BACKGROUND: The inter-reviewer reliability of the risk of bias (RoB) assessment lacked agreement in previous studies. It is important to analyse these disagreements to improve the repeatability of RoB assessment. The objective of the study was to evaluate the frequency and reasons for disagreements in RoB assessments for randomised controlled trials (RCTs) that were included in multiple Cochrane reviews in the field of hypertension.
METHODS: A cross-sectional study was employed. We retrieved any RCTs that had been included in multiple Cochrane reviews in the field of hypertension from ARCHIE. The results of the RoB assessments were extracted, and the distributions of agreements and possible reasons for disagreement were analyzed.
RESULTS: Twenty-six Cochrane reviews were included in this study. A total of 78 RCTs appeared in more than one Cochrane review. The level of agreement ranged from domain to domain. \"Blinding of outcome assessment\" showed a reasonably high level of agreement (94.9%), while \"incomplete outcome data\", \"selective outcome reporting\" and \"other sources of bias\" showed moderate levels of agreement (74.6%, 79.2% and 75.6%, respectively). However, the domains of \"allocation concealment\", \"random sequence generation\" and \"blinding of participants and personnel\" showed low levels of agreement (24.4%, 23.5%, and 47.4%, respectively). In the domains of \"allocation concealment\" and \"blinding of participants and personnel\", the agreement group had higher proportion of publication year ≤ 1996 than the disagreement group (P = 0.008 and P < 0.001, respectively). In the \"blinding of participants and personnel\", the impact factor was higher in the agreement group (P < 0.001). By analyzing the support text, we found that the most likely reason for disagreement was extracting different information from the same RCT.
CONCLUSIONS: For Cochrane reviews in the field of hypertension using the 2011 version of the RoB tool, there was a large disagreement in the RoB assessment. It is suggested that the results of RoB assessments in systematic reviews that used the 2011 version of the RoB tool need to be interpreted with caution. More accurate information from RCTs needs to be collected when we synthesize clinical evidence.

Peritoneal dialysis (PD) and haemodialysis (HD) are two possible modalities for people with kidney failure commencing dialysis. Only a few randomised controlled trials (RCTs) have evaluated PD versus HD. The benefits and harms of the two modalities remain uncertain. This review includes both RCTs and non-randomised studies of interventions (NRSIs).
To evaluate the benefits and harms of PD, compared to HD, in people with kidney failure initiating dialysis.
We searched the Cochrane Kidney and Transplant Register of Studies from 2000 to June 2024 using search terms relevant to this review. Studies in the Register were identified through searches of CENTRAL, MEDLINE, and EMBASE, conference proceedings, the International Clinical Trials Registry Platform (ICTRP) Search Portal, and ClinicalTrials.gov. MEDLINE and EMBASE were searched for NRSIs from 2000 until 28 March 2023.
RCTs and NRSIs evaluating PD compared to HD in people initiating dialysis were eligible.
Two investigators independently assessed if the studies were eligible and then extracted data. Risk of bias was assessed using standard Cochrane methods, and relevant outcomes were extracted for each report. The primary outcome was residual kidney function (RKF). Secondary outcomes included all-cause, cardiovascular and infection-related death, infection, cardiovascular disease, hospitalisation, technique survival, life participation and fatigue.
