Advertisement

Prognostic Risk Modelling for Patients Undergoing Major Lower Limb Amputation: An Analysis of the UK National Vascular Registry

Open ArchivePublished:December 26, 2019DOI:https://doi.org/10.1016/j.ejvs.2019.12.006

      Objective

      Major lower limb amputation is the highest risk lower limb procedure in vascular surgery. Despite this, few high quality studies have examined factors contributing to mortality. The aim was to identify independent risk factors for peri-operative morbidity and mortality and develop reliable models for estimating risk.

      Methods

      All patients undergoing lower limb amputation above the ankle entered into the UK National Vascular Registry (January 2014–December 2016) were included. Missing data were handled using multiple imputation. Models were developed to evaluate independent risk factors for mortality (the primary outcome) and morbidity using logistic regression, minimising the Bayesian information criterion to balance complexity and model fit. Ethical approval for the study was granted (Wales REC 3 ref:16/WA/0353).

      Results

      All 9549 above ankle joint amputations in the registry were included. Overall, 865 patients (9.1%) died before leaving hospital. Independent factors associated with mortality were emergency admission, bilateral operation, age, American Society of Anesthesiologists' grade, abnormal electrocardiogram, and increased white cell count or creatinine (p < .01 for all). Independent factors reducing mortality were transtibial operation, increased albumin or patient weight, and previous ipsilateral revascularisation procedures (p < .01 for all). A risk model incorporating these factors had good discrimination (C-statistic 0.79, 95% confidence interval 0.77–0.80) and excellent calibration. Morbidity rates were high, with 6.6%, 9.7%, and 4.3% of patients suffering cardiac, respiratory, and renal complications, respectively. The risk model was also predictive of morbidity outcomes (C-statistics 0.74, 0.69, and 0.74, respectively).

      Conclusion

      Morbidity and mortality after lower limb amputation are high in the UK. Some potentially modifiable factors for quality improvement initiatives have been identified and accurate predictive models that could assist patient counselling and decision making have been developed.

      Keywords

      This observational study used data from the UK National Vascular Registry to look at risk factors for in hospital mortality and morbidity following major lower limb amputation. Eleven independent risk factors were identified, including emergency admission, raised white cell count or creatinine, low albumin or patient weight, age, American Society of Anesthesiologists' grade, amputation level, and prior intervention to the amputated limb. A risk model was developed that had excellent calibration and good discrimination (C-statistic 0.79, 95% confidence interval 0.77–0.80). An online calculator for the model is available, making it easy to use.

      Introduction

      Major lower limb amputation is one of the highest risk surgical procedures in the UK, with in hospital mortality rates of approximately 6% for below knee and 12% for above knee amputation.
      • Waton S.
      • Johal A.
      • Heikkela K.
      • Cromwell D.
      • Loftus I.
      National vascular registry annual report.
      While this is largely related to patient fitness rather than operative complexity, it is nonetheless important to identify prognostic factors contributing to mortality rates so that quality improvement programmes can be implemented. Although limited work exists for UK populations, the only large studies come from parts of the world with radically different healthcare systems, calling generalisability into question (Table S1; see Supplementary Material).
      • Tang T.Y.
      • Prytherch D.R.
      • Walsh S.R.
      • Athanassoglou V.
      • Seppi V.
      • Sadat U.
      • et al.
      The development of a VBHOM-based outcome model for lower limb amputation performed for critical ischaemia.
      • Patterson A.J.
      • Degnan A.J.
      • Walsh S.R.
      • Eltayeb M.
      • Scout E.F.
      • Clarke J.M.
      • et al.
      Efficacy of VBHOM to predict outcome following major lower limb amputation.
      • Pell J.
      • Stonebridge P.
      Association between age and survival following major amputation. The Scottish Vascular Audit Group.
      • Easterlin M.C.
      • Chang D.C.
      • Wilson S.E.
      A practical index to predict 30-day mortality after major amputation.
      • Yamada K.
      • Yasunaga H.
      • Kadono Y.
      • Chikuda H.
      • Ogata T.
      • Horiguchi H.
      • et al.
      Postoperative outcomes of major lower extremity amputations in patients with diabetes and peripheral artery disease: analysis using the Diagnosis Procedure Combination database in Japan.
      • Feinglass J.
      • Pearce W.H.
      • Martin G.J.
      • Gibbs J.
      • Cowper D.
      • Sorensen M.
      • et al.
      Postoperative and late survival outcomes after major amputation: findings from the department of Veterans Affairs national surgical quality improvement program.
      Mortality is not the only negative outcome experienced by this cohort; they face long hospital stays, a high rate of peri-operative complications, and frequent re-admissions.
      • National Confidential Enquiry Into Patient Outcome and Death
      Lower limb amputation: working together.
      Prognostic risk modelling into leading causes of morbidity is therefore also important. Individualising consent
      • Fryar C.
      Clarifying the montgomery judgment.
      and risk adjustment of surgeon specific outcome data
      • Walker K.
      • Neuburger J.
      • Groene O.
      • Cromwell D.A.
      • van der Meulen J.
      Public reporting of surgeon outcomes: low numbers of procedures lead to false complacency.
      are also enabled by robust risk models. There is some evidence that the broad adoption of the EUROSCORE
      • Nashef S.A.
      • Roques F.
      • Michel P.
      • Gauducheau E.
      • Lemeshow S.
      • Salamon R.
      European system for cardiac operative risk evaluation (EuroSCORE).
      risk prediction tool in cardiac surgery helped lead to the dramatic improvement in cardiac surgical outcomes that occurred around the turn of the millennium.
      • Nashef S.A.
      • Roques F.
      • Sharples L.D.
      • Nilsson J.
      • Smith C.
      • Goldstone A.R.
      • et al.
      EuroSCORE II.
      It is possible that by facilitating appropriate targeting of resources to higher risk patients it may be possible to replicate these results in other fields. Thus, development of accurate risk models for patients undergoing major lower limb amputation may have a multitude of benefits.
      The objectives of this study were therefore to identify the independent risk factors for peri-operative mortality and leading causes of morbidity in UK patients and develop robust prognostic models. The ability of these models to accurately predict mortality in a contemporary UK data set were then compared with previously published models.

