Multiple Imputation Methods Research Articles

Effect of Spesolimab on Achieving Sustained Disease Remission in Patients with Generalized Pustular Psoriasis: Results from the Effisayil 2 Study

Introduction: Generalized pustular psoriasis (GPP) is a chronic, rare and potentially life-threatening disease characterized by flares of widespread skin pustulation. Sustained skin clearance is a key treatment objective. Intravenous (IV) spesolimab, an anti-interleukin-36 receptor monoclonal antibody, is approved for GPP flare treatment; however, the optimal dosing and long-term efficacy of subcutaneous (SC) spesolimab treatment in patients with GPP has not yet been reported. We report the effect of SC spesolimab on achieving sustained disease remission over 48 weeks in Effisayil 2 (NCT04399837), a trial that evaluated the efficacy and safety of SC spesolimab in preventing GPP flares.
 Methods: Patients with a history of GPP were randomized (1:1:1:1) to receive placebo (n=31), low- (n=31, 300 mg loading dose [LD]; 150 mg every 12 weeks [q12w]), medium- (n=31, 600 mg LD; 300 mg q12w) or high-dose (n=30, 600 mg LD; 300 mg every 4 weeks [q4w]) SC spesolimab over 48 weeks. Sustained remission was defined as GPP Physician Global Assessment (GPPGA) total score of 0 or 1 at all visits up to Week 48, or GPPGA total score of 0 or 1 and all GPPGA subscores ≤2 at all visits up to Week 48. Sustained pustular clearance was defined as GPPGA pustulation subscore of 0 at all visits from Week 4 to Week 48. Any use of IV spesolimab or another standard of care for GPP worsening was considered a failure of remission. Missing data were imputed by a sequential logistic regression multiple imputation method. Adjusted risk differences (RD) were calculated by the Mantel−Haenszel type-weighted average of differences.
 Results: 31 patients received placebo, and 30 patients received high-dose spesolimab. Compared to placebo, more patients had sustained remission in the high-dose arm using both GPPGA total score 0 or 1 (63.3% vs 29.0%; RD [95% CI] 0.35 [0.10–0.59]), and GPPGA total score 0 or 1 and all GPPGA subscores ≤2 (60.4% vs 29.0%; RD [95% CI] 0.32 [0.07–0.56]). Compared to placebo, more patients in the high-dose arm had sustained pustular clearance over 48 weeks (63.6% vs 25.8%; RD [95% CI] 0.38 [0.14–0.62]).
 Conclusion: Overall, high-dose SC spesolimab q4w is effective for the long-term management of GPP skin symptoms.

Abstract 16725: Cost-Effectiveness Analysis of Intensive Treatment of Systolic Hypertension: Results From Sprint Patient-Level Data

Background: The Systolic Blood Pressure Intervention Trial (SPRINT) showed that treating patients at high risk for cardiovascular events but without diabetes or a history of stroke to a systolic blood pressure (SBP) target goal of <120 mm Hg (Intensive Care, IC), compared with <140 mm Hg (Standard Care, SC), resulted in lower rates of fatal and nonfatal major cardiovascular events and death. We examined the cost-effectiveness of IC vs SC over a life-time horizon, based on patient-level data, from a health-sector perspective. Methods: Patient-level data from SPRINT were used to calculate and estimate cost, effectiveness, quality adjusted life years (QALY), and the incremental cost-effectiveness ratio (ICER). For participants who died during the trial, life years and QALYs were estimated from in-trial analysis; for who survived to the end of the trial, pooled, harmonized data from six epidemiologic cohorts included in the National Heart, Lung, and Blood Institute Pooled Cohorts Study was used to estimate risk of lifetime primary fatal and non-fatal events and calculate life-years. Published utilities estimates were used to calculate QALYs. For participants without SPRINT lifetime primary event estimates, we applied multiple imputation method via machine learning to impute the life-years and QALYs. The ICER was expressed as cost per QALY gained. Results: IC participants were estimated to gain an average of 0.51 QALY (95% CI: 0.21, 0.81) relative to SC participants. Estimated total lifetime costs were $17,449 higher with IC (95% CI: $10,697, $24,201) than SC. The ICER for IC vs SC was $34,214 per QALY gained, with 68.5% of bootstrap-derived estimates <$50,000/QALY. The results from multiple imputation chained equations for addressing missing data showed that IC is cost effectiveness at $50,000/QALY. Across various scenario analyses, we consistently found a more than 60% probability that IC is cost-effective over a lifetime when the willingness-to-pay value exceeds $100,000 per QALY. Conclusions: Based on participant-level trial and epidemiologic cohort data, IC is cost-effective, with significant clinical benefit at acceptable higher lifetime cost. Recommending intensive SBP targets in patients at high CVD risk is an effective, high-value therapy.

