Doug Altman, medical statistician par excellence: What can radiologists learn from his legacy?

This narrative review describes our experience of working with Doug Altman, the most highly cited medical statistician in the world. Doug was particularly interested in diagnostics, and imaging studies in particular. We describe how his insights helped improve our own radiological research studies and we provide advice for other researchers hoping to improve their own research practice.

Access to statistical support for medical imaging research: questionnaire survey of UK radiology trainees.

AIM: To investigate the level of statistical support available to UK radiology trainees, and to gather opinions regarding how support may affect their current and future research aspirations. MATERIALS AND METHODS: An online survey was developed, piloted, and distributed to radiology trainees via the UK Radiology Academic Network for Trainees and training programme directors. Research experience, research aspirations, available and desired statistical support, and attitudes to statistics were surveyed and responses were collated. RESULTS: Seventy-nine responses were received, only two (3%) of whom had allocated time for research. Only three (4%) respondents were content with their statistical support whereas 25 (32%) reported insufficient statistical support; 13 (52%) of these believed this impacted "considerably" on research aspirations. Sixty-six (84%) respondents desired dedicated statistical support, 40 (61%) of whom stated the amount required would likely be "moderate" and 26 (39%) "significant". Respondents believed support would be most helpful to analyse data already collected (41 responses, 54%) rather than research planning (25, 33%). Most respondents (60, 76%) had used self-help methods to learn research statistics but only 21 (35%) found this useful. CONCLUSION: Training schemes must improve the provision, access to, and awareness of statistical support so that any research efforts are performed to a high standard. Trainees should not be expected to participate in research without sufficient time, mentorship, and statistical support.

Reporting guideline for interventional trials of primary and incisional ventral hernia repair.

BACKGROUND: Primary and incisional ventral hernia trials collect unstandardized inconsistent data, limiting data interpretation and comparison. This study aimed to create two minimum data sets for primary and incisional ventral hernia interventional trials to standardize data collection and improve trial comparison. To support these data sets, standardized patient-reported outcome measures and trial methodology criteria were created. METHODS: To construct these data sets, nominal group technique methodology was employed, involving 15 internationally recognized abdominal wall surgeons and two patient representatives. Initially a maximum data set was created from previous systematic and panellist reviews. Thereafter, three stages of voting took place: stage 1, selection of the number of variables for data set inclusion; stage 2, selection of variables to be included; and stage 3, selection of variable definitions and detection methods. A steering committee interpreted and analysed the data. RESULTS: The maximum data set contained 245 variables. The three stages of voting commenced in October 2019 and had been completed by July 2020. The final primary ventral hernia data set included 32 variables, the incisional ventral hernia data set included 40 variables, the patient-reported outcome measures tool contained 25 questions, and 40 methodological criteria were chosen. The best known variable definitions were selected for accurate variable description. CT was selected as the optimal preoperative descriptor of hernia morphology. Standardized follow-up at 30 days, 1 year, and 5 years was selected. CONCLUSION: These minimum data sets, patient-reported outcome measures, and methodological criteria have allowed creation of a manual for investigators aiming to undertake primary ventral hernia or incisional ventral hernia interventional trials. Adopting these data sets will improve trial methods and comparisons.

What Exactly is Meant by "Loss of Domain" for Ventral Hernia? Systematic Review of Definitions.

Large ventral hernias are a significant surgical challenge. "Loss of domain" (LOD) expresses the relationship between hernia and abdominal volume, and is used to predict operative difficulty and success. This systematic review assessed whether different definitions of LOD are used in the literature. The PubMed database was searched for articles reporting large hernia repairs that explicitly described LOD. Two reviewers screened citations and extracted data from selected articles, focusing on the definitions used for LOD, study demographics, study design, and reporting surgical specialty. One hundred and seven articles were identified, 93 full-texts examined, and 77 were included in the systematic review. Sixty-seven articles were from the primary literature, and 10 articles were from the secondary literature. Twenty-eight articles (36%) gave a written definition for loss of domain. These varied and divided into six broad groupings; four described the loss of the right of domain, six described abdominal strap muscle contraction, five described the "second abdomen", five describing large irreducible hernias. Six gave miscellaneous definitions. Two articles gave multiple definitions. Twenty articles (26%) gave volumetric definitions; eight used the Tanaka method [hernia sac volume (HSV)/abdominal cavity volume] and five used the Sabbagh method [(HSV)/total peritoneal volume]. The definitions used for loss of domain were not dependent on the reporting specialty. Our systematic review revealed that multiple definitions of loss of domain are being used. These vary and are not interchangeable. Expert consensus on this matter is necessary to standardise this important concept for hernia surgeons.