      Methods

      Data

      All patients recorded in the UK National Vascular Registry (NVR) as undergoing major lower limb amputation (below knee, through knee, above knee, hip disarticulation, and hind quarter amputation) from 1 January 2014 until 31 December 2016 were included in the study. Data were formally requested through and approved by the UK Healthcare Quality Improvement Partnership, who are the data controllers for English and Welsh data within the NVR, and through the Audit and Quality Improvement Committee of the Vascular Society of Great Britain and Ireland, who are the data controllers for Scottish and Northern Irish data within the NVR. Data were retrieved in March 2018 to allow time for completion of the index admission, as well as site data entry.
      A list of the variables applied for and their type is given in Table S2 (see Supplementary Material).

      Outcomes

      The primary outcome was in hospital mortality. Secondary outcomes were return to theatre during admission, re-admission to a higher level of care, post-operative length of stay, and post-operative complications (subdivided into several different categories: cardiac, respiratory, cerebral [stroke], renal failure, haemorrhage, and limb ischaemia).

      Ethical approval and study registration

      The study was approved by Wales Research Ethics Committee 3 (reference number 16/WA/0353), and the protocol was registered in the Australia and New Zealand Clinical Trial Registry as ACTRN12618000356268. This report has been prepared in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) and Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD) statements.
      • von Elm E.
      • Altman D.G.
      • Egger M.
      • Pocock S.J.
      • Gøtzsche P.C.
      • Vandenbroucke J.P.
      The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.
      ,
      • Collins G.S.
      • Reitsma J.B.
      • Alstman D.G.
      • Moons K.G.M.
      Transparent reporting of a multivariable prediction model for individual Prognosis or diagnosis (TRIPOD): the TRIPOD statement.

      Statistical methodology

      All statistical analysis was performed in the R statistical programming environment, version 3.5.1.
      R Core Team
      R: a language and environment for statistical computing.
      Multiple imputation using the mice package version 3.3.0 was used to handle missing data,
      • van Buuren S.
      • Groothuis-Oudshoorn K.
      Mice: multivariate imputation by chained equations in R.
      excluding parameters with >50% of values missing from multivariable modelling. Data were imputed using 45 replicates with 45 iterations of the chained equations algorithm for each replicate. In order to explore the differences there might be between analysis based on the imputed data and the unimputed data, a sensitivity analysis was done by performing univariable analysis using both complete case analysis compared with the multiply imputed data.
      Univariable analysis was performed using univariable logistic regression, together with application of Rubin's rules to pool estimates for multiple imputation.
      • Little R.J.
      • Rubin D.B.
      Statistical analysis with missing data.
      Continuous variables were kept as such and odds ratios (ORs) are given per unit change in value rather than dichotomised into “high” and “low” values. Multivariable analysis was performed using multivariable logistic regression analysis to develop models using pre-operative predictors. Parameters were selected for inclusion in prognostic models using stepwise selection and following an information criterion based analysis, by minimising the Schwarz–Bayes Criterion.
      • Schwarz G.
      Estimating the dimension of a model.
      This was done separately for each of the 45 replicates and terms that were present in at least half of the replicates were retained.
      The NVR was deliberately designed as a fairly minimal reporting set in response to high rates of missing data in the previous national registry, which included a large number of parameters, so that all recorded predictors were felt to be potentially important for outcomes. Therefore, all the predictors listed in Table S2 that had missing data levels <50% for inclusion in the prognostic models were evaluated for inclusion in the prognostic models. Receiver operating curve (ROC) analysis was used to assess model discrimination using the pROC package, version 1.12.1.
      • Robin X.
      • Turck N.
      • Hainard A.
      • Tiberti N.
      • Lisacek F.
      • Sanchez J.-C.
      • et al.
      pROC: an open-source package for R and S+ to analyze and compare ROC curves.
      The Delong method was then used to calculate confidence intervals (CIs) for the area under the ROC curve (C-statistic) and test whether performance was different to the estimated C-statistics of existing models.
      • DeLong E.R.
      • DeLong D.M.
      • Clarke-Pearson D.L.
      Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach.
      Comparison with existing models is hampered by the fact that three of the four models found in a literature search include terms that are not recorded in the NVR, so any estimation of the discriminatory power of these models is hampered by the fact that these parameters can only be set to default values.
      • Tang T.Y.
      • Prytherch D.R.
      • Walsh S.R.
      • Athanassoglou V.
      • Seppi V.
      • Sadat U.
      • et al.
      The development of a VBHOM-based outcome model for lower limb amputation performed for critical ischaemia.
      ,
      • Easterlin M.C.
      • Chang D.C.
      • Wilson S.E.
      A practical index to predict 30-day mortality after major amputation.
      ,
      • Feinglass J.
      • Pearce W.H.
      • Martin G.J.
      • Gibbs J.
      • Cowper D.
      • Sorensen M.
      • et al.
      Postoperative and late survival outcomes after major amputation: findings from the department of Veterans Affairs national surgical quality improvement program.
      The revised Vascular Biochemistry and Haematology Outcome Model does not suffer from this problem, so comparison with this model can be viewed as “fair”.
      • Patterson A.J.
      • Degnan A.J.
      • Walsh S.R.
      • Eltayeb M.
      • Scout E.F.
      • Clarke J.M.
      • et al.
      Efficacy of VBHOM to predict outcome following major lower limb amputation.
      The Hosmer–Lemeshow goodness of fit test was used to assess calibration of the models.
      • Hosmer D.W.
      • Lemeshow S.
      • Sturdivant R.X.
      Applied logistic regression.