A total of 153 reports of 84 studies (2 RCTs, 82 NRSIs) were included. Studies varied widely in design (small single-centre studies to international registry analyses) and in the included populations (broad inclusion criteria versus restricted to more specific participants). Additionally, treatment delivery (e.g. automated versus continuous ambulatory PD, HD with catheter versus arteriovenous fistula or graft, in-centre versus home HD) and duration of follow-up varied widely. The two included RCTs were deemed to be at high risk of bias in terms of blinding participants and personnel and blinding outcome assessment for outcomes pertaining to quality of life. However, most other criteria were assessed as low risk of bias for both studies. Although the risk of bias (Newcastle-Ottawa Scale) was generally low for most NRSIs, studies were at risk of selection bias and residual confounding due to the constraints of the observational study design. In children, there may be little or no difference between HD and PD on all-cause death (6 studies, 5752 participants: RR 0.81, 95% CI 0.62 to 1.07; I2 = 28%; low certainty) and cardiovascular death (3 studies, 7073 participants: RR 1.23, 95% CI 0.58 to 2.59; I2 = 29%; low certainty), and was unclear for infection-related death (4 studies, 7451 participants: RR 0.98, 95% CI 0.39 to 2.46; I2 = 56%; very low certainty). In adults, compared with HD, PD had an uncertain effect on RKF (mL/min/1.73 m2) at six months (2 studies, 146 participants: MD 0.90, 95% CI 0.23 to 3.60; I2 = 82%; very low certainty), 12 months (3 studies, 606 participants: MD 1.21, 95% CI -0.01 to 2.43; I2 = 81%; very low certainty) and 24 months (3 studies, 334 participants: MD 0.71, 95% CI -0.02 to 1.48; I2 = 72%; very low certainty). PD had uncertain effects on residual urine volume at 12 months (3 studies, 253 participants: MD 344.10 mL/day, 95% CI 168.70 to 519.49; I2 = 69%; very low certainty). PD may reduce the risk of RKF loss (3 studies, 2834 participants: RR 0.55, 95% CI 0.44 to 0.68; I2 = 17%; low certainty). Compared with HD, PD had uncertain effects on all-cause death (42 studies, 700,093 participants: RR 0.87, 95% CI 0.77 to 0.98; I2 = 99%; very low certainty). In an analysis restricted to RCTs, PD may reduce the risk of all-cause death (2 studies, 1120 participants: RR 0.53, 95% CI 0.32 to 0.86; I2 = 0%; moderate certainty). PD had uncertain effects on both cardiovascular (21 studies, 68,492 participants: RR 0.96, 95% CI 0.78 to 1.19; I2 = 92%) and infection-related death (17 studies, 116,333 participants: RR 0.90, 95% CI 0.57 to 1.42; I2 = 98%) (both very low certainty). Compared with HD, PD had uncertain effects on the number of patients experiencing bacteraemia/bloodstream infection (2 studies, 2582 participants: RR 0.34, 95% CI 0.10 to 1.18; I2 = 68%) and the number of patients experiencing infection episodes (3 studies, 277 participants: RR 1.23, 95% CI 0.93 to 1.62; I2 = 20%) (both very low certainty). PD may reduce the number of bacteraemia/bloodstream infection episodes (2 studies, 2637 participants: RR 0.44, 95% CI 0.27 to 0.71; I2 = 24%; low certainty). Compared with HD; It is uncertain whether PD reduces the risk of acute myocardial infarction (4 studies, 110,850 participants: RR 0.90, 95% CI 0.74 to 1.10; I2 = 55%), coronary artery disease (3 studies, 5826 participants: RR 0.95, 95% CI 0.46 to 1.97; I2 = 62%); ischaemic heart disease (2 studies, 58,374 participants: RR 0.86, 95% CI 0.57 to 1.28; I2 = 95%), congestive heart failure (3 studies, 49,511 participants: RR 1.10, 95% CI 0.54 to 2.21; I2 = 89%) and stroke (4 studies, 102,542 participants: RR 0.94, 95% CI 0.90 to 0.99; I2 = 0%) because of low to very low certainty evidence. Compared with HD, PD had uncertain effects on the number of patients experiencing hospitalisation (4 studies, 3282 participants: RR 0.90, 95% CI 0.62 to 1.30; I2 = 97%) and all-cause hospitalisation events (4 studies, 42,582 participants: RR 1.02, 95% CI 0.81 to 1.29; I2 = 91%) (very low certainty). None of the included studies reported specifically on life participation or fatigue. However, two studies evaluated employment. Compared with HD, PD had uncertain effects on employment at one year (2 studies, 593 participants: RR 0.83, 95% CI 0.20 to 3.43; I2 = 97%; very low certainty).
The comparative effectiveness of PD and HD on the preservation of RKF, all-cause and cause-specific death risk, the incidence of bacteraemia, other vascular complications (e.g. stroke, cardiovascular events) and patient-reported outcomes (e.g. life participation and fatigue) are uncertain, based on data obtained mostly from NRSIs, as only two RCTs were included.