Temporal and external validation of a prediction model for adverse outcomes among inpatients with diabetes.

AIM: To temporally and externally validate our previously developed prediction model, which used data from University Hospitals Birmingham to identify inpatients with diabetes at high risk of adverse outcome (mortality or excessive length of stay), in order to demonstrate its applicability to other hospital populations within the UK. METHODS: Temporal validation was performed using data from University Hospitals Birmingham and external validation was performed using data from both the Heart of England NHS Foundation Trust and Ipswich Hospital. All adult inpatients with diabetes were included. Variables included in the model were age, gender, ethnicity, admission type, intensive therapy unit admission, insulin therapy, albumin, sodium, potassium, haemoglobin, C-reactive protein, estimated GFR and neutrophil count. Adverse outcome was defined as excessive length of stay or death. RESULTS: Model discrimination in the temporal and external validation datasets was good. In temporal validation using data from University Hospitals Birmingham, the area under the curve was 0.797 (95% CI 0.785-0.810), sensitivity was 70% (95% CI 67-72) and specificity was 75% (95% CI 74-76). In external validation using data from Heart of England NHS Foundation Trust, the area under the curve was 0.758 (95% CI 0.747-0.768), sensitivity was 73% (95% CI 71-74) and specificity was 66% (95% CI 65-67). In external validation using data from Ipswich, the area under the curve was 0.736 (95% CI 0.711-0.761), sensitivity was 63% (95% CI 59-68) and specificity was 69% (95% CI 67-72). These results were similar to those for the internally validated model derived from University Hospitals Birmingham. CONCLUSIONS: The prediction model to identify patients with diabetes at high risk of developing an adverse event while in hospital performed well in temporal and external validation. The externally validated prediction model is a novel tool that can be used to improve care pathways for inpatients with diabetes. Further research to assess clinical utility is needed.

Intraoperative transfusion practices in Europe.

BACKGROUND: Transfusion of allogeneic blood influences outcome after surgery. Despite widespread availability of transfusion guidelines, transfusion practices might vary among physicians, departments, hospitals and countries. Our aim was to determine the amount of packed red blood cells (pRBC) and blood products transfused intraoperatively, and to describe factors determining transfusion throughout Europe. METHODS: We did a prospective observational cohort study enrolling 5803 patients in 126 European centres that received at least one pRBC unit intraoperatively, during a continuous three month period in 2013. RESULTS: The overall intraoperative transfusion rate was 1.8%; 59% of transfusions were at least partially initiated as a result of a physiological transfusion trigger- mostly because of hypotension (55.4%) and/or tachycardia (30.7%). Haemoglobin (Hb)- based transfusion trigger alone initiated only 8.5% of transfusions. The Hb concentration [mean (sd)] just before transfusion was 8.1 (1.7) g dl(-1) and increased to 9.8 (1.8) g dl(-1) after transfusion. The mean number of intraoperatively transfused pRBC units was 2.5 (2.7) units (median 2). CONCLUSION: Although European Society of Anaesthesiology transfusion guidelines are moderately implemented in Europe with respect to Hb threshold for transfusion (7-9 g dl(-1)), there is still an urgent need for further educational efforts that focus on the number of pRBC units to be transfused at this threshold. CLINICAL TRIAL REGISTRATION: NCT 01604083.

Alterations in coagulation following major liver resection.