      Results

      Demographics and outcomes

      There were 12 593 amputations entered into the registry during the study period, of which 9 549 were above the ankle and so comprised the study population. Of these, 4 516 (47.3%) were transtibial, 4 369 (45.8%) transfemoral, 442 (4.6%) through knee, 32 (0.3%) hip disarticulation, and 190 (2%) were simultaneous bilateral procedures. Table 1 summarises the baseline characteristics of the study population, together with the number of missing data for each parameter.
      Table 1Baseline characteristics of patients undergoing major lower limb amputation in the UK National Vascular Registry (NVR)
      ParameterPatients (n = 9 549)Missing valuesn (%)
      Age – years70 (60–78)5 (0)
      Sex
       Male6 729 (70.5)
       Female2 820 (29.5)
      Hospital type
       Teaching4 544 (47.6)
       Non-teaching5 005 (52.4)
      Emergency admission7 489 (78.4)
      Comorbidities17 (0)
       Diabetes5 065 (53.0)
       Ischaemic heart disease3 788 (39.7)
       Congestive heart failure1 004 (10.5)
       Chronic lung disease1 939 (20.3)
       Chronic kidney disease1 968 (20.6)
       Hypertension5 812 (60.9)
       Stroke1 085 (11.4)
      Smoking30 (0)
       Never1 948
       Ex-smoker4 721
       Current2 850
      Pre-operative blood tests
       White cell count – 109 cells/L11.7 (9.0–15.4)14 (0)
       Haemoglobin – g/L112 (97–148)3416 (36)
       Sodium – mmol/L136 (133–139)38 (0)
       Potassium – mmol/L4.5 (4.1–4.9)18 (0)
       Creatinine – μmol/L81 (61–118)11 (0)
       Albumin – g/L30 (24–35)2824 (30)
      Abnormal ECG3 672 (42.9)988 (10)
      ASA grade2 (0)
       190
       2756
       36 164
       42 462
       575
      Indication
      Amputations for malignancy are not recorded in the NVR.
      16 (0)
       Acute limb ischaemia1 563 (16.4)
       Chronic limb ischaemia1 981 (20.8)
       Neuropathy140 (1.5)
       Tissue loss3 567 (37.4)
       Uncontrolled infection2 129 (22.3)
       Trauma94 (1.0)
       Aneurysm59 (0.6)
      Pre-operative medications3 (0)
       Antiplatelet agent6 783 (71.1)
       Statin6 701 (70.2)
       Beta blocker2 560 (26.8)
       ACE inhibitor/ARB3 035 (32.0)
      Weight – kilograms75 (63–87)2 607 (27)
      Data are presented as median (interquartile range) for continuous variables and n (%) for categorical variables, apart from smoking status, and American Society of Anesthesiologists' (ASA) grade. Abnormal ECG is a field in the NVR with only two options – normal or abnormal. ECG = electrocardiogram; ACE = angiotensin converting enzyme; ARB = angiotensin II receptor blocker.
      Amputations for malignancy are not recorded in the NVR.
      Overall, 865 patients (9.1%) died before leaving hospital. The mortality rate for below knee amputations (5.8%) was lower than for above knee procedures (12.0%; p < .001). Data were not available on cause of death or re-amputation rates. There was also a high rate of post-operative morbidity in the cohort, with 6.6%, 9.7%, and 4.3% of patients suffering cardiac, respiratory, and renal complications, respectively. Less than 1% of patients were recorded as having a post-operative stroke or bleeding complication, and 4.4% had a complication relating to limb ischaemia. Ten per cent (n = 966) of patients had an unplanned return to theatre, while 3.8% (n = 363) were re-admitted to critical care. The median post-operative length of stay was 16 days (interquartile range [IQR] 9–28 days), with an overall median length of stay of 24 days (IQR 14–42 days).

      Post-operative mortality risk factors

      Univariable analysis revealed that increased patient age; a history of ischaemic heart disease, congestive heart failure, chronic lung disease, chronic kidney disease, or stroke; a raised white cell count, raised serum creatinine, or low serum albumin; an abnormal electrocardiogram (ECG); increased American Society of Anesthesiologists' (ASA) grade; emergency admission and pre-operative beta blocker therapy all increased the odds of in hospital mortality. Male sex, previous intervention on the same side, below knee amputation, current smoking, statin or angiotensin converting enzyme inhibitor/angiotensin II receptor blocker therapy, and increased weight all had protective effects (Table 2).
      Table 2Univariable analysis showing odds ratios (OR) of being discharged alive according to different risk factors for patients undergoing major lower limb amputation in the UK National Vascular Registry
      ParameterMultiple imputationComplete case analysis
      OR95% CIp valueOR95% CI
      Age – per 10 year ↑0.7630.720–0.809<.0010.7640.721–0.810
      Sex – male vs. female1.2411.070–1.439.0041.2411.070–1.439
      Hospital type – teaching vs. non-teaching1.0340.899–1.190.641.0340.899–1.190
      Emergency admission0.2630.203–0.342<.0010.2630.203–0.342
      Previous intervention on same side1.6171.406–1.861<.0011.6181.406–1.861
      Below knee amputation vs. higher level2.2161.907–2.575<.0012.2161.907–2.575
      Comorbidities (Yes vs. No)
       Diabetes1.0690.930–1.230.351.0690.930–1.230
       Ischaemic heart disease0.6340.551–0.730<.0010.6350.552–0.731
       Congestive heart failure0.4780.397–0.576<.0010.4790.398–0.577
       Chronic lung disease0.6900.588–0.810<.0010.6900.588–0.811
       Chronic kidney disease0.4770.411–0.555<.0010.4770.411–0.556
       Hypertension0.8670.750–1.003.0540.8680.750–1.003
       Stroke0.7850.640–0.963.0210.7850.640–0.963
      Smoking – current1.2081.031–1.415.0191.2061.030–1.413
      Pre-operative blood tests
       White cell count – per 109 cells/L ↑0.9680.961–0.975<.0010.9680.961–0.975
       Haemoglobin – per g/L ↑1.0050.996–1.014.311.0050.996–1.014
       Sodium – per mmol/L ↑0.9950.980–1.010.510.9950.980–1.010
       Potassium – per mmol/L ↑0.9930.924–1.069.860.9940.924–1.069
       Creatinine – per 10 μmol/L ↑0.9730.968–0.977<.0010.9730.968–0.977
       Albumin – per g/L ↑1.0611.051–1.072<.0011.0621.051–1.072
      Abnormal ECG0.4110.353–0.478<00010.4000.343–0.467
      ASA grade – per grade ↑0.2480.219–0.282<00010.2480.219–0.282
      Pre-operative medications
       Antiplatelet agent1.1130.957–1.295.171.1130.957–1.295
       Statin1.2581.086–1.459.0021.2591.086–1.459
       Beta blocker0.7180.619–0.834<.0010.7190.619–0.834
       ACE inhibitor/ARB1.1691.002–1.363.051.1691.002–1.364
      Weight – per 10 kg ↑1.0851.042–1.129<00011.0881.043–1.134
      The last two columns present a sensitivity analysis using complete cases only. The ↑ symbol is used to indicate an increase in value, for example “per 10 year ↑” indicates that the ORs are those associated with a 10 year increase in age. Numbers >1 indicate greater odds of being discharged alive. CI = confidence interval; Y = yes; N = no; ECG = electrocardiogram; ASA = American Society of Anesthesiologists; ACE = angiotensin converting enzyme; ARB = angiotensin II receptor blocker.
      Analysis was repeated using complete case analysis to assess sensitivity to the imputation methodology. Results were almost identical to the multiple imputation analysis, giving confidence that the imputation methodology had not introduced significant bias (Table 2).
      Multivariable regression modelling revealed that independent factors associated with worse in hospital mortality were emergency admission (OR 2.47, 95% CI 1.89–3.24), bilateral operation (OR 2.19, 95% CI 1.48–3.25), age (OR per 10 year increase 1.21, 95% CI 1.13–1.29), ASA grade (OR per unit increase 2.60, 95% CI 2.27–2.98), abnormal ECG (OR 1.52, 95% CI 1.28–1.79), and increased white blood cell count (OR per 109 cells/L increase 1.02, 95% CI 1.01–1.03) or serum creatinine (OR per 10 μmol/L increase 1.02, 95% CI 1.02–1.03).
      Independent protective factors reducing in hospital mortality were transtibial operation (OR 0.61, 95% CI 0.52–0.72), increased serum albumin (OR per g/L increase 0.97, 95% CI 0.95–0.98), previous procedures to the amputated limb (OR 0.79, 95% CI 0.68–0.92), and increased patient weight (OR per 10 kg increase 0.95, 95% CI 0.91–0.99).

