Access to big datasets from e-health records and individual participant data (IPD) meta-analysis is signalling a new advent of external validation studies for clinical prediction models. In this ...article, the authors illustrate novel opportunities for external validation in big, combined datasets, while drawing attention to methodological challenges and reporting issues.
A core outcome set (COS) is a standardised set of outcomes which should be measured and reported, as a minimum, in all effectiveness trials for a specific health area. This will allow results of ...studies to be compared, contrasted and combined as appropriate, as well as ensuring that all trials contribute usable information. The COMET (Core Outcome Measures for Effectiveness Trials) Initiative aims to support the development, reporting and adoption of COS. Central to this is a publically accessible online resource, populated with all available COS. The aim of the review we report here was to identify studies that sought to determine which outcomes or domains to measure in all clinical trials in a specific condition and to describe the methodological techniques used in these studies.
We developed a multi-faceted search strategy for electronic databases (MEDLINE, SCOPUS, and Cochrane Methodology Register). We included studies that sought to determine which outcomes/domains to measure in all clinical trials in a specific condition.
A total of 250 reports relating to 198 studies were judged eligible for inclusion in the review. Studies covered various areas of health, most commonly cancer, rheumatology, neurology, heart and circulation, and dentistry and oral health. A variety of methods have been used to develop COS, including semi-structured discussion, unstructured group discussion, the Delphi Technique, Consensus Development Conference, surveys and Nominal Group Technique. The most common groups involved were clinical experts and non-clinical research experts. Thirty-one (16%) studies reported that the public had been involved in the process. The geographic locations of participants were predominantly North America (n = 164; 83%) and Europe (n = 150; 76%).
This systematic review identified many health areas where a COS has been developed, but also highlights important gaps. It is a further step towards a comprehensive, up-to-date database of COS. In addition, it shows the need for methodological guidance, including how to engage key stakeholder groups, particularly members of the public.
Large differences exist in size at birth and in rates of impaired fetal growth worldwide. The relative effects of nutrition, disease, the environment, and genetics on these differences are often ...debated. In clinical practice, various references are often used to assess fetal growth and newborn size across populations and ethnic origins, whereas international standards for assessing growth in infants and children have been established. In the INTERGROWTH-21(st) Project, our aim was to assess fetal growth and newborn size in eight geographically defined urban populations in which the health and nutrition needs of mothers were met and adequate antenatal care was provided.
For this study, fetal growth and newborn size were measured in two INTERGROWTH-21(st) component studies using prespecified markers and the same methods, equipment, and selection criteria. In the Fetal Growth Longitudinal Study (FGLS), we studied educated, affluent, healthy women, with adequate nutritional status who were at low risk of intrauterine growth restriction. The primary markers of fetal growth were ultrasound measurements of fetal crown-rump length at less than 14 weeks and 0 days of gestation and fetal head circumference from 14 weeks and 0 days to 40 weeks and 0 days of gestation, and birthlength for newborn size. In the concomitant, population-based Newborn Cross-Sectional Study (NCSS), we measured birthlength in all newborn babies from the eight geographically defined urban populations with the same methods, instruments, and staff as in FGLS. From this large NCSS cohort, we selected an FGLS-like subpopulation to match FGLS with the same eligibility criteria.
Between May 14, 2009, and Aug 2, 2013, we enrolled 4607 women in FGLS and 59 137 women in NCSS. From NCSS, 20 486 (34·6%) women met the FGLS eligibility criteria, and constituted the FGLS-like subpopulation. With variance component analysis, only between 1·9% and 3·5% of the total variability in crown-rump length, fetal head circumference, and newborn birthlength could be attributed to between-site differences. With standardised site effect analysis in 16 gestational age windows from 9 weeks and 0 days of gestation to birth for the three measures (128 comparisons), only one was marginally higher than 0·5 SD of the standardised site difference range. Sensitivity analyses, excluding individual populations in turn from the pooling of all-site centiles across gestational ages, showed no noticeable effect on the 3rd, 50th, and 97th centiles derived from the remaining populations. Our populations were consistent at birth with those in the WHO Multicentre Growth Reference Study (MGRS). The mean birthlength for term newborn babies in that study was 49·5 cm (SD 1·9), which was very similar to that in the FGLS cohort (49·4 cm 1·9) and the NCSS derived FGLS-like subpopulation (49·3 cm 1·8).
Fetal growth and newborn length are similar across diverse geographical settings when mothers' nutritional and health needs are met, and environmental constraints on growth are low. The findings for birthlength are in strong agreement with those of the WHO MGRS. These results provide the conceptual frame to create international standards for growth from conception to newborn baby, which will extend the present infant to childhood WHO MGRS standards.
Bill & Melinda Gates Foundation.