The international normalised ratio is frequently raised in patients who have undergone major liver resection, and is assumed to represent a potential bleeding risk. However, these patients have an increased risk of venous thromboembolic events, despite conventional coagulation tests indicating hypocoagulability. This prospective, observational study of patients undergoing major hepatic resection analysed the serial changes in coagulation in the early postoperative period. Thrombin generation parameters and viscoelastic tests of coagulation (thromboelastometry) remained within normal ranges throughout the study period. Levels of the procoagulant factors II, V, VII and X initially fell, but V and X returned to or exceeded normal range by postoperative day five. Levels of factor VIII and Von Willebrand factor were significantly elevated from postoperative day one (p < 0.01). Levels of the anticoagulants, protein C and antithrombin remained significantly depressed on postoperative day five (p = 0.01). Overall, the imbalance between pro- and anticoagulant factors suggested a prothrombotic environment in the early postoperative period.

Patient blood management to reduce surgical risk.

BACKGROUND: Preoperative anaemia and perioperative blood transfusion are both identifiable and preventable surgical risks. Patient blood management is a multimodal approach to address this issue. It focuses on three pillars of care: the detection and treatment of preoperative anaemia; the reduction of perioperative blood loss; and harnessing and optimizing the patient-specific physiological reserve of anaemia, including restrictive haemoglobin transfusion triggers. This article reviews why patient blood management is needed and strategies for its incorporation into surgical pathways. METHODS: Studies investigating the three pillars of patient blood management were identified using PubMed, focusing on recent evidence-based guidance for perioperative management. RESULTS: Anaemia is common in surgical practice. Both anaemia and blood transfusion are independently associated with adverse outcomes. Functional iron deficiency (iron restriction due to increased levels of hepcidin) is the most common cause of preoperative anaemia, and should be treated with intravenous iron. Intraoperative blood loss can be reduced with antifibrinolytic drugs such as tranexamic acid, and cell salvage should be used. A restrictive transfusion practice should be the standard of care after surgery. CONCLUSION: The significance of preoperative anaemia appears underappreciated, and its detection should lead to routine investigation and treatment before elective surgery. The risks of unnecessary blood transfusion are increasingly being recognized. Strategic adoption of patient blood management in surgical practice is recommended, and will reduce costs and improve outcomes in surgery.

An in vitro study comparing two dose regimes of fresh frozen plasma on conventional and thromboelastographic tests of coagulation after major hepatic resection.

BACKGROUND: After hepatic resection, post-operative increases in international normalised ratio (INR) are frequent, but rarely associated with bleeding complication. Coagulation as assessed by thrombin generation may be normal, despite the increased INR. This study tests the hypothesis that viscoelastic tests (VET) of coagulation remain normal, despite mild elevations in INR, examines the efficacy of fresh frozen plasma (FFP) in reversing prolongation of INR in such patients and determines the effect of FFP on VET. METHODS: A prospective cohort study of 47 patients undergoing major hepatectomy. In vitro spiking with FFP (7·5 and 15 mL kg(-1)) was carried out if post-operative day 2 (POD2) INR levels were ≥1·5. Thromboelastography (TEG®) and INR were measured before and after FFP spiking. RESULTS: Blood from patients with an INR ≥ 1·5 on POD2 was spiked with FFP. There was a significant reduction in the INR from 1·94 [standard deviation (SD): 0·59] to 1·46 (SD: 0·27, P = 0·005) and 1·36 (SD: 0·18, P = 0·0007) with FFP 7·5 or 15 mL kg(-1), respectively. At baseline, the TEG R-time [6·17 min (NR, 9-27 min)] and maximum amplitude (MA) [66·9 mm (NR, 44-64 mm)] were hypercoagulable, and remained so on POD2 for the R-time (6·7 min), but fell to within the normal range for the MA (54·0 mm). FFP spiking had no significant effect on TEG variables. CONCLUSIONS: Despite the rise in INR after hepatectomy, VET do not show evidence of hypocoagulability. In vitro addition of FFP had no significant effect on TEG parameters. Clinical use of FFP in this situation is questionable.