      Development of a prognostic model for post-operative mortality

      A multivariable logistic regression model using the factors identified above to predict the chances of surviving to hospital discharge was constructed. Hosmer–Lemeshow goodness of fit analysis revealed good model fit (p = .348 for evidence of miscalibration). A calibration table is given in Table S3 (see Supplementary Material), along with details of the model formula. This is displayed graphically in Figure S1 (see Supplementary Material).
      ROC curve analysis showed that the model (labelled “UKAmpRisk”) has good (bordering on excellent) discrimination (C-statistic 0.79, 95% CI 0.77–0.80). A plot of the ROC curve is shown in Fig. 1.
      Figure 1
      Figure 1Receiver operating curve for UKAmpRisk, the prognostic model developed here, for patients undergoing major lower limb amputation in the UK National Vascular Registry (NVR), with best estimates of the Veterans Affairs Model (VAM),
      • Feinglass J.
      • Pearce W.H.
      • Martin G.J.
      • Gibbs J.
      • Cowper D.
      • Sorensen M.
      • et al.
      Postoperative and late survival outcomes after major amputation: findings from the department of Veterans Affairs national surgical quality improvement program.
      Vascular Biochemistry and Haematology Outcomes Model (VBHOM),
      • Tang T.Y.
      • Prytherch D.R.
      • Walsh S.R.
      • Athanassoglou V.
      • Seppi V.
      • Sadat U.
      • et al.
      The development of a VBHOM-based outcome model for lower limb amputation performed for critical ischaemia.
      ,
      • Patterson A.J.
      • Degnan A.J.
      • Walsh S.R.
      • Eltayeb M.
      • Scout E.F.
      • Clarke J.M.
      • et al.
      Efficacy of VBHOM to predict outcome following major lower limb amputation.
      and National Surgical Quality Improvement Programme (NSQIP) models.
      • Easterlin M.C.
      • Chang D.C.
      • Wilson S.E.
      A practical index to predict 30-day mortality after major amputation.
      Some predictive variables for VAM, VBHOM, and NSQIP models are not available in the NVR.

      Comparison to existing models

      The C-statistic was 0.59 (95% CI 0.56–0.61) for the Vascular Biochemistry and Haematology Outcomes Model (VBHOM),
      • Tang T.Y.
      • Prytherch D.R.
      • Walsh S.R.
      • Athanassoglou V.
      • Seppi V.
      • Sadat U.
      • et al.
      The development of a VBHOM-based outcome model for lower limb amputation performed for critical ischaemia.
      0.65 (95% CI 0.63–0.67) for the revised VBHOM model (VBHOM2),
      • Patterson A.J.
      • Degnan A.J.
      • Walsh S.R.
      • Eltayeb M.
      • Scout E.F.
      • Clarke J.M.
      • et al.
      Efficacy of VBHOM to predict outcome following major lower limb amputation.
      0.68 (95% CI 0.66–0.70) for the Veterans Affairs Model (VAM),
      • Feinglass J.
      • Pearce W.H.
      • Martin G.J.
      • Gibbs J.
      • Cowper D.
      • Sorensen M.
      • et al.
      Postoperative and late survival outcomes after major amputation: findings from the department of Veterans Affairs national surgical quality improvement program.
      and 0.65 (95% CI 0.64–0.68) for the National Surgical Quality Improvement Programme (NSQIP) model.
      • Easterlin M.C.
      • Chang D.C.
      • Wilson S.E.
      A practical index to predict 30-day mortality after major amputation.
      All four models showed inferior discrimination compared with the present model (p < .001 for all comparisons). Fig. 1 shows all five ROC curves on the same graph for comparison. The NSQIP, VBHOM, and VBHOM2 models all failed the Hosmer–Lemeshow goodness of fit test (p < .001 in each case), suggesting that they are also poorly calibrated for this patient cohort. The intercept coefficient was not published for the VAM model, so it was not possible to assess the calibration for that model.

      Risk factors for secondary outcomes

      Multivariable regression modelling revealed that low serum albumin and high ASA grade were consistent predictors of most morbidity outcomes. Other predictors frequently associated with outcome were amputations done as an emergency and a raised serum creatinine level. Full details of the parameters which were independently associated with the secondary outcomes are given in Table S4 (see Supplementary Material). The ability of models based on these factors to discriminate between patients who did or did not suffer these morbidity outcomes was again assessed using the C-statistic (Table S4).
      The predictive model for in hospital mortality was again a good predictor of several of the morbidity outcomes, including cardiac, respiratory, and renal complications (C-statistics 0.74, 0.69, and 0.74, respectively).

