From complexity to clarity: how directed acyclic graphs enhance the study design of systematic reviews and meta-analyses

While frameworks to systematically assess bias in systematic reviews and meta-analyses (SRMAs) and frameworks on causal inference are well established, they are less frequently integrated beyond the data analysis stages. This paper proposes the use of Directed Acyclic Graphs (DAGs) in the design stage of SRMAs. We hypothesize that DAGs created and registered a priori can offer a useful approach to more effective and efficient evidence synthesis. DAGs provide a visual representation of the complex assumed relationships between variables within and beyond individual studies prior to data analysis, facilitating discussion among researchers, guiding data analysis, and may lead to more targeted inclusion criteria or set of data extraction items. We illustrate this argument through both experimental and observational case examples.

Quality of motivational interviewing matters: the effect on participation in health-promotion activities in a cluster randomized controlled trial

Objectives This study aims to evaluate the effectiveness of a "blended" workplace health-promotion intervention, alongside identification of key components beneficial for future implementation strategies. Methods Within a cluster randomized controlled trial, 491 employees at increased risk of cardiovascular disease were allocated to the limited (N=217; 9 clusters) or extensive (N=274; 8 clusters) intervention. The extensive intervention consisted of motivational interviewing (MI) within the framework of a web-based health risk assessment (HRA), a blended care approach. The limited intervention received solely the web-based HRA. Occupational health physicians (N=21) within three organizations delivered the intervention. Implementation components investigated included: HRA (reach and participation), newsletters (percentage read), and MI sessions (number and quality). MI quality was determined by scoring audiotaped MI sessions, using the MI treatment integrity code. After 6 and 12 months, effects on participation in health-promotion activities and its associations with components of implementation were determined by mixed-effects models. Results Over 80% of employees participated in health-promotion activities, with an additional 8% in the extensive compared to the limited group. In the extensive intervention, those with more or better quality MI sessions were more likely to participate in health-promotion activities. Increased MI quality was associated with sustained participation. Conclusions This study suggests that participation in health-promotion activities can be increased by adding MI to a web-based approach and improving the quality of the MI delivered. Interventions with MI should include optimized delivery and quality of MI sessions.

Evaluation of T2-W MR imaging and diffusion-weighted imaging for the early post-treatment local response assessment of patients treated conservatively for cervical cancer: a multicentre study

Objectives

To compare MR imaging with or without DWI and clinical response evaluation (CRE) in the local control evaluation of cervical carcinoma after radiotherapy.

Methods

In a multicentre university setting, we prospectively included 107 patients with primary cervical cancer treated with radiotherapy. Sensitivity and specificity for CRE and MR imaging (with pre-therapy MR imaging as reference) (2 readers) were evaluated using cautious and strict criteria for identifying residual tumour. Nested logistic regression models were constructed for CRE, subsequently adding MR imaging with and without DWI as independent variables, as well as the pre- to post-treatment change in apparent diffusion coefficient (delta ADC).

Results

Using cautious criteria, CRE and MR imaging with DWI (reader 1/reader 2) have comparable high specificity (83% and 89%/95%, respectively), whereas MR imaging without DWI showed significantly lower specificity (63%/53%) than CRE. Using strict criteria, CRE and MR imaging with DWI both showed very high specificity (99% and 92%/95%, respectively), whereas MR imaging without DWI showed significantly lower specificity (89%/77%) than CRE. All sensitivities were not significantly different. Addition of MR imaging with DWI to CRE has statistically significant incremental value in identifying residual tumour (reader 1: estimate, 1.06; p = 0.001) (reader 2: estimate, 0.62; p = 0.02). Adding the delta ADC did not have significant incremental value in detecting residual tumour.

Conclusions

DWI significantly increases the specificity of MR imaging in the detection of local residual tumour. Furthermore, MR imaging with DWI has significant incremental diagnostic value over CRE, whereas adding the delta ADC has no incremental diagnostic value.

Key Points

• If MR imaging is used for response evaluation, DWI should be incorporated

• MR imaging with DWI has diagnostic value comparable/complementary to clinical response evaluation

• Inter-reader agreement is moderate to fair for two experienced radiologist readers

• Quantitative measurements of ADC early post-therapy have limited diagnostic value

Prediction of intracranial findings on CT-scans by alternative modelling techniques

Background

Prediction rules for intracranial traumatic findings in patients with minor head injury are designed to reduce the use of computed tomography (CT) without missing patients at risk for complications. This study investigates whether alternative modelling techniques might improve the applicability and simplicity of such prediction rules.