Thromboelastography (TEG®) compared with total platelet count in thrombocytopenia haematological malignancy patients with bleeding: a pilot observational study.

BACKGROUND: There has been no reported use of Thromboelastography(®) (TEG(®) ) in assessing thrombocytopenic haematological malignancy (HM) patients experiencing bleeding. OBJECTIVES: To assess whether there are differences in TEG(®) variables in thrombocytopenic HM patients experiencing clinically significant bleeding compared with those not experiencing bleeding. METHODS: Thirty adult patients with HM and a total platelet count (TPC) of ≤ 30 × 10(9) L(-1) were observed for greater than or equal to grade two World Health Organisation (WHO) bleeding episodes for 72 h. They had TPC, TEG(®) parameters [maximal amplitude (MA), reaction time (R-time), alpha (α-angle) and functional fibrinogen (FF) levels], activated partial thromboplastin time (APTT) and prothrombin time (PT) measured at inclusion, 24 and 48 h. RESULTS: Five patients had nine significant bleeding episodes. Patients bleeding (n) were all more hypocoaguable when measured by TEG(®) than patients not bleeding at inclusion n = 4 (MA: 16.9 vs 31.8 mm, P < 0.01; R-time: 10.1 vs 7.5 min, P = 0.02; α-angle: 18.8 vs 28.4, P < 0.01; FF: 182.4 vs 204 mg dL(-1), P < 0.01), at 24 h n = 3 (MA 18 vs 29.4 mm, P = 0.02; R-time: 9.4 vs 7.4 mins, P = 0.02; α-angle: 21.6 vs 26.5, P = 0.04 and FF: 168 vs 201 mg dL(-1) P = 0.01) and at 48 h n = 2 (MA: 29.7 vs 23.0, P = 0.02; R-time: 8.1 vs 6.7 mins, P = 0.05; α-angle: 22.6 vs 26.5, P = 0.04; FF: 170 vs 208 mg dL(-1), P = 0.01). There were no significant differences in TPC, APTT or PT between patients that bled and those that did not at each time point. CONCLUSIONS: Thrombocytopenic HM patients experiencing bleeding are hypocoaguable on TEG(®).

Point-of-care monitoring of haemostasis.

Recent research in the management of haemorrhage has led to several changes in clinical practice. Evidence is accumulating that point-of-care testing results in fewer transfusions, improved patient outcomes, and reduced hospital costs. However, there is still insufficient high quality evidence to support transfusion guidelines and algorithms based on point-of-care tests alone, and more robust studies are needed. The implementation of point-of-care testing requires institutional support and senior clinical leadership to realise the benefits, with educational programmes, audit, and feedback regarding transfusion practice. A change in philosophy is required, from performing testing only when there is an obvious bleeding problem, towards the concept of routinely monitoring high-risk patients throughout the surgical procedure. This informs clinical practice, establishes normal ranges for that population, identifies patients at risk and allows early identification and treatment of evolving coagulopathy.

Risk of bleeding and adverse outcomes predicted by thromboelastography platelet mapping in patients taking clopidogrel within 7 days of non-cardiac surgery.