      Discussion

      Eleven factors have been identified that are independently associated with in hospital mortality in patients undergoing major lower limb amputation. While some of these factors (principally age and ASA grade) are not modifiable, the majority are potentially amenable to modification through improved clinical care.
      Many, such as emergency admission and a raised white cell count, are linked to management of patients at a late stage in their disease and may reflect late presentation or recognition. This highlights the critical role of healthcare staff in recognising the deteriorating foot in the community, and robust in hospital systems and teams to treat patients quickly. Earlier recognition will reduce the number of patients undergoing amputation as an emergency when they are septic, with increased risk of both kidney and cardiac dysfunction, often following a period of chronic low grade foot sepsis resulting in malnutrition and low albumin. Amputation is often followed by long periods in hospital. In the authors' experience, much of this time results from social or organisational factors, including the need to assess a patient's home for wheelchair suitability and carry out any necessary modifications. Earlier recognition would allow amputation to be handled in a more elective manner, so that this could be done ahead of time, facilitating shorter hospital admissions and thus reduced healthcare costs. Such systems are already in place for many patients in the form of the diabetic foot service and could be rolled out to all patients with chronic limb threatening ischaemia. Evidence is already emerging that such “limb salvage” services can make an impressive impact by reducing amputation rates.
      • Flores A.M.
      • Mell M.W.
      • Dalman R.L.
      • Chandra V.
      Benefit of multidisciplinary wound care center on the volume and outcomes of a vascular surgery practice.
      However, the present work highlights the fact that limb salvage must not be the only measure of the success of “limb salvage” services. Early recognition that limb salvage is unlikely to succeed could facilitate early discussion about the options and outcomes of amputation – there is increasing recognition of this fact from those running limb salvage services.
      • Neville R.F.
      • Kayssi A.
      Development of a limb-preservation program.
      This, in turn, could improve the outcomes of those patients who decide to have an amputation rather than continuing to pursue efforts at limb salvage, which are ultimately likely to be unsuccessful.
      Some of these patients are frail and, with co-existing cognitive impairment, they may benefit little from amputation. Early recognition of the deteriorating limb will help to prompt conversations about whether amputation is something the patient would want or benefit from, allowing decisions about active or palliative management to be made based on accurate prognostication.
      A model has been developed that could be used to aid counselling and decision making, either in clinic or at the bedside, by quantifying the probability of the patient surviving to hospital discharge. A web calculator has been developed for easy use in the clinic (available from www.ambler.me.uk/Vascular) and could easily be converted into a standalone smartphone app for offline use. By having a model that can more reliably predict mortality, discussions about options can be more fully explored with patients, enhancing shared decision making.
      • Elwyn G.
      • Frosch D.
      • Thomson R.
      • Joseph-Williams N.
      • Lloyd A.
      • Kinnersley P.
      • et al.
      Shared decision making: a model for clinical practice.
      Multiple previous studies have shown that surgeons systematically underestimate the chances of a patient surviving an operation.
      • Timmermans D.
      • Kievit J.
      • van Bockel H.
      How do surgeons' probability estimates of operative mortality compare with a decision analytic model?.
      ,
      • Glasgow R.E.
      • Hawn M.T.
      • Hosokawa P.W.
      • Henderson W.G.
      • Min S.J.
      • Richman J.S.
      • et al.
      Comparison of prospective risk estimates for postoperative complications: human vs computer model.
      As the choice between amputation and conservative management is sometimes the choice between amputation and palliation, it is critically important that these discussions are conducted in the context of reliable risk estimates. One possible use of this calculator might be in highlighting those patients who are unlikely to have a good outcome from amputation. For example, an underweight 85 year old patient with sepsis who would need an above knee amputation would have a low chance of surviving amputation. Rather than intervening, and with a high chance of prolonged morbidity and ultimately death, it may be better to consider palliative management, having now identified them accurately with this prognostic model.
      Ten per cent of patients returned to theatre during their index admission. This is quite a high proportion, and it would have been useful to examine the reasons for these returns to theatre. Unfortunately, no detail is given in the data set so some of these patients will have had a minor debridement procedure, while some will have had a major revision of amputation level. This may be the reason why it was difficult to generate a model with good discriminatory power for this outcome (Table S4).
      There are some indications that the medical management of the patients in this study was not optimal, with only 71% and 70% of patients being treated with antiplatelet or statin therapy, respectively. The recent Global Vascular Guidelines on the management of chronic limb threatening ischaemia give a level 1A recommendation to both of these,
      • Conte M.S.
      • Bradbury A.W.
      • Kolh P.
      • White J.V.
      • Dick F.
      • Fitridge R.
      • et al.
      Global vascular guidelines on the management of chronic limb-threatening ischemia.
      so there is clearly room for improvement in this patient cohort, although these rates were far better than the rates of 39% and 11%, respectively, revealed by a recent prospective study of patients with peripheral arterial disease in the UK.
      • Saratzis A.
      • Jaspers N.E.M.
      • Gwilym B.
      • Thomas O.
      • Tsui A.
      • Lefroy R.
      • et al.
      Observational study of the medical management of patients with peripheral artery disease.
      Of note, although statin use was associated with reduced mortality on univariable analysis, use of antiplatelet agents had no significant effect on any measured outcome.
      Previous procedures to the amputated limb reduced mortality rates. While it is possible that having had previous procedures is simply a surrogate for “fitness” in some way, it may also be that intervention to facilitate healing at a transtibial rather than transfemoral level might have multiple benefits, both in terms of improved short term outcomes and also in terms of the improvement in long term functional outcomes. Supporting this hypothesis is the fact that 51% of patients with a previous procedure had a transtibial amputation, whereas only 43% of patients without a previous procedure had a transtibial amputation. While it is possible that some of the effect seen for transtibial amputation is due to unmeasured confounding, the association was strong, even when all measured confounders were taken into account in multivariable modelling (adjusted OR for in hospital mortality 0.61, 95% CI 0.52–0.72).
      Some of the effects seen in the univariable analysis highlight the importance of confounder correction using multivariable analysis or careful matching of patient groups. For example, examining the univariable analysis it appears that smoking is protective and beta blockers are strongly associated with an increased risk of death. These effects are almost certainly artefacts caused by confounding. Indeed, the mean age of smokers was 62 years, whereas that of non-smokers was 70 years. Likewise, 60% of those on beta blockers had a history of ischaemic heart disease, whereas only 30% of those not on a beta blocker had a history of ischaemic heart disease.
      Strengths of this work include the large, national database used, the rigorous statistical methods used both to handle missing data, and the information criterion approach to reduce the chances of over fitting. The model presented has shown discrimination which improves on the previously published models (Fig. 1).
      Limitations of the study include the fact that the case completion rate in the NVR is known to be only around 60%.
      • Waton S.
      • Johal A.
      • Heikkela K.
      • Cromwell D.
      • Miller F.
      • Boyle J.R.
      National vascular registry annual report.
      This is a dramatic improvement over the situation 10 years ago, when only half this number of cases was entered.
      • Lees T.
      • Stansby G.
      • Baker S.
      • Mitchell D.
      • Holt P.
      • Taylor P.
      • et al.
      National vascular database report 2009.
      The UK NVR reports for the past two years have highlighted the fact that case ascertainment rates vary widely between Vascular Networks.
      • Waton S.
      • Johal A.
      • Heikkela K.
      • Cromwell D.
      • Miller F.
      • Boyle J.R.
      National vascular registry annual report.
      ,
      • Waton S.
      • Johal A.
      • Heikkela K.
      • Cromwell D.
      • Loftus I.M.
      • Boyle J.R.
      National vascular registry annual report.
