BMC Medical Research Methodology最新文献

Background: Latent class models can be used to estimate diagnostic accuracy without a gold standard test. Early studies often assumed independence between tests given the true disease state, however this can lead to biased estimates when there are inter-test dependencies. Residual correlation plots and chi-squared statistics have been commonly utilized to assess the validity of the conditional independence assumption and, when it does not hold, identify which test pairs are conditionally dependent. We aimed to assess the performance of these tools with a simulation study covering a wide range of scenarios.

Methods: We generated data sets from a model with four tests and a dependence between tests 1 and 2 within the diseased group. We varied sample size, prevalence, covariance, sensitivity and specificity, with 504 combinations of these in total, and 1000 data sets for each combination. We fitted the conditional independence model in a Bayesian framework, and reported absolute bias, coverage, and how often the residual correlation plots, $G^2$ and ${\chi }^2$ statistics indicated lack-of-fit globally or for each test pair.

Results: Across all settings, residual correlation plots, pairwise $G^2$ and ${\chi }^2$ detected the correct correlated pair of tests only 12.1%, 10.3%, and 10.3% of the time, respectively, but incorrectly suggested dependence between tests 3 and 4 64.9%, 49.7%, and 49.5% of the time. We observed some variation in this across parameter settings, with these tools appearing to perform more as intended when tests 3 and 4 were both much more accurate than tests 1 and 2. Residual correlation plots, $G^2$ and ${\chi }^2$ statistics identified a lack of overall fit in 74.3%, 64.5% and 67.5% of models, respectively. The conditional independence model tended to overestimate the sensitivities of the correlated tests (median bias across all scenarios 0.094, 2.5th and 97.5th percentiles -0.003, 0.397) and underestimate prevalence and the specificities of the uncorrelated tests.

Conclusions: Residual correlation plots and chi-squared statistics cannot be relied upon to identify which tests are conditionally dependent, and also have relatively low power to detect lack of overall fit. This is important since failure to account for conditional dependence can lead to highly biased parameter estimates.

Background: Monitoring of somatic development through the assessment of anthropometric variables such as weight, height, and BMI is vital for evaluating the physical development and nutritional status of children. This approach aids in the early identification of somatic developmental disorders, enabling timely medical interventions. It traditionally relies on Z-scores, which compare anthropometric variables with reference standards. In addition to somatic development monitoring, the early detection and management of pediatric and adolescent hypertension are crucial due to potential long-term health risks. However, manual calculations of Z-scores are time-consuming and error-prone, impeding timely interventions for at-risk children. This article introduces an innovative open-code solution for real-time Z-score assessments directly within the electronic data capture platform, Research Electronic Data Capture, (REDCap™), aiming to streamline the monitoring of somatic development in children.

Methods: Leveraging the World Health Organization (WHO) growth standards and National Health and Nutrition Examination Survey (NHANES) references, our approach integrates Z-score computations directly into REDCap, providing a secure and user-friendly environment for healthcare professionals and researchers. We employed Bland-Altman analyses to compare our method with established calculators (Knirps and Growth XP™) using synthetic data values for all variables.

Results: Bland-Altman plots demonstrated strong agreement between our REDCap calculations and the Knirps and Growth XP systems. Z-scores for height, BMI, and blood pressure consistently aligned, affirming the accuracy of our approach across the measurement range.

Conclusion: The integration with REDCap streamlines data collection and analysis, eliminating the need for separate software and data exports. Moreover, our solution uses the World Health Organization (WHO) growth standards and National Health and Nutrition Examination Survey (NHANES) references. This not only ensures calculation accuracy but also enhances its suitability for diverse research contexts. The Bland-Altman analyses establish the reliability of our method, contributing to a more effective approach to child growth and blood pressure monitoring.

Background: Longitudinal cohort studies are pivotal to understand how socioeconomic, environmental, biological, and lifestyle factors influence health and disease. The added value of cohort studies increases as they accumulate life course data and expand across generations. Ensuring that participants stay motivated to contribute over decades of follow-up is, however, challenging. Participant engagement and involvement (PEI) aims to secure the long-term commitment of participants and promote researcher-participant interaction. This study explored PEI practices in a selection of pregnancy and birth, twin, and family-based population cohort studies.