Mounting evidence suggests that there is frequently considerable variation in the risk of the outcome of interest in clinical trial populations. These differences in risk will often cause clinically ...important heterogeneity in treatment effects (HTE) across the trial population, such that the balance between treatment risks and benefits may differ substantially between large identifiable patient subgroups; the "average" benefit observed in the summary result may even be non-representative of the treatment effect for a typical patient in the trial. Conventional subgroup analyses, which examine whether specific patient characteristics modify the effects of treatment, are usually unable to detect even large variations in treatment benefit (and harm) across risk groups because they do not account for the fact that patients have multiple characteristics simultaneously that affect the likelihood of treatment benefit. Based upon recent evidence on optimal statistical approaches to assessing HTE, we propose a framework that prioritizes the analysis and reporting of multivariate risk-based HTE and suggests that other subgroup analyses should be explicitly labeled either as primary subgroup analyses (well-motivated by prior evidence and intended to produce clinically actionable results) or secondary (exploratory) subgroup analyses (performed to inform future research). A standardized and transparent approach to HTE assessment and reporting could substantially improve clinical trial utility and interpretability.
Adding, omitting or changing outcomes after a systematic review protocol is published can result in bias because it increases the potential for unacknowledged or post hoc revisions of the planned ...analyses. The main objective of this study was to look for discrepancies between primary outcomes listed in protocols and in the subsequent completed reviews published on the Cochrane Library. A secondary objective was to quantify the risk of bias in a set of meta-analyses where discrepancies between outcome specifications in protocols and reviews were found.
New reviews from three consecutive issues of the Cochrane Library were assessed. For each review, the primary outcome(s) listed in the review protocol and the review itself were identified and review authors were contacted to provide reasons for any discrepancies. Over a fifth (64/288, 22%) of protocol/review pairings were found to contain a discrepancy in at least one outcome measure, of which 48 (75%) were attributable to changes in the primary outcome measure. Where lead authors could recall a reason for the discrepancy in the primary outcome, there was found to be potential bias in nearly a third (8/28, 29%) of these reviews, with changes being made after knowledge of the results from individual trials. Only 4(6%) of the 64 reviews with an outcome discrepancy described the reason for the change in the review, with no acknowledgment of the change in any of the eight reviews containing potentially biased discrepancies. Outcomes that were promoted in the review were more likely to be significant than if there was no discrepancy (relative risk 1.66 95% CI (1.10, 2.49), p = 0.02).
In a review, making changes after seeing the results for included studies can lead to biased and misleading interpretation if the importance of the outcome (primary or secondary) is changed on the basis of those results. Our assessment showed that reasons for discrepancies with the protocol are not reported in the review, demonstrating an under-recognition of the problem. Complete transparency in the reporting of changes in outcome specification is vital; systematic reviewers should ensure that any legitimate changes to outcome specification are reported with reason in the review.
Critics of systematic reviews have argued that these studies often fail to inform clinical decision making because their results are far too general, that the data are sparse, such that findings ...cannot be applied to individual patients or for other decision making. While there is some consensus on methods for investigating statistical and methodological heterogeneity, little attention has been paid to clinical aspects of heterogeneity. Clinical heterogeneity, true effect heterogeneity, can be defined as variability among studies in the participants, the types or timing of outcome measurements, and the intervention characteristics. The objective of this project was to develop recommendations for investigating clinical heterogeneity in systematic reviews.
We used a modified Delphi technique with three phases: (1) pre-meeting item generation; (2) face-to-face consensus meeting in the form of a modified Delphi process; and (3) post-meeting feedback. We identified and invited potential participants with expertise in systematic review methodology, systematic review reporting, or statistical aspects of meta-analyses, or those who published papers on clinical heterogeneity.
Between April and June of 2011, we conducted phone calls with participants. In June 2011 we held the face-to-face focus group meeting in Ann Arbor, Michigan. First, we agreed upon a definition of clinical heterogeneity: Variations in the treatment effect that are due to differences in clinically related characteristics. Next, we discussed and generated recommendations in the following 12 categories related to investigating clinical heterogeneity: the systematic review team, planning investigations, rationale for choice of variables, types of clinical variables, the role of statistical heterogeneity, the use of plotting and visual aids, dealing with outlier studies, the number of investigations or variables, the role of the best evidence synthesis, types of statistical methods, the interpretation of findings, and reporting.
Clinical heterogeneity is common in systematic reviews. Our recommendations can help guide systematic reviewers in conducting valid and reliable investigations of clinical heterogeneity. Findings of these investigations may allow for increased applicability of findings of systematic reviews to the management of individual patients.
Haemorrhage is a common cause of death in trauma patients. Although transfusions are extensively used in the care of bleeding trauma patients, there is uncertainty about the balance of risks and ...benefits and how this balance depends on the baseline risk of death. Our objective was to evaluate the association of red blood cell (RBC) transfusion with mortality according to the predicted risk of death.