      Therefore, there is optimism that much of the missing data relate largely to institutional and administrative factors rather than patient related factors, but the potential that the non-submitted cases might be systematically different from submitted cases, introducing bias into the results, must be acknowledged. In addition to missing cases, there was also a degree of missing data items within otherwise completed cases. However, it must also be stressed that given the size of the database, even in cases where up to 50% of values are missing for a parameter, there will still be data from around 5 000 (or more) cases. This, together with the rigorous multiple imputation methodology (the gold standard method for handling missing data in clinical research),
      • Little R.J.
      • D'Agostino R.
      • Cohen M.L.
      • Dickersin K.
      • Emerson S.S.
      • Farrar J.T.
      • et al.
      The prevention and treatment of missing data in clinical trials.
      gives confidence that the results are valid. Sensitivity analysis using only complete cases gave very similar results (Table 2), so there is no evidence that these missing values have introduced significant bias.
      Validation of data within the NVR is also lacking. This is a general criticism of registry based studies, as to the authors' knowledge, no national registry of major lower limb amputation cases has been rigorously validated. The Swedish Vascular registry (Swedvasc) and the Hungarian registry have been validated, although the former only for aortic aneurysm repair and carotid surgery, while the latter also for infra-inguinal arterial reconstruction.
      • Venermo M.
      • Lees T.
      International vascunet validation of the Swedvasc registry.
      ,
      • Bergqvist D.
      • Bjorck M.
      • Lees T.
      • Menyhei G.
      Validation of the VASCUNET registry – pilot study.
      Plans are in place for a validation exercise of the UK NVR in 2020, but this may not include the major lower limb amputation subset.
      A further weakness of this study is limitations of the data recorded in the NVR. For example, it is increasingly recognised that frailty is an important risk factor for peri-operative complications, including mortality.
      • Ambler G.K.
      • Brooks D.E.
      • Al Zuhir N.
      • Ali A.
      • Gohel M.S.
      • Hayes P.D.
      • et al.
      Effect of frailty on short- and mid-term outcomes in vascular surgical patients.
      Dependent functional status has been shown in other work to be important for predicting mortality in patients undergoing amputation.
      • Easterlin M.C.
      • Chang D.C.
      • Wilson S.E.
      A practical index to predict 30-day mortality after major amputation.
      ,
      • Feinglass J.
      • Pearce W.H.
      • Martin G.J.
      • Gibbs J.
      • Cowper D.
      • Sorensen M.
      • et al.
      Postoperative and late survival outcomes after major amputation: findings from the department of Veterans Affairs national surgical quality improvement program.
      However, until recently no measure of frailty or functional status has been recorded in the NVR. A measure of frailty has recently been added to the NVR data set, which will allow further investigation of this factor in the future.
      In hospital mortality has been modelled, as that is the audit standard within the UK Vascular Registry. Unfortunately, this is different from many other national audit databases such as Swedvasc, which reports 30 day mortality. Inconsistency in outcome reporting presents difficulties for clinical audit and research, as it makes pooling of information between studies (meta-analysis) challenging. Efforts are underway to develop Core Outcome Sets for patients undergoing major lower limb amputation for complications of peripheral vascular disease.
      • Ambler G.K.
      • Bosanquet D.C.
      • Brookes-Howell L.
      • Thomas-Jones E.
      • Waldron C.A.
      • Edwards A.G.K.
      • et al.
      Development of a core outcome set for studies involving patients undergoing major lower limb amputation for peripheral arterial disease: study protocol for a systematic review and identification of a core outcome set using a Delphi survey.
      Several of the predictive factors found herein are similar to those found in previous work. Increasing age was found to be an independent predictor of mortality in almost all previous studies, including this one. Emergency admission and level of amputation were also found to be predictive of mortality in several other studies, including work from large administrative databases in Japan and the USA.
      • Yamada K.
      • Yasunaga H.
      • Kadono Y.
      • Chikuda H.
      • Ogata T.
      • Horiguchi H.
      • et al.
      Postoperative outcomes of major lower extremity amputations in patients with diabetes and peripheral artery disease: analysis using the Diagnosis Procedure Combination database in Japan.
      ,
      • Feinglass J.
      • Pearce W.H.
      • Martin G.J.
      • Gibbs J.
      • Cowper D.
      • Sorensen M.
      • et al.
      Postoperative and late survival outcomes after major amputation: findings from the department of Veterans Affairs national surgical quality improvement program.
      Evidence of systemic sepsis in the form of a raised white cell count has also been identified as a significant factor in previous studies.
      • Tang T.Y.
      • Prytherch D.R.
      • Walsh S.R.
      • Athanassoglou V.
      • Seppi V.
      • Sadat U.
      • et al.
      The development of a VBHOM-based outcome model for lower limb amputation performed for critical ischaemia.
      ,
      • Easterlin M.C.
      • Chang D.C.
      • Wilson S.E.
      A practical index to predict 30-day mortality after major amputation.
      By contrast, bilateral procedures have not been shown previously in multivariable analysis to have a worse outcome than unilateral procedures, and increased patient weight has never been identified as an independent protective factor previously.
      Further studies are needed to identify whether attempts to modify any of the factors have a clinically relevant impact on outcomes. Firstly, improvements might be made through quality improvement programmes designed to facilitate earlier identification and treatment of patients for whom further attempts at limb salvage are at high risk of failure. Improved shared decision making using risk quantified by these data should be encouraged, perhaps supported by a decision aid.
      • Joseph-Williams N.
      • Lloyd A.
      • Edwards A.
      • Stobbart L.
      • Tomson D.
      • Macphail S.
      • et al.
      Implementing shared decision making in the NHS: lessons from the MAGIC programme.
      Secondly, there are more speculative options that would require testing in prospective interventional studies. One of these is a proposed intervention to facilitate healing at a more distal level, as both prior procedures to the amputated limb and below knee operations appear to have a protective role. Increased patient weight and serum albumin have similar protective effects, so it is possible that pre-operative dietary intervention or other “pre-habilitation” might also be helpful for patients with stable but unreconstructable arterial disease. There is increasing interest in the putative benefits of “pre-habilitation”, with multiple ongoing studies,
      • McIsaac D.I.
      • Saunders C.
      • Hladkowicz E.
      • Bryson G.L.
      • Forster A.J.
      • Gagne S.
      • et al.
      PREHAB study: a protocol for a prospective randomised clinical trial of exercise therapy for people living with frailty having cancer surgery.
      although little concrete evidence of benefit, at this time.
      • Milder D.A.
      • Pillinger N.L.
      • Kam P.C.A.
      The role of prehabilitation in frail surgical patients: a systematic review.
      Further work is also needed to externally validate the predictive model. The importance of this was highlighted within vascular surgery recently with the publication of the draft National Institute for Health and Care Excellence (NICE) guidelines for the treatment of abdominal aortic aneurysm (AAA),
      National Institute for Health and Care Excellence
      Abdominal aortic aneurysm: diagnosis and management.
      which found that none of the models for the patient group that had been subjected to external validation was found to have good discriminatory power. The NICE panel also suggested that the literature on risk scoring in AAA repair reports several unvalidated risk scores and concluded that there is no justification to develop further models in this context, until the external validity of existing ones is assessed. The situation for major limb amputation is somewhat different, with very few available models. Those that do exist are difficult to validate in the UK as they use parameters that are not routinely recorded in the NVR. The authors are therefore optimistic that the developed model has the potential to contribute to improvement of outcomes for patients with chronic limb threatening ischaemia. Work on validating the model is planned for next year once data from the next two years is available.
      In conclusion, independent risk factors for mortality and morbidity following major lower limb amputation have been identified and a prognostic model for in hospital mortality with good predictive power developed. Important next steps include further external validation, and, if supported, the development of quality improvement programmes that focus on modification of the factors identified, and adjustment of published surgical outcomes using this model.