Methods: Purposive sampling was used to identify cohorts in Europe, Australia and New Zealand. Fourteen semi-structured digital interviews were conducted with cohort study representatives to explore strategies for participant recruitment, informed consent, communication of general and individual information to participants, data collection, and participant involvement. Experiences, resources allocated to PEI, and reflections on future PEI, were discussed. The interview data were analyzed using a content analysis approach and summary results were reviewed and discussed by the representatives.

Results: The cohort studies used various strategies to recruit participants including support from health professionals and organizations combined with information on the studies' web sites and social media. New approaches such as intra-cohort recruitment, were being investigated. Most cohorts transitioned from paper-based to digital solutions to collect the participants' consent and data. While digital solutions increased efficiency, they also brought new challenges. The studies experimented with the use of participant advisory panels and focus groups to involve participants in making decisions, although their success varied across age and socio-economic background. Most representatives reported PEI resources to be limited and called for more human, technical, educational and financial resources to maximize the positive effects of PEI.

Conclusions: This study of PEI among well-established cohort studies underscores the importance of PEI for project sustainability and highlights key factors to consider in developing PEI. Our analysis shows that knowledge gaps exist regarding which approaches have highest impact on retention rates and are best suited for different participant groups. Research is needed to support the development of best practices for PEI as well as knowledge exchange between cohorts through network building.

Background: This study aims to analyse the effects of reducing Received Dose Intensity (RDI) in chemotherapy treatment for osteosarcoma patients on their survival by using a novel approach. Previous research has highlighted discrepancies between planned and actual RDI, even among patients randomized to the same treatment regimen. To mitigate toxic side effects, treatment adjustments, such as dose reduction or delayed courses, are necessary. Toxicities are therefore risk factors for mortality and predictors of future exposure levels. Toxicity introduces post-assignment confounding when assessing the causal effect of chemotherapy RDI on survival outcomes, a topic of ongoing debate.

Methods: Chemotherapy administration data from BO03 and BO06 Randomized Clinical Trials (RCTs) in ostosarcoma are employed to emulate a target trial with three RDI-based exposure strategies: 1) standard, 2) reduced, and 3) highly-reduced RDI. Investigations are conducted between subgroups of patients characterised by poor or good Histological Responses (HRe), i.e., the strongest known prognostic factor for survival in osteosarcoma. Inverse Probability of Treatment Weighting (IPTW) is first used to transform the original population into a pseudo-population which mimics the target randomized cohort. Then, a Marginal Structural Cox Model with effect modification is employed. Conditional Average Treatment Effects (CATEs) are ultimately measured as the difference between the Restricted Mean Survival Time of reduced/highly-reduced RDI strategy and the standard one. Confidence Intervals for CATEs are obtained using a novel IPTW-based bootstrap procedure.

Results: Significant effect modifications based on HRe were found. Increasing RDI-reductions led to contrasting trends for poor and good responders: the higher the reduction, the better (worsen) was the survival in poor (good) reponders. Due to their intrinsic resistance to chemotherapy, poor reponders could benefit from reduced RDI, with an average gain of 10.2 and 15.4 months at 5-year for reduced and highly-reduced exposures, respectively.

Conclusions: This study introduces a novel approach to (i) comprehensively address the challenges related to the analysis of chemotherapy data, (ii) mitigate the toxicity-treatment-adjustment bias, and (iii) repurpose existing RCT data for retrospective analyses extending beyond the original trials' intended scopes.

Background: Soil-transmitted helminthiasis (STH) are a parasitic infection that predominantly affects impoverished regions. Model-based geostatistics (MBG) has been established as a set of modern statistical methods that enable mapping of disease risk in a geographical area of interest. We investigate how the use of remotely sensed covariates can help to improve the predictive inferences on STH prevalence using MBG methods. In particular, we focus on how the covariates impact on the classification of areas into distinct class of STH prevalence.