A secondary analysis of the CRASH-2 trial (which originally evaluated the effect of tranexamic acid on mortality in trauma patients) was conducted. The trial included 20,127 trauma patients with significant bleeding from 274 hospitals in 40 countries. We evaluated the association of RBC transfusion with mortality in four strata of predicted risk of death: <6%, 6%-20%, 21%-50%, and >50%. For this analysis the exposure considered was RBC transfusion, and the main outcome was death from all causes at 28 days. A total of 10,227 patients (50.8%) received at least one transfusion. We found strong evidence that the association of transfusion with all-cause mortality varied according to the predicted risk of death (p-value for interaction <0.0001). Transfusion was associated with an increase in all-cause mortality among patients with <6% and 6%-20% predicted risk of death (odds ratio OR 5.40, 95% CI 4.08-7.13, p<0.0001, and OR 2.31, 95% CI 1.96-2.73, p<0.0001, respectively), but with a decrease in all-cause mortality in patients with >50% predicted risk of death (OR 0.59, 95% CI 0.47-0.74, p<0.0001). Transfusion was associated with an increase in fatal and non-fatal vascular events (OR 2.58, 95% CI 2.05-3.24, p<0.0001). The risk associated with RBC transfusion was significantly increased for all the predicted risk of death categories, but the relative increase was higher for those with the lowest (<6%) predicted risk of death (p-value for interaction <0.0001). As this was an observational study, the results could have been affected by different types of confounding. In addition, we could not consider haemoglobin in our analysis. In sensitivity analyses, excluding patients who died early; conducting propensity score analysis adjusting by use of platelets, fresh frozen plasma, and cryoprecipitate; and adjusting for country produced results that were similar.
The association of transfusion with all-cause mortality appears to vary according to the predicted risk of death. Transfusion may reduce mortality in patients at high risk of death but increase mortality in those at low risk. The effect of transfusion in low-risk patients should be further tested in a randomised trial.
www.ClinicalTrials.gov NCT01746953.
Abstract Objective To compare different methods to handle treatment when developing a prognostic model that aims to produce accurate probabilities of the outcome of individuals if left untreated. ...Study Design and Setting Simulations were performed based on two normally distributed predictors, a binary outcome, and a binary treatment, mimicking a randomised trial or an observational study. Comparison was made between simply ignoring treatment (SIT), restricting the analytical dataset to untreated individuals (AUT), inverse probability weighting (IPW), and explicit modelling of treatment (MT). Methods were compared in terms of predictive performance of the model and the proportion of incorrect treatment decisions. Results Omitting a genuine predictor of the outcome from the prognostic model decreased model performance, in both an observational study and a randomised trial. In randomised trials, the proportion of incorrect treatment decisions was smaller when applying AUT or MT, compared to SIT and IPW. In observational studies, MT was superior to all other methods regarding the proportion of incorrect treatment decisions. Conclusion If a prognostic model aims to produce correct probabilities of the outcome in the absence of treatment, ignoring treatments that affect that outcome can lead to suboptimal model performance and incorrect treatment decisions. Explicitly modeling treatment is recommended.
Different methods of evaluating diagnostic performance when comparing diagnostic tests may lead to different results. We compared two such approaches, sensitivity and specificity with area under the ...Receiver Operating Characteristic Curve (ROC AUC) for the evaluation of CT colonography for the detection of polyps, either with or without computer assisted detection.
In a multireader multicase study of 10 readers and 107 cases we compared sensitivity and specificity, using radiological reporting of the presence or absence of polyps, to ROC AUC calculated from confidence scores concerning the presence of polyps. Both methods were assessed against a reference standard. Here we focus on five readers, selected to illustrate issues in design and analysis. We compared diagnostic measures within readers, showing that differences in results are due to statistical methods.
Reader performance varied widely depending on whether sensitivity and specificity or ROC AUC was used. There were problems using confidence scores; in assigning scores to all cases; in use of zero scores when no polyps were identified; the bimodal non-normal distribution of scores; fitting ROC curves due to extrapolation beyond the study data; and the undue influence of a few false positive results. Variation due to use of different ROC methods exceeded differences between test results for ROC AUC.
The confidence scores recorded in our study violated many assumptions of ROC AUC methods, rendering these methods inappropriate. The problems we identified will apply to other detection studies using confidence scores. We found sensitivity and specificity were a more reliable and clinically appropriate method to compare diagnostic tests.
The COMET Initiative database is a repository of studies relevant to the development of core outcome sets (COS). Use of the website continues to increase, with more than 16,500 visits in 2014 (36 % ...increase over 2013), 12,257 unique visitors (47 % increase), 9780 new visitors (43 % increase) and a rise in the proportion of visits from outside the UK (8565 visits; 51 % of all visits). By December 2014, a total of 6588 searches had been completed, with 2383 in 2014 alone (11 % increase). The growing awareness of the need for COS is reflected in the website and database usage figures.