In healthcare cost‐effectiveness analysis, probability distributions are typically skewed and missing data are frequent. Bootstrap and multiple imputation are well‐established resampling methods for ...handling skewed and missing data. However, it is not clear how these techniques should be combined. This paper addresses combining multiple imputation and bootstrap to obtain confidence intervals of the mean difference in outcome for two independent treatment groups. We assessed statistical validity and efficiency of 10 candidate methods and applied these methods to a clinical data set. Single imputation nested in the bootstrap percentile method (with added noise to reflect the uncertainty of the imputation) emerged as the method with the best statistical properties. However, this method can require extensive computation times and the lack of standard software makes this method not accessible for a larger group of researchers. Using a standard unpaired t‐test with standard multiple imputation without bootstrap appears to be a robust alternative with acceptable statistical performance for which standard multiple imputation software is available.
Ill‐defined research questions could be particularly problematic in an epidemiological setting where measurements fluctuate over time due to intercurrent events, such as medication use. When a ...research question fails to specify how medication use should be handled methodologically, arbitrary decisions may be made during the analysis phase, which likely leads to a mismatch between the intended question and the performed analysis. The mismatch can result in vastly different or meaningless interpretations of estimated effects. Thus, a research question such as “what is the effect of X on Y?” requires further elaboration, and it should consider whether and how medication use has affected the measurements of interest. In our study, we will discuss how well‐defined questions can be formulated when medication use is involved in observational studies. We will distinguish between a situation where an exposure is affected by medication use and where the outcome of interest is affected by medication use. For each setting, we will give examples of different research questions that could be asked depending on how medication use is considered in the estimand and discuss methodological considerations under each question.
Purpose
In epidemiological research, measurements affected by medication, for example, blood pressure lowered by antihypertensives, are common. Different ways of handling medication are required ...depending on the research questions and whether the affected measurement is the exposure, the outcome, or a confounder. This study aimed to review handling of medication use in observational research.
Methods
PubMed was searched for etiological studies published between 2015 and 2019 in 15 high‐ranked journals from cardiology, diabetes, and epidemiology. We selected studies that analyzed blood pressure, glucose, or lipid measurements (whether exposure, outcome or confounder) by linear or logistic regression. Two reviewers independently recorded how medication use was handled and assessed whether the methods used were in accordance with the research aim. We reported the methods used per variable category (exposure, outcome, confounder).
Results
A total of 127 articles were included. Most studies did not perform any method to account for medication use (exposure 58%, outcome 53%, and confounder 45%). Restriction (exposure 22%, outcome 23%, and confounders 10%), or adjusting for medication use using a binary indicator were also used frequently (exposure: 18%, outcome: 19%, confounder: 45%). No advanced methods were applied. In 60% of studies, the methods' validity could not be judged due to ambiguous reporting of the research aim. Invalid approaches were used in 28% of the studies, mostly when the affected variable was the outcome (36%).
Conclusion
Many studies ambiguously stated the research aim and used invalid methods to handle medication use. Researchers should consider a valid methodological approach based on their research question.
Clinical prediction models are estimated using a sample of limited size from the target population, leading to uncertainty in predictions, even when the model is correctly specified. Generally, not ...all patient profiles are observed uniformly in model development. As a result, sampling uncertainty varies between individual patients' predictions. We aimed to develop an intuitive measure of individual prediction uncertainty. The variance of a patient's prediction can be equated to the variance of the sample mean outcome in n∗$$ {n}_{\ast } $$ hypothetical patients with the same predictor values. This hypothetical sample size n∗$$ {n}_{\ast } $$ can be interpreted as the number of similar patients neff$$ {n}_{\mathrm{eff}} $$ that the prediction is effectively based on, given that the model is correct. For generalized linear models, we derived analytical expressions for the effective sample size. In addition, we illustrated the concept in patients with acute myocardial infarction. In model development, neff$$ {n}_{\mathrm{eff}} $$ can be used to balance accuracy versus uncertainty of predictions. In a validation sample, the distribution of neff$$ {n}_{\mathrm{eff}} $$ indicates which patients were more and less represented in the development data, and whether predictions might be too uncertain for some to be practically meaningful. In a clinical setting, the effective sample size may facilitate communication of uncertainty about predictions. We propose the effective sample size as a clinically interpretable measure of uncertainty in individual predictions. Its implications should be explored further for the development, validation and clinical implementation of prediction models.
Disturbed cognitive function is associated with several causes of mortality; however, the association between cognitive function and the risk of cancer death has not been extensively investigated ...yet. We aimed to evaluate the association of cognitive function with the risk of cancer death and all-cause mortality in the PROspective Study of Pravastatin in the Elderly at Risk (PROSPER) and Leiden 85-plus Study. Additionally, a systematic review and meta-analysis of longitudinal studies were conducted to evaluate the association of cognitive function and risk of cancer death.