      Conflicts of Interest

      None.

      Funding

      GKA was supported by the National Institute for Health Research (NIHR) Biomedical Research Centre at University Hospitals Bristol NHS Foundation Trust and the University of Bristol for part of the time that this research was carried out. The views expressed in this publication are those of the author(s) and not necessarily those of the National Health Service, the NIHR, or the Department of Health and Social Care.

      Acknowledgements

      The authors would like to thank Sam Waton, the National Vascular Registry Manager, for his help in obtaining the data and answering queries about the database. This paper was presented at the Vascular Society of Great Britain and Ireland annual scientific meeting in Glasgow in November 2018.

      Appendix A. Supplementary data

      The following are the Supplementary data to this article:

      References

        • Waton S.
        • Johal A.
        • Heikkela K.
        • Cromwell D.
        • Loftus I.
        National vascular registry annual report.
        Vascular Society of Great Britain and Ireland, London2016
        • Tang T.Y.
        • Prytherch D.R.
        • Walsh S.R.
        • Athanassoglou V.
        • Seppi V.
        • Sadat U.
        • et al.
        The development of a VBHOM-based outcome model for lower limb amputation performed for critical ischaemia.
        Eur J Vasc Endovasc Surg. 2009; 37: 62-66
        • Patterson A.J.
        • Degnan A.J.
        • Walsh S.R.
        • Eltayeb M.
        • Scout E.F.
        • Clarke J.M.
        • et al.
        Efficacy of VBHOM to predict outcome following major lower limb amputation.
        Vasc Endovascular Surg. 2012; 46: 369-373
        • Pell J.
        • Stonebridge P.
        Association between age and survival following major amputation. The Scottish Vascular Audit Group.
        Eur J Vasc Endovasc Surg. 1999; 17: 166-169
        • Easterlin M.C.
        • Chang D.C.
        • Wilson S.E.
        A practical index to predict 30-day mortality after major amputation.
        Ann Vasc Surg. 2013; 27: 909-917
        • Yamada K.
        • Yasunaga H.
        • Kadono Y.
        • Chikuda H.
        • Ogata T.
        • Horiguchi H.
        • et al.
        Postoperative outcomes of major lower extremity amputations in patients with diabetes and peripheral artery disease: analysis using the Diagnosis Procedure Combination database in Japan.
        Am J Surg. 2016; 212: 446-450
        • Feinglass J.
        • Pearce W.H.
        • Martin G.J.
        • Gibbs J.
        • Cowper D.
        • Sorensen M.
        • et al.
        Postoperative and late survival outcomes after major amputation: findings from the department of Veterans Affairs national surgical quality improvement program.
        Surgery. 2001; 130: 21-29
        • National Confidential Enquiry Into Patient Outcome and Death
        Lower limb amputation: working together.
        NCEPOD, London2014
        • Fryar C.
        Clarifying the montgomery judgment.
        BMJ. 2015; 350: h2217
        • Walker K.
        • Neuburger J.
        • Groene O.
        • Cromwell D.A.
        • van der Meulen J.
        Public reporting of surgeon outcomes: low numbers of procedures lead to false complacency.
        Lancet. 2013; 382: 1674-1677
        • Nashef S.A.
        • Roques F.
        • Michel P.
        • Gauducheau E.
        • Lemeshow S.
        • Salamon R.
        European system for cardiac operative risk evaluation (EuroSCORE).
        Eur J Cardiothorac Surg. 1999; 16: 9-13
        • Nashef S.A.
        • Roques F.
        • Sharples L.D.
        • Nilsson J.
        • Smith C.
        • Goldstone A.R.
        • et al.
        EuroSCORE II.
        Eur J Cardiothorac Surg. 2012; 41: 734-744
        • von Elm E.
        • Altman D.G.
        • Egger M.
        • Pocock S.J.
        • Gøtzsche P.C.
        • Vandenbroucke J.P.
        The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.
        J Clin Epidemiol. 2008; 61: 344-349
        • Collins G.S.
        • Reitsma J.B.
        • Alstman D.G.
        • Moons K.G.M.
        Transparent reporting of a multivariable prediction model for individual Prognosis or diagnosis (TRIPOD): the TRIPOD statement.
        Ann Intern Med. 2015; 162: 55-63
        • R Core Team
        R: a language and environment for statistical computing.
        R Foundation for Statistical Computing, Vienna2018
        • van Buuren S.
        • Groothuis-Oudshoorn K.
        Mice: multivariate imputation by chained equations in R.
        J Stat Softw. 2011; 45: 1-67
        • Little R.J.
        • Rubin D.B.
        Statistical analysis with missing data.
        John Wiley, New York2002
        • Schwarz G.
        Estimating the dimension of a model.
        Ann Stat. 1978; 6: 461-464
        • Robin X.
        • Turck N.
        • Hainard A.
        • Tiberti N.
        • Lisacek F.
        • Sanchez J.-C.
        • et al.
        pROC: an open-source package for R and S+ to analyze and compare ROC curves.
        BMC Bioinformatics. 2011; 12: 77
        • DeLong E.R.
        • DeLong D.M.
        • Clarke-Pearson D.L.
        Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach.
        Biometrics. 1988; 44: 837-845
        • Hosmer D.W.
        • Lemeshow S.
        • Sturdivant R.X.
        Applied logistic regression.