This is an update of a review first published in 2010. Use of topical fluoride has become more common over time. Excessive fluoride consumption from topical fluorides in young children could potentially lead to dental fluorosis in permanent teeth.
To describe the relationship between the use of topical fluorides in young children and the risk of developing dental fluorosis in permanent teeth.
We carried out electronic searches of the Cochrane Oral Health Trials Register, CENTRAL, MEDLINE, Embase, three other databases, and two trials registers. We searched the reference lists of relevant articles. The latest search date was 28 July 2022.
We included randomized controlled trials (RCTs), quasi-RCTs, cohort studies, case-control studies, and cross-sectional surveys comparing fluoride toothpaste, mouth rinses, gels, foams, paint-on solutions, and varnishes to a different fluoride therapy, placebo, or no intervention. Upon the introduction of topical fluorides, the target population was children under six years of age.
We used standard methodological procedures expected by Cochrane and used GRADE to assess the certainty of the evidence. The primary outcome measure was the percentage prevalence of fluorosis in the permanent teeth. Two authors extracted data from all included studies. In cases where both adjusted and unadjusted risk ratios or odds ratios were reported, we used the adjusted value in the meta-analysis.
We included 43 studies: three RCTs, four cohort studies, 10 case-control studies, and 26 cross-sectional surveys. We judged all three RCTs, one cohort study, one case-control study, and six cross-sectional studies to have some concerns for risk of bias. We judged all other observational studies to be at high risk of bias. We grouped the studies into five comparisons. Comparison 1. Age at which children started toothbrushing with fluoride toothpaste Two cohort studies (260 children) provided very uncertain evidence regarding the association between children starting to use fluoride toothpaste for brushing at or before 12 months versus after 12 months and the development of fluorosis (risk ratio (RR) 0.98, 95% confidence interval (CI) 0.81 to 1.18; very low-certainty evidence). Similarly, evidence from one cohort study (3939 children) and two cross-sectional studies (1484 children) provided very uncertain evidence regarding the association between children starting to use fluoride toothpaste for brushing before or after the age of 24 months (RR 0.83, 95% CI 0.61 to 1.13; very low-certainty evidence) or before or after four years (odds ratio (OR) 1.60, 95% CI 0.77 to 3.35; very low-certainty evidence), respectively. Comparison 2. Frequency of toothbrushing with fluoride toothpaste Two case-control studies (258 children) provided very uncertain evidence regarding the association between children brushing less than twice per day versus twice or more per day and the development of fluorosis (OR 1.63, 95% CI 0.81 to 3.28; very low-certainty evidence). Two cross-sectional surveys (1693 children) demonstrated that brushing less than once per day versus once or more per day may be associated with a decrease in the development of fluorosis in children (OR 0.62, 95% CI 0.53 to 0.74; low-certainty evidence). Comparison 3. Amount of fluoride toothpaste used for toothbrushing Two case-control studies (258 children) provided very uncertain evidence regarding the association between children using less than half a brush of toothpaste, versus half or more of the brush, and the development of fluorosis (OR 0.77, 95% CI 0.41 to 1.46; very low-certainty evidence). The evidence from cross-sectional surveys was also very uncertain (OR 0.92, 95% CI 0.66 to 1.28; 3 studies, 2037 children; very low-certainty evidence). Comparison 4. Fluoride concentration in toothpaste There was evidence from two RCTs (1968 children) that lower fluoride concentration in the toothpaste used by children under six years of age likely reduces the risk of developing fluorosis: 550 parts per million (ppm) fluoride versus 1000 ppm (RR 0.75, 95% CI 0.57 to 0.99; moderate-certainty evidence); 440 ppm fluoride versus 1450 ppm (RR 0.72, 95% CI 0.58 to 0.89; moderate-certainty evidence). The age at which the toothbrushing commenced was 24 months and 12 months, respectively. Two case-control studies (258 children) provided very uncertain evidence regarding the association between fluoride concentrations under 1000 ppm, versus concentrations of 1000 ppm or above, and the development of fluorosis (OR 0.89, 95% CI 0.52 to 1.52; very low-certainty evidence). Comparison 5. Age at which topical fluoride varnish was applied There was evidence from one RCT (123 children) that there may be little to no difference between a fluoride varnish application before four years, versus no application, and the development of fluorosis (RR 0.77, 95% CI 0.45 to 1.31; low-certainty evidence). There was low-certainty evidence from two cross-sectional surveys (982 children) that the application of topical fluoride varnish before four years of age may be associated with the development of fluorosis in children (OR 2.18, 95% CI 1.46 to 3.25).