BACKGROUND: Patients often fail to stop clopidogrel appropriately before non-cardiac surgery. Thromboelastography platelet mapping (TEG-PM) can be used to measure the percentage adenosine 5'-diphosphate platelet receptor inhibition (ADP-PRI) by clopidogrel in these patients. METHODS: This prospective case-control study investigated the risk of bleeding in patients who had taken clopidogrel within 7 days of scheduled operation. Patients underwent TEG-PM to stratify their bleeding risk. Low-risk (ADP-PRI below 30 per cent) and urgent priority high-risk (ADP-PRI 30 per cent or more) patients proceeded to surgery. The outcomes of these patients were compared with those of matched controls. Regression analysis, with bootstrapping validation, was used to identify independent risk factors for bleeding and an optimal cut-off value of ADP-PRI for cancellation of surgery. RESULTS: From May 2008 to October 2013, 182 patients failed to discontinue clopidogrel. No correlation was observed between duration of clopidogrel omission and percentage ADP-PRI; 112 low-risk and 19 high-risk patients proceeded to surgery. High-risk patients had significantly greater intraoperative packed red blood cell (PRBC) transfusion in comparison with their matched controls, and a strong positive correlation between percentage ADP-PRI and units of intraoperative PRBCs transfused (r = 0·749, 95 per cent confidence interval (c.i.) 0·410 to 0·940; P < 0·001). Percentage ADP-PRI was the only independent risk factor for intraoperative PRBC transfusion (odds ratio 1·07, 95 per cent c.i. 1·02 to 1·13; P = 0·005). CONCLUSION: An objective measure of platelet inhibition with TEG-PM, using an ADP-PRI cut-off of 34 per cent, can be used to prevent unnecessary cancellations, while minimizing patient risk.

Spawning activity of the Australian lungfish Neoceratodus forsteri in an impoundment.

This study assessed the spawning activity of the threatened Australian lungfish Neoceratodus forsteri by measuring egg densities within the artificial habitat of a large impoundment (Lake Wivenhoe, Australia). Eggs were sampled (August to November 2009) from multiple locations across the impoundment, but occurred at highest densities in water shallower than 40 cm along shorelines with a dense cover of submerged terrestrial vegetation. The numbers of eggs declined over the study period and all samples were dominated by early developmental stages and high proportions of unviable eggs. The quality of the littoral spawning habitats declined over the study as flooded terrestrial grasses decomposed and filamentous algae coverage increased. Water temperatures at the spawning site exhibited extreme variations, ranging over 20·4° C in water shallower than 5 cm. Dissolved oxygen concentrations regularly declined to <1 mg l⁻¹ at 40 and 80 cm water depth. Spawning habitats utilised by N. forsteri within impoundments expose embryos to increased risk of desiccation or excessive submergence through water-level variations, and extremes in temperature and dissolved oxygen concentration that present numerous challenges for successful spawning and recruitment of N. forsteri in large impoundment environments.

Corrigendum to 'Patient blood management and perioperative anaemia' [BJA Education 17 (2017) 28-34].

[This corrects the article DOI: 10.1093/bjaed/mkw061.].

Epidural anaesthesia and analgesia for liver resection.

Although epidural analgesia is routinely used in many institutions for patients undergoing hepatic resection, there are unresolved issues regarding its safety and efficacy in this setting. We performed a review of papers published in the area of anaesthesia and analgesia for liver resection surgery and selected four areas of current controversy for the focus of this review: the safety of epidural catheters with respect to postoperative coagulopathy, a common feature of this type of surgery; analgesic efficacy; associated peri-operative fluid administration; and the role of epidural analgesia in enhanced recovery protocols. In all four areas, issues are raised that question whether epidural anaesthesia is always the best choice for these patients. Unfortunately, the evidence available is insufficient to provide definitive answers, and it is clear that there are a number of areas of controversy that would benefit from high-quality clinical trials.

Bystander cardiopulmonary resuscitation for paediatric out-of-hospital cardiac arrest in England: An observational registry cohort study.