        3rd ed. John Wiley, Hoboken, NJ2013
        • Flores A.M.
        • Mell M.W.
        • Dalman R.L.
        • Chandra V.
        Benefit of multidisciplinary wound care center on the volume and outcomes of a vascular surgery practice.
        J Vasc Surg. 2019; 70: 1612-1619
        • Neville R.F.
        • Kayssi A.
        Development of a limb-preservation program.
        Blood Purif. 2017; 43: 218-225
        • Elwyn G.
        • Frosch D.
        • Thomson R.
        • Joseph-Williams N.
        • Lloyd A.
        • Kinnersley P.
        • et al.
        Shared decision making: a model for clinical practice.
        J Gen Intern Med. 2012; 27: 1361-1367
        • Timmermans D.
        • Kievit J.
        • van Bockel H.
        How do surgeons' probability estimates of operative mortality compare with a decision analytic model?.
        Acta Psychol. 1996; 93: 107-120
        • Glasgow R.E.
        • Hawn M.T.
        • Hosokawa P.W.
        • Henderson W.G.
        • Min S.J.
        • Richman J.S.
        • et al.
        Comparison of prospective risk estimates for postoperative complications: human vs computer model.
        J Am Coll Surg. 2014; 218: 237-245
        • Conte M.S.
        • Bradbury A.W.
        • Kolh P.
        • White J.V.
        • Dick F.
        • Fitridge R.
        • et al.
        Global vascular guidelines on the management of chronic limb-threatening ischemia.
        Eur J Vasc Endovasc Surg. 2019; 58: S1-S109
        • Saratzis A.
        • Jaspers N.E.M.
        • Gwilym B.
        • Thomas O.
        • Tsui A.
        • Lefroy R.
        • et al.
        Observational study of the medical management of patients with peripheral artery disease.
        Br J Surg. 2019; 106: 1168-1177
        • Waton S.
        • Johal A.
        • Heikkela K.
        • Cromwell D.
        • Miller F.
        • Boyle J.R.
        National vascular registry annual report.
        Vascular Society of Great Britain and Ireland, London2018
        • Lees T.
        • Stansby G.
        • Baker S.
        • Mitchell D.
        • Holt P.
        • Taylor P.
        • et al.
        National vascular database report 2009.
        Vascular Society of Great Britain and Ireland, London2009
        • Waton S.
        • Johal A.
        • Heikkela K.
        • Cromwell D.
        • Loftus I.M.
        • Boyle J.R.
        National vascular registry annual report.
        Vascular Society of Great Britain and Ireland, London2017
        • Little R.J.
        • D'Agostino R.
        • Cohen M.L.
        • Dickersin K.
        • Emerson S.S.
        • Farrar J.T.
        • et al.
        The prevention and treatment of missing data in clinical trials.
        N Engl J Med. 2012; 367: 1355-1360
        • Venermo M.
        • Lees T.
        International vascunet validation of the Swedvasc registry.
        Eur J Vasc Endovasc Surg. 2015; 50: 802-808
        • Bergqvist D.
        • Bjorck M.
        • Lees T.
        • Menyhei G.
        Validation of the VASCUNET registry – pilot study.
        Vasa. 2014; 43: 141-144
        • Ambler G.K.
        • Brooks D.E.
        • Al Zuhir N.
        • Ali A.
        • Gohel M.S.
        • Hayes P.D.
        • et al.
        Effect of frailty on short- and mid-term outcomes in vascular surgical patients.
        Br J Surg. 2015; 102: 638-645
        • Ambler G.K.
        • Bosanquet D.C.
        • Brookes-Howell L.
        • Thomas-Jones E.
        • Waldron C.A.
        • Edwards A.G.K.
        • et al.
        Development of a core outcome set for studies involving patients undergoing major lower limb amputation for peripheral arterial disease: study protocol for a systematic review and identification of a core outcome set using a Delphi survey.
        Trials. 2017; 18: 628
        • Joseph-Williams N.
        • Lloyd A.
        • Edwards A.
        • Stobbart L.
        • Tomson D.
        • Macphail S.
        • et al.
        Implementing shared decision making in the NHS: lessons from the MAGIC programme.
        BMJ. 2017; 357: j1744
        • McIsaac D.I.
        • Saunders C.
        • Hladkowicz E.
        • Bryson G.L.
        • Forster A.J.
        • Gagne S.
        • et al.
        PREHAB study: a protocol for a prospective randomised clinical trial of exercise therapy for people living with frailty having cancer surgery.
        BMJ Open. 2018; 8
        • Milder D.A.
        • Pillinger N.L.
        • Kam P.C.A.
        The role of prehabilitation in frail surgical patients: a systematic review.
        Acta Anaesthesiol Scand. 2018; 62: 1356-1366
        • National Institute for Health and Care Excellence
        Abdominal aortic aneurysm: diagnosis and management.
        (Available at:)

      Comments

      Commenting Guidelines

      To submit a comment for a journal article, please use the space above and note the following:

      • We will review submitted comments as soon as possible, striving for within two business days.
      • This forum is intended for constructive dialogue. Comments that are commercial or promotional in nature, pertain to specific medical cases, are not relevant to the article for which they have been submitted, or are otherwise inappropriate will not be posted.
      • We require that commenters identify themselves with names and affiliations.
      • Comments must be in compliance with our Terms & Conditions.
      • Comments are not peer-reviewed.