Most evidence identified mild fluorosis as a potential adverse outcome of using topical fluoride at an early age. There is low- to very low-certainty and inconclusive evidence on the risk of having fluorosis in permanent teeth for: when a child starts receiving topical fluoride varnish application; toothbrushing with fluoride toothpaste; the amount of toothpaste used by the child; and the frequency of toothbrushing. Moderate-certainty evidence from RCTs showed that children who brushed with 1000 ppm or more fluoride toothpaste from one to two years of age until five to six years of age probably had an increased chance of developing dental fluorosis in permanent teeth. It is unethical to propose new RCTs to assess the development of dental fluorosis. However, future RCTs focusing on dental caries prevention could record children\'s exposure to topical fluoride sources in early life and evaluate the dental fluorosis in their permanent teeth as a long-term outcome. In the absence of these studies and methods, further research in this area will come from observational studies. Attention needs to be given to the choice of study design, bearing in mind that prospective controlled studies will be less susceptible to bias than retrospective and uncontrolled studies.

Plasma cell-free DNA (cfDNA) fragmentation patterns are emerging directions in cancer liquid biopsy with high translational significance. Conventionally, the cfDNA sequencing reads are aligned to a reference genome to extract their fragmentomic features. In this study, through cfDNA fragmentomics profiling using different reference genomes on the same datasets in parallel, we report systematic biases in such conventional reference-based approaches. The biases in cfDNA fragmentomic features vary among races in a sample-dependent manner and therefore might adversely affect the performances of cancer diagnosis assays across multiple clinical centers. In addition, to circumvent the analytical biases, we develop Freefly, a reference-free approach for cfDNA fragmentomics profiling. Freefly runs ∼60-fold faster than the conventional reference-based approach while generating highly consistent results. Moreover, cfDNA fragmentomic features reported by Freefly can be directly used for cancer diagnosis. Hence, Freefly possesses translational merit toward the rapid and unbiased measurement of cfDNA fragmentomics.

Collaborative efforts in artificial intelligence (AI) are increasingly common between high-income countries (HICs) and low- to middle-income countries (LMICs). Given the resource limitations often encountered by LMICs, collaboration becomes crucial for pooling resources, expertise, and knowledge. Despite the apparent advantages, ensuring the fairness and equity of these collaborative models is essential, especially considering the distinct differences between LMIC and HIC hospitals. In this study, we show that collaborative AI approaches can lead to divergent performance outcomes across HIC and LMIC settings, particularly in the presence of data imbalances. Through a real-world COVID-19 screening case study, we demonstrate that implementing algorithmic-level bias mitigation methods significantly improves outcome fairness between HIC and LMIC sites while maintaining high diagnostic sensitivity. We compare our results against previous benchmarks, utilizing datasets from four independent United Kingdom Hospitals and one Vietnamese hospital, representing HIC and LMIC settings, respectively.