Methods

We included 3181 patients with minor head injury who had received CT scans between February 2002 and August 2004. Of these patients 243 (7.6%) had intracranial traumatic findings and 17 (0.5%) underwent neurosurgical intervention. We analyzed sensitivity, specificity and area under the ROC curve (AUC-value) to compare the performance of various modelling techniques by 10 × 10 cross-validation. The techniques included logistic regression, Bayes network, Chi-squared Automatic Interaction Detection (CHAID), neural net, support vector machines, Classification And Regression Trees (CART) and "decision list" models.

Results

The cross-validated performance was best for the logistic regression model (AUC 0.78), followed by the Bayes network model and the neural net model (both AUC 0.74). The other models performed poorly (AUC < 0.70). The advantage of the Bayes network model was that it provided a graphical representation of the relationships between the predictors and the outcome.

Conclusions

No alternative modelling technique outperformed the logistic regression model. However, the Bayes network model had a presentation format which provided more detailed insights into the structure of the prediction problem. The search for methods with good predictive performance and an attractive presentation format should continue.

Diagnostic performance of the platelet function analyzer (PFA-100®) for the detection of disorders of primary haemostasis in patients with a bleeding history–a systematic review and meta-analysis

The Platelet Function Analyzer (PFA-100) is increasingly being used in the workup of patients with a bleeding diathesis. A profound knowledge of the possible diagnostic performance of this test is essential in order to make sound clinical decisions based on its results. It was the aim of this study to systematically review the published literature and provide valid estimates of the diagnostic performance of the PFA-100 for detecting disorders of primary haemostasis in newly presenting patients with a bleeding diathesis. A comprehensive literature search was performed for studies published between January 1994 and February 2006. Studies were eligible for the systematic review if they provided data supposed to be applicable to the determination of the diagnostic performance of the PFA-100. Furthermore, they were included in a meta-analysis if study reporting allowed calculation of sensitivity and specificity and if study quality ensured minimized biases of these estimates for the described clinical setting. Pooled weighted sensitivity, specificity and diagnostic odds ratio were calculated applying random effects modelling and constructing summary operator characteristic curves. This was done separately for the available test modifications using either collagen/epinephrine (PFA-EPI) or collagen/adenosine-diphosphate (PFA-ADP) for platelet activation. Thirty-six articles were included in the systematic review. Six studies met our eligibility criteria for a meta-analysis. The major reason for exclusion from the meta-analysis was a case-control design. A total of 1486 and 1259 patients were included in the meta-analysis of the diagnostic performance of the PFA-EPI and PFA-ADP, respectively. Pooled weighted sensitivity and specificity of the PFA-EPI/PFA-ADP in detecting a disorder of primary haemostasis were: 82.5/66.9% (95%-confidence interval (95%-CI): 76.0-88.9%/57.9-75.9%), and 88.7/85.5% (95%-CI: 84.3-93.1%/82.0-89.1%). 83/75% of patients with a positive PFA-EPI/PFA-ADP result do have a disorder of primary haemostasis whereas 88/79% with a negative PFA-EPI/PFA-ADP result do not. The PFA-EPI appeared to have a higher sensitivity and better predictive values than the PFA-ADP in detecting disorders of primary haemostasis, although a rigorous gold standard definition for a disorder of primary haemostasis, particularly for platelet disorders, was not applied in most studies. The majority of the studies lacked important requirements for quality and reporting, precluding a more precise and definitive characterization of the clinical utility of the PFA-100. This emphasizes the need for an evidence-based critical appraisal of diagnostic studies in haemostasis research in order to promote the conducting of studies that produce clinically relevant results.

The use of correspondence analysis in the assessment of morphologic changes during carcinogenesis

Correspondence analysis is a statistical technique originally developed for the analysis of frequency and contingency tables. It may also be successfully applied to the problem of the assessment of the characteristics of samples of cytological material. The present article describes the use of this formalism on feature data measured on urothelial cells in populations obtained from rats at different time points after exposition to a carcinogenic agent. After the application of correspondence analysis, the projections of the samples corresponding to the different time points onto the plane spanned by the first two factorial axes constitute a trajectory, preserving the time ordering. The locations of the projections of cell samples of untreated animals (controls) with respect to this trajectory suggest that two different processes inducing different morphological changes are at work: Early abnormal samples are characterized by a lower-than-normal value of nuclear stain content, whereas late abnormals are mainly characterized by a higher-than-normal nuclear stain value. It is shown that populations of selected atypical cells show less variation throughout the carcinogenic process than populations of randomly selected cells.