Risk of cancer death and all-cause mortality were reported using hazard ratios (HRs) with 95% confidence interval (CI) in tertiles of cognitive function of PROSPER and Leiden85-Plus Study. Additionally, PubMed, Embase, Web of Science, Cochrane, PsycINFO, Academic Search Premier, CINHAL, and Emcare were searched up to November 1st, 2020 to perform a systematic review and meta-analysis. The relative risks (RRs) with 95%CI of cancer death per each standard deviation lower performance in cognitive measurements were calculated.
Participants of PROSPER had 1.65-fold (95%CI 1.11-2.47) greater risk of cancer death (P for trend = 0.016) and 1.85-fold (95%CI 1.46-2.34) higher risk of all-cause mortality (P for trend<0.001), in multivariable models. Results of the Leiden-85 Plus Study showed that subjects with MMSE score below 24 had a lower chance of cancer death (HR 0.79, 95%CI 0.36-1.70, P for trend = 0.820) but had 2.18-fold (95%CI 1.57-3.02) higher risk of all-cause mortality compared to the reference group (P for trend<0.001). Besides, the results of systematic review and meta-analysis showed that per each standard deviation lower performance in cognitive function, individuals were at a 10% higher chance of cancer death (RR 1.10, 95%CI 1.00-1.20, P-value = 0.044).
Lower cognitive function performance is associated with a marginally increased risk of cancer death, in line with a significantly greater risk of all-cause mortality.
Background
Treatment of patients with hemophilia has advanced over the past decades, but it is unknown whether this has resulted in a normal life expectancy in the Netherlands.
Objective
This ...observational cohort study aimed to assess all‐cause and cause‐specific mortality in patients with hemophilia in the Netherlands between 2001 and 2018 and to compare mortality and life expectancy with previous survival assessments from 1973 onward.
Patients/methods
All 1066 patients with hemophilia who participated in a nationwide survey in 2001 were followed until July 2018.
Results
Information on 1031 individuals (97%) was available, of whom 142 (14%) deceased during follow‐up. Compared with the general Dutch male population, mortality of patients with hemophilia was still increased (standardized mortality ratio: 1.4, 95% confidence interval: 1.2–1.7). Intracranial bleeding and malignancies were the most common causes of death. Estimated median life expectancy of patients with hemophilia was 77 years, 6 years lower than the median life expectancy of the general Dutch male population (83 years). Over the past 45 years, death rates of patients with hemophilia have consistently decreased, approaching the survival experience of the general population. Over the past decades, mortality due to human immunodeficiency virus and hepatitis C virus infections has decreased, death due to intracranial hemorrhages has increased, and death due to ischemic heart disease has remained consistently low over time.
Conclusions
Survival in patients with hemophilia in the Netherlands has improved over time but is still lower than that of the general population.
Objective
Hematopoietic stem cell transplantation (HSCT) can be a curative treatment for malignant and nonmalignant diseases in children but is associated with significant late effects including ...growth failure. Growth hormone treatment (GHRx) is offered to improve growth, but limited data are available on its effect on adult height (AH). We aim to evaluate the effectiveness of GHRx.
Design
Single‐center retrospective study.
Patients
Thirty‐four patients who had received GHRx for ≥1 year were matched with two controls each, without GHRx, based on sex, indication for HSCT (malignancy, benign haematological disease or immunodeficiency), age at HSCT and conditioning with/without total body irradiation (TBI). All had reached AH.
Measurements
The primary outcome measure was the difference between AH and predicted AH (PAH) at start of GHRx or the equivalent age in controls (AH−PAH), calculated according to Bailey and Pinneau.
Results
GHRx was started at age 12.0 ± 2.6 years; median treatment duration was 3.8 years (range 1.7−9.2). AH−PAH standard deviation score (SDS) was significantly higher in growth hormone (GH) treated boys (−0.5 ± 0.7 SDS) than in controls (−1.5 ± 1.0 SDS, p < .001). Girls also had a higher AH−PAH after GHRx (+0.5 ± 0.6 SDS) compared to controls (−0.2 SDS ±0.7, p < .01). AH remained approximately 2 SDS below target height (TH) in treated and untreated individuals. Among GH‐treated children, AH−PAH was higher in those who had received busulfan‐based compared to TBI‐based conditioning.
Conclusion
GHRx had a significant positive effect on AH compared to PAH, although AH remained far below TH. Higher AH−PAH was observed in girls and in those conditioned without TBI.