Doubly adaptive biased coin design (DBCD), a response-adaptive randomization scheme, aims to skew subject assignment probabilities based on accrued responses for ethical considerations. Recent years have seen substantial advances in understanding DBCD\'s theoretical properties, assuming correct model specification for the responses. However, concerns have been raised about the impact of model misspecification on its design and analysis. In this paper, we assess the robustness to both design model misspecification and analysis model misspecification under DBCD. On one hand, we confirm that the consistency and asymptotic normality of the allocation proportions can be preserved, even when the responses follow a distribution other than the one imposed by the design model during the implementation of DBCD. On the other hand, we extensively investigate three commonly used linear regression models for estimating and inferring the treatment effect, namely difference-in-means, analysis of covariance (ANCOVA) I, and ANCOVA II. By allowing these regression models to be arbitrarily misspecified, thereby not reflecting the true data generating process, we derive the consistency and asymptotic normality of the treatment effect estimators evaluated from the three models. The asymptotic properties show that the ANCOVA II model, which takes covariate-by-treatment interaction terms into account, yields the most efficient estimator. These results can provide theoretical support for using DBCD in scenarios involving model misspecification, thereby promoting the widespread application of this randomization procedure.

The Asian water tower (AWT) serves as the source of 10 major Asian river systems and supports the lives of ~2 billion people. Obtaining reliable precipitation data over the AWT is a prerequisite for understanding the water cycle within this pivotal region. Here, we quantitatively reveal that the \"observed\" precipitation over the AWT is considerably underestimated in view of observational evidence from three water cycle components, namely, evapotranspiration, runoff, and accumulated snow. We found that three paradoxes appear if the so-called observed precipitation is corrected, namely, actual evapotranspiration exceeding precipitation, unrealistically high runoff coefficients, and accumulated snow water equivalent exceeding contemporaneous precipitation. We then explain the cause of precipitation underestimation from instrumental error caused by wind-induced gauge undercatch and the representativeness error caused by sparse-uneven gauge density and the complexity of local surface conditions. These findings require us to rethink previous results concerning the water cycle, prompting the study to discuss potential solutions.

UNASSIGNED: Individuals from minoritized racial and ethnic backgrounds experience pernicious and pervasive health disparities that have emerged, in part, from clinician bias.
UNASSIGNED: We used a natural language processing approach to examine whether linguistic markers in electronic health record (EHR) notes differ based on the race and ethnicity of the patient. To validate this methodological approach, we also assessed the extent to which clinicians perceive linguistic markers to be indicative of bias.
UNASSIGNED: In this cross-sectional study, we extracted EHR notes for patients who were aged 18 years or older; had more than 5 years of diabetes diagnosis codes; and received care between 2006 and 2014 from family physicians, general internists, or endocrinologists practicing in an urban, academic network of clinics. The race and ethnicity of patients were defined as White non-Hispanic, Black non-Hispanic, or Hispanic or Latino. We hypothesized that Sentiment Analysis and Social Cognition Engine (SEANCE) components (ie, negative adjectives, positive adjectives, joy words, fear and disgust words, politics words, respect words, trust verbs, and well-being words) and mean word count would be indicators of bias if racial differences emerged. We performed linear mixed effects analyses to examine the relationship between the outcomes of interest (the SEANCE components and word count) and patient race and ethnicity, controlling for patient age. To validate this approach, we asked clinicians to indicate the extent to which they thought variation in the use of SEANCE language domains for different racial and ethnic groups was reflective of bias in EHR notes.
UNASSIGNED: We examined EHR notes (n=12,905) of Black non-Hispanic, White non-Hispanic, and Hispanic or Latino patients (n=1562), who were seen by 281 physicians. A total of 27 clinicians participated in the validation study. In terms of bias, participants rated negative adjectives as 8.63 (SD 2.06), fear and disgust words as 8.11 (SD 2.15), and positive adjectives as 7.93 (SD 2.46) on a scale of 1 to 10, with 10 being extremely indicative of bias. Notes for Black non-Hispanic patients contained significantly more negative adjectives (coefficient 0.07, SE 0.02) and significantly more fear and disgust words (coefficient 0.007, SE 0.002) than those for White non-Hispanic patients. The notes for Hispanic or Latino patients included significantly fewer positive adjectives (coefficient -0.02, SE 0.007), trust verbs (coefficient -0.009, SE 0.004), and joy words (coefficient -0.03, SE 0.01) than those for White non-Hispanic patients.
UNASSIGNED: This approach may enable physicians and researchers to identify and mitigate bias in medical interactions, with the goal of reducing health disparities stemming from bias.