Methods: This study uses secondary data obtained from a sample of 1551 schools in Kenya, gathered through a combination of longitudinal and cross-sectional surveys. We compare the performance of two geostatistical models: one that does not make use of any spatially referenced covariate; and a second model that uses remotely sensed covariates to assist STH prevalence prediction. We also carry out a simulation study in which we compare the performance of the two models in the classifications of areal units with varying sample sizes and prevalence levels.

Results: The model with covariates generated lower levels of uncertainty and was able to classify 88 more districts into prevalence classes than the model without covariates, which instead left those as "unclassified". The simulation study showed that the model with covariates also yielded a higher proportion of correct classification of at least 40% for all sub-counties.

Conclusion: Covariates can substantially reduce the uncertainty of the predictive inference generated from geostatistical models. Using covariates can thus contribute to the design of more effective STH control strategies by reducing sample sizes without compromising the predictive performance of geostatistical models.

A research protocol is a document that outlines the proposed research idea and is submitted to funding agencies, institutions, or journals for approval. Writing a research protocol represents a challenge, particularly for early-career researchers. In this guide, we aim to provide detailed guidance with the key components and offer practical tips for crafting a research protocol in line with the various study designs. Specifically, the structure of a research protocol should contain the following items: (1) a title that is specific, catchy, and impressive within the word limitation; (2) an abstract that briefs the critical points of the study; (3) an introduction highlighting the study context from broad to narrow and defining the knowledge gap; (4) a justification underlining the significance of the proposed study; (5) Specific, Measurable, Attainable, Relevant, and Time-bound (SMART) objective(s) and aim(s); (6) a methodology covering seven sub-items, including [i] study design and settings, [ii] study subjects, [iii] sample size calculation and sampling, [iv] participants recruitment and follow-up, [v] questionnaire development, [vi] potential variables and outcomes, and [vii] data analysis plan; (7) dissemination of the results; (8) ethics and conflict of interests; (9) budgets analysis/ funding disclosure; and (10) references. This guide will give an overview of these steps and provide clear and concise tips on how to successfully draft a scientific protocol. With careful planning and appropriate guidance, it is possible to develop a well-structured and compelling protocol to obtain approval for the conduction of the study or funding from agencies, institutions, or organizations.

Background: Whether machine learning approaches are superior to classical statistical models for survival analyses, especially in the case of lack of proportionality, is unknown.

Objectives: To compare model performance and predictive accuracy of classic regressions and machine learning approaches using data from the Inspiring Families programme.

Methods: The Inspiring Families programme aims to support members of families with complex issues to return to work. We explored predictors of time to return to work with proportional hazards (Semi-Parametric Cox in Stata) and (Flexible Parametric Parmar-Royston in Stata) against the Survival penalised regression with Elastic Net penalty (scikit-survival), (conditional) Survival Forest algorithm (pySurvival), and (kernel) Survival Support Vector Machine (pySurvival).

Results: At baseline we obtained data on 61 binary variables from all 3161 participants. No model appeared superior, with a low predictive power (concordance index between 0.51 and 0.61). The median time for finding the first job was about 254 days. The top five contributing variables were 'family issues and additional barriers', 'restriction of hours', 'available CV', 'self-employment considered' and 'education'. The Harrell's Concordance index was range from 0.60 (Cox model) to 0.71 (Random Survival Forest) suggesting a better fit for the machine learning approaches. However, the comparison for predicting median time on a selected scenario based showed only minor differences.

Conclusion: Implementing a series of survival models with and without proportional hazards background provides a useful insight as well as better interpretation of the coefficients affected by non-linearities. However, that better fit does not translate to substantially higher predictive power and accuracy from using machine learning approaches. Further tuning of the machine learning algorithms may provide improved results.

Background: The hazard ratio of the Cox proportional hazards model is widely used in randomized controlled trials to assess treatment effects. However, two properties of the hazard ratio including the non-collapsibility and built-in selection bias need to be further investigated.