INTRODUCTION: Bystander cardiopulmonary resuscitation (BCPR) is strongly advocated by resuscitation councils for paediatric out-of-hospital cardiac arrests (OHCAs). However, there are limited reports on rates of BCPR in children and its relationship with return of spontaneous circulation (ROSC) or survival outcomes. OBJECTIVE: We describe the rate of BCPR and its association with any ROSC and survival- to- hospital-discharge. METHODS: We conducted retrospective analysis of prospectively collected paediatric (<18 years of age) OHCA cases in England; we included specialist registry patients treated by emergency medical services (EMS) with known BCPR status and outcome between January 2014 and November 2018. Data included patient demographics, aetiology, witness status, initial rhythm, EMS, season, time of day and bystander status. Associations between BCPR, and any ROSC and survival-to-hospital-discharge outcomes were explored using multivariable logistic regression. RESULTS: There were 2363 paediatric OHCAs treated across 11 EMS regions. BCPR was performed in 69.6% (1646/2363) of the cases overall (range 57.7% (206/367) to 83.7% (139/166) across EMS regions). Only 34.9% (550/1572) of BCPR cases were witnessed. Overall, any ROSC was achieved in 22.8% (523/2289) and survival to hospital discharge in 10.8% (225/2066). Adjusted odds ratio (aOR) for any ROSC was significantly improved following BCPR compared to no BCPR (aOR 1.37, 95% CI 1.03-1.81), but adjusted odds ratio for survival-to-hospital-discharge were similar (aOR 1.01, 95% CI 0.66-1.55). CONCLUSIONS: BCPR was associated with improved rates of any ROSC but not survival-to-hospital-discharge. Variations in EMS BCPR rates may indicate opportunities for regional targeted increase in public BCPR education.

Identifying predictors of ventral hernia recurrence: systematic review and meta-analysis.

BACKGROUND: Ventra hernias are increasing in prevalence and many recur despite attempted repair. To date, much of the literature is underpowered and divergent. As a result there is limited high quality evidence to inform surgeons succinctly which perioperative variables influence postoperative recurrence. This systematic review aimed to identify predictors of ventral hernia recurrence. METHODS: PubMed was searched for studies reporting prognostic data of ventral hernia recurrence between 1 January 1995 and 1 January 2018. Extracted data described hernia type (primary/incisional), definitions of recurrence, methods used to detect recurrence, duration of follow-up, and co-morbidity. Data were extracted for all potential predictors, estimates and thresholds described. Random-effects meta-analysis was used. Bias was assessed with a modified PROBAST (Prediction model Risk Of Bias ASsessment Tool). RESULTS: Screening of 18 214 abstracts yielded 274 individual studies for inclusion. Hernia recurrence was defined in 66 studies (24.1 per cent), using 41 different unstandardized definitions. Three patient variables (female sex, age 65 years or less, and BMI greater than 25, 30, 35 or 40 kg/m2), five patient co-morbidities (smoking, diabetes, chronic obstructive pulmonary disease, ASA grade III-IV, steroid use), two hernia-related variables (incisional/primary, recurrent/primary), six intraoperative variables (biological mesh, bridged repair, open versus laparoscopic surgery, suture versus mesh repair, onlay/retrorectus, intraperitoneal/retrorectus), and six postoperative variables (any complication, surgical-site occurrence, wound infection, seroma, haematoma, wound dehiscence) were identified as significant prognostic factors for hernia recurrence. CONCLUSION: This study summarized the current evidence base for predicting ventral hernia recurrence. Results should inform best practice and future research.

Reporting of prognostic studies of tumour markers: a review of published articles in relation to REMARK guidelines.

BACKGROUND: Poor reporting compromises the reliability and clinical value of prognostic tumour marker studies. We review articles to assess the reporting of patients and events using REMARK guidelines, at the time of guideline publication. METHODS: We sampled 50 prognostic tumour marker studies from higher impact cancer journals between 2006 and 2007. The inclusion criteria were cancer; focus on single biological tumour marker; survival analysis; multivariable analysis; and not gene array or proteomic data. Articles were assessed for the REMARK profile and other REMARK guideline items. We propose a reporting aid, the REMARK profile, motivated by the CONSORT flowchart. RESULTS: In 50 studies assessed for the REMARK profile, the number of eligible patients (56% of articles), excluded patients (54%) and patients in analyses (98%) was reported. Only 50% of articles reported the number of outcome events. In multivariable analyses, 54% and 30% of articles reported patient and event numbers for all variables. Of the studies, 66% used archival samples, indicating a potentially biased patient selection. Only 36% of studies reported clearly defined outcomes. CONCLUSIONS: Good reporting is critical for the interpretability and clinical applicability of prognostic studies. Current reporting of key information, such as the number of outcome events in all patients and subgroups, is poor. Use of the REMARK profile would greatly improve reporting and enhance prognostic research.