Hypertensive patients often do not make the most favorable choices and behaviors for managing disease. Behavioral economics strategies offer new ideas for guiding patients toward health behavior. The scoping review aimed to summarize behavioral economics strategies designed to improve hypertension self-management behaviors. A literature search was conducted in September 2022 using the following electronic databases: Embase, Medline, CINAHL, PsycINFO, Web of Science, Cochrane Library, CNKI, Wan Fang Database for Chinese Periodicals, and CBM-SinoMed. We screened the literature for experimental studies written in Chinese or English reporting on BE strategies designed to improve self-management behavior in hypertension. We searched 17 820 records and included 18 articles in the final scoping review. We performed qualitative synthesis by the categories of choice architecture. The most common BE strategies were those targeting decision information and decision assistance, such as changing the presentation of information, making information visible, and providing reminders for actions. Most strategies targeted BP, diet, medication adherence, and physical activity behavior. Ten out of 18 studies reported statistically significant improvement in self-management behavior. Further research on BE strategies should focus on addressing the challenges, including changing the decision structure, encompassing a more comprehensive range of target behaviors, and examining the long-term effects of BE strategies.
Self-management of hypertension is a long-term effort, but people often make bounded rational decisions and act in ways that deviate from health goals. Behavioral Economics (BE) strategies make small changes in the decision-making environment to alter choices, steering individuals’ behavior consistent with their goals or preferences. We summarized the BE strategies to improve self-management behavior in hypertension and described the study results using the categories of choice architecture. We found that the most widely used BE strategy is changing the presentation of information in the decision-making environment. Most BE strategies positively affect the target behaviors, which have the potential of BE strategies to enhance self-management behavior for hypertension. Further research is needed to identify the origins of these strategies, modify decision-making structures, and incorporate a broader range of health behaviors to showcase the practicality and sustainability of implementing BE strategies.

UNASSIGNED: Large language models (LLMs) may facilitate the labor-intensive process of systematic reviews. However, the exact methods and reliability remain uncertain.
UNASSIGNED: To explore the feasibility and reliability of using LLMs to assess risk of bias (ROB) in randomized clinical trials (RCTs).
UNASSIGNED: A survey study was conducted between August 10, 2023, and October 30, 2023. Thirty RCTs were selected from published systematic reviews.
UNASSIGNED: A structured prompt was developed to guide ChatGPT (LLM 1) and Claude (LLM 2) in assessing the ROB in these RCTs using a modified version of the Cochrane ROB tool developed by the CLARITY group at McMaster University. Each RCT was assessed twice by both models, and the results were documented. The results were compared with an assessment by 3 experts, which was considered a criterion standard. Correct assessment rates, sensitivity, specificity, and F1 scores were calculated to reflect accuracy, both overall and for each domain of the Cochrane ROB tool; consistent assessment rates and Cohen κ were calculated to gauge consistency; and assessment time was calculated to measure efficiency. Performance between the 2 models was compared using risk differences.
UNASSIGNED: Both models demonstrated high correct assessment rates. LLM 1 reached a mean correct assessment rate of 84.5% (95% CI, 81.5%-87.3%), and LLM 2 reached a significantly higher rate of 89.5% (95% CI, 87.0%-91.8%). The risk difference between the 2 models was 0.05 (95% CI, 0.01-0.09). In most domains, domain-specific correct rates were around 80% to 90%; however, sensitivity below 0.80 was observed in domains 1 (random sequence generation), 2 (allocation concealment), and 6 (other concerns). Domains 4 (missing outcome data), 5 (selective outcome reporting), and 6 had F1 scores below 0.50. The consistent rates between the 2 assessments were 84.0% for LLM 1 and 87.3% for LLM 2. LLM 1\'s κ exceeded 0.80 in 7 and LLM 2\'s in 8 domains. The mean (SD) time needed for assessment was 77 (16) seconds for LLM 1 and 53 (12) seconds for LLM 2.
UNASSIGNED: In this survey study of applying LLMs for ROB assessment, LLM 1 and LLM 2 demonstrated substantial accuracy and consistency in evaluating RCTs, suggesting their potential as supportive tools in systematic review processes.