoperative care with an attempt to reduce postoperative morbidity, and thereby total LOS and CCU LOS.

The low mortality associated with the procedure and low rates of graft loss unfortunately preclude the analysis of these factors as outcome measures in this cohort. Larger studies would need to be carried out to include these outcomes and also to correct for the numerous variables involved with PT.

Conclusions

This prospective study comparing the predictive capability of a number of commonly used general surgical risk scoring systems in SPK transplantation does indicate that the Waterlow score may have value in the prediction of outcome following the procedure. Further validation with larger cohorts is required but this would allow for more accurate stratification of patients and for multivariate analysis to be conducted, thereby aiding both clinicians and prospective transplant recipients in assessing risk.