Methods: We conduct simulations to differentiate the non-collapsibility effect and built-in selection bias from the difference between the marginal and the conditional hazard ratio. Meanwhile, we explore the performance of the Cox model with inverse probability of treatment weighting for covariate adjustment when estimating the marginal hazard ratio. The built-in selection bias is further assessed in the period-specific hazard ratio.

Results: The conditional hazard ratio is a biased estimate of the marginal effect due to the non-collapsibility property. In contrast, the hazard ratio estimated from the inverse probability of treatment weighting Cox model provides an unbiased estimate of the true marginal hazard ratio. The built-in selection bias only manifests in the period-specific hazard ratios even when the proportional hazards assumption is satisfied. The Cox model with inverse probability of treatment weighting can be used to account for confounding bias and provide an unbiased effect under the randomized controlled trials setting when the parameter of interest is the marginal effect.

Conclusions: We propose that the period-specific hazard ratios should always be avoided due to the profound effects of built-in selection bias.

Background: Time-to-event analysis is associated with methodological complexities. Previous research identified flaws in the reporting of time-to-event analyses in randomized trial publications. These hardships impose challenges for meta-analyses of time-to-event outcomes based on aggregate data. We examined the characteristics, reporting and methods of systematic reviews including such analyses.

Methods: Through a systematic search (02/2017-08/2020), we identified 50 Cochrane Reviews with ≥ 1 meta-analysis based on the hazard ratio (HR) and a corresponding random sample (n = 50) from core clinical journals (Medline; 08/02/2021). Data was extracted in duplicate and included outcome definitions, general and time-to-event specific methods and handling of time-to-event relevant trial characteristics.

Results: The included reviews analyzed 217 time-to-event outcomes (Median: 2; IQR 1-2), most frequently overall survival (41%). Outcome definitions were provided for less than half of time-to-event outcomes (48%). Few reviews specified general methods, e.g., included analysis types (intention-to-treat, per protocol) (35%) and adjustment of effect estimates (12%). Sources that review authors used for retrieval of time-to-event summary data from publications varied substantially. Most frequently reported were direct inclusion of HRs (64%) and reference to established guidance without further specification (46%). Study characteristics important to time-to-event analysis, such as variable follow-up, informative censoring or proportional hazards, were rarely reported. If presented, complementary absolute effect estimates calculated based on the pooled HR were incorrectly calculated (14%) or correct but falsely labeled (11%) in several reviews.

Conclusions: Our findings indicate that limitations in reporting of trial time-to-event analyses translate to the review level as well. Inconsistent reporting of meta-analyses of time-to-event outcomes necessitates additional reporting standards.

Background: Patient-reported outcomes (PROs) play an increasing role in the evaluation of oncology treatments. At the same time, single-arm trials are commonly included in regulatory approval submissions. Because of the high risk of biases, results from single-arm trials require careful interpretation. This benefits from a clearly defined estimand, or target of estimation. In this case study, we demonstrated how the ICH E9 (R1) estimand framework can be implemented in SATs with PRO endpoints.

Methods: For the global quality of life outcome in a real single-arm lung cancer trial, a range of possible estimands was defined. We focused on the choice of the variable of interest and strategies to deal with intercurrent events (death, treatment discontinuation and disease progression). Statistical methods were described for each estimand and the corresponding results on the trial data were shown.

Results: Each intercurrent event handling strategy resulted in its own estimated mean global quality of life over time, with a specific interpretation, suitable for a corresponding clinical research aim. In the setting of this case study, a 'while alive' strategy for death and a 'treatment policy' strategy for non-terminal intercurrent events were deemed aligned with a descriptive research aim to inform clinicians and patients about expected quality of life after the start of treatment.

Conclusions: The results show that decisions made in the estimand framework are not trivial. Trial results and their interpretation strongly depend on the chosen estimand. The estimand framework provides a structure to match a research question with a clear target of estimation, supporting specific clinical decisions. Adherence to this framework can help improve the quality of data collection, analysis and reporting of PROs in SATs, impacting decision making in clinical practice.