The use of thromboprophylaxis to prevent clinically apparent venous thromboembolism after knee arthroscopy or casting of the lower leg is disputed. We compared the incidence of symptomatic venous ...thromboembolism after these procedures between patients who received anticoagulant therapy and those who received no anticoagulant therapy.
We conducted two parallel, pragmatic, multicenter, randomized, controlled, open-label trials with blinded outcome evaluation: the POT-KAST trial, which included patients undergoing knee arthroscopy, and the POT-CAST trial, which included patients treated with casting of the lower leg. Patients were assigned to receive either a prophylactic dose of low-molecular-weight heparin (for the 8 days after arthroscopy in the POT-KAST trial or during the full period of immobilization due to casting in the POT-CAST trial) or no anticoagulant therapy. The primary outcomes were the cumulative incidences of symptomatic venous thromboembolism and major bleeding within 3 months after the procedure.
In the POT-KAST trial, 1543 patients underwent randomization, of whom 1451 were included in the intention-to-treat population. Venous thromboembolism occurred in 5 of the 731 patients (0.7%) in the treatment group and in 3 of the 720 patients (0.4%) in the control group (relative risk, 1.6; 95% confidence interval CI, 0.4 to 6.8; absolute difference in risk, 0.3 percentage points; 95% CI, -0.6 to 1.2). Major bleeding occurred in 1 patient (0.1%) in the treatment group and in 1 (0.1%) in the control group (absolute difference in risk, 0 percentage points; 95% CI, -0.6 to 0.7). In the POT-CAST trial, 1519 patients underwent randomization, of whom 1435 were included in the intention-to-treat population. Venous thromboembolism occurred in 10 of the 719 patients (1.4%) in the treatment group and in 13 of the 716 patients (1.8%) in the control group (relative risk, 0.8; 95% CI, 0.3 to 1.7; absolute difference in risk, -0.4 percentage points; 95% CI, -1.8 to 1.0). No major bleeding events occurred. In both trials, the most common adverse event was infection.
The results of our trials showed that prophylaxis with low-molecular-weight heparin for the 8 days after knee arthroscopy or during the full period of immobilization due to casting was not effective for the prevention of symptomatic venous thromboembolism. (Funded by the Netherlands Organization for Health Research and Development; POT-KAST and POT-CAST ClinicalTrials.gov numbers, NCT01542723 and NCT01542762 , respectively.).
Objective To assess the effect of provision of free glasses on academic performance in rural Chinese children with myopia.Design Cluster randomized, investigator masked, controlled trial.Setting 252 ...primary schools in two prefectures in western China, 2012-13.Participants 3177 of 19 934 children in fourth and fifth grades (mean age 10.5 years) with visual acuity <6/12 in either eye without glasses correctable to >6/12 with glasses. 3052 (96.0%) completed the study.Interventions Children were randomized by school (84 schools per arm) to one of three interventions at the beginning of the school year: prescription for glasses only (control group), vouchers for free glasses at a local facility, or free glasses provided in class.Main outcome measures Spectacle wear at endline examination and end of year score on a specially designed mathematics test, adjusted for baseline score and expressed in standard deviations.Results Among 3177 eligible children, 1036 (32.6%) were randomized to control, 988 (31.1%) to vouchers, and 1153 (36.3%) to free glasses in class. All eligible children would benefit from glasses, but only 15% wore them at baseline. At closeout glasses wear was 41% (observed) and 68% (self reported) in the free glasses group, and 26% (observed) and 37% (self reported) in the controls. Effect on test score was 0.11 SD (95% confidence interval 0.01 to 0.21) when the free glasses group was compared with the control group. The adjusted effect of providing free glasses (0.10, 0.002 to 0.19) was greater than parental education (0.03, −0.04 to 0.09) or family wealth (0.01, −0.06 to 0.08). This difference between groups was significant, but was smaller than the prespecified 0.20 SD difference that the study was powered to detect.Conclusions The provision of free glasses to Chinese children with myopia improves children’s performance on mathematics testing to a statistically significant degree, despite imperfect compliance, although the observed difference between groups was smaller than the study was originally designed to detect. Myopia is common and rarely corrected in this setting.Trial Registration Current Controlled Trials ISRCTN03252665.
In this paper we study approaches for dealing with treatment when developing a clinical prediction model. Analogous to the estimand framework recently proposed by the European Medicines Agency for ...clinical trials, we propose a ‘predictimand’ framework of different questions that may be of interest when predicting risk in relation to treatment started after baseline. We provide a formal definition of the estimands matching these questions, give examples of settings in which each is useful and discuss appropriate estimators including their assumptions. We illustrate the impact of the predictimand choice in a dataset of patients with end-stage kidney disease. We argue that clearly defining the estimand is equally important in prediction research as in causal inference.