A Machine Learning-Based Multiple Imputation Method for the Health and Aging Brain Study–Health Disparities

The Health and Aging Brain Study–Health Disparities (HABS–HD) project seeks to understand the biological, social, and environmental factors that impact brain aging among diverse communities. A common issue for HABS–HD is missing data. It is impossible to achieve accurate machine learning (ML) if data contain missing values. Therefore, developing a new imputation methodology has become an urgent task for HABS–HD. The three missing data assumptions, (1) missing completely at random (MCAR), (2) missing at random (MAR), and (3) missing not at random (MNAR), necessitate distinct imputation approaches for each mechanism of missingness. Several popular imputation methods, including listwise deletion, min, mean, predictive mean matching (PMM), classification and regression trees (CART), and missForest, may result in biased outcomes and reduced statistical power when applied to downstream analyses such as testing hypotheses related to clinical variables or utilizing machine learning to predict AD or MCI. Moreover, these commonly used imputation techniques can produce unreliable estimates of missing values if they do not account for the missingness mechanisms or if there is an inconsistency between the imputation method and the missing data mechanism in HABS–HD. Therefore, we proposed a three-step workflow to handle missing data in HABS–HD: (1) missing data evaluation, (2) imputation, and (3) imputation evaluation. First, we explored the missingness in HABS–HD. Then, we developed a machine learning-based multiple imputation method (MLMI) for imputing missing values. We built four ML-based imputation models (support vector machine (SVM), random forest (RF), extreme gradient boosting (XGB), and lasso and elastic-net regularized generalized linear model (GLMNET)) and adapted the four ML-based models to multiple imputations using the simple averaging method. Lastly, we evaluated and compared MLMI with other common methods. Our results showed that the three-step workflow worked well for handling missing values in HABS–HD and the ML-based multiple imputation method outperformed other common methods in terms of prediction performance and change in distribution and correlation. The choice of missing handling methodology has a significant impact on the accompanying statistical analyses of HABS–HD. The conceptual three-step workflow and the ML-based multiple imputation method perform well for our Alzheimer’s disease models. They can also be applied to other disease data analyses.

Correlation between malignant peritoneal cytology and survival in patients with uterine leiomyosarcoma and endometrial stromal sarcoma

ObjectiveThis study aimed to examine the correlation between malignant peritoneal cytology and overall survival among patients with uterine leiomyosarcoma and endometrial stromal sarcoma.MethodsPatients with uterine leiomyosarcoma and endometrial stromal sarcoma...

Cost-utility analysis and cross-country comparison of pharmacogenomics-guided treatment in colorectal cancer patients participating in the U-PGx PREPARE study

ObjectivesA cost-utility analysis was conducted to evaluate pharmacogenomic (PGx)-guided treatment compared to the standard-of-care intervention among patients diagnosed with colorectal cancer (CRC) in Italy. MethodsData derived from a prospective, open-label, block randomized clinical trial, as a part of the largest PGx study worldwide (355 patients in both arms) were used. Mortality was used as the primary health outcome to estimate life years (LYs) gained in treatment arms within a survival analysis context. PGx-guided treatment was based on established drug-gene interactions between capecitabine, 5-fluorouracil and irinotecan with DPYD and/or UGT1A1 genomic variants. Utility values for the calculation of Quality Adjusted Life Year (QALY) was based on Visual Analog Scale (VAS) score. Missing data were imputed via the Multiple Imputation method and linear interpolation, when possible, while censored cost data were corrected via the Replace-From-The-Right algorithm. The Incremental Cost-Effectiveness Ratio (ICER) was calculated for QALYs. Raw data were bootstrapped 5000 times in order to produce 95% Confidence Intervals based on non-parametric percentile method and to construct a cost-effectiveness acceptability curve. Cost differences for study groups were investigated via a generalized linear regression model analysis. Total therapy cost per patient reflected all resources expended in the management of any adverse events, including medications, diagnostics tests, devices, surgeries, the utilization of intensive care units, and wards. ResultsThe total cost of the study arm was estimated at €380 (∼ US$416; 95%CI: 195–596) compared to €565 (∼ US$655; 95%CI: 340–724) of control arm while the mean survival in study arm was estimated at 1.58 (+0.25) LYs vs 1.50 (+0.26) (Log Rank test, X2 = 4.219, df=1, p-value=0.04). No statistically significant difference was found in QALYs. ICER was estimated at €13418 (∼ US$14695) per QALY, while the acceptability curve indicated that when the willingness-to-pay was under €5000 (∼ US$5476), the probability of PGx being cost-effective overcame 70%. The most frequent adverse drug event in both groups was neutropenia of severity grade 3 and 4, accounting for 82.6% of total events in the study arm and 65.0% in the control arm. Apart from study arm, smoking status, Body-Mass-Index and Cumulative Actionability were also significant predictors of total cost. Subgroup analysis conducted in actionable patients (7.9% of total patients) indicated that PGx-guided treatment was a dominant option over its comparator with a probability greater than 92%. In addition, a critical literature review was conducted, and these findings are in line with those reported in other European countries. ConclusionPGx-guided treatment strategy may represent a cost-saving option compared to the existing conventional therapeutic approach for colorectal cancer patient management in the National Health Service of Italy.

Estimation of zero-inflated bivariate Poisson regression with missing covariates

Zero-inflated bivariate Poisson regression (ZIBP) models are most often applied to correlated bivariate count data that contain a large proportion of zeros. These models have been applied in various fields, such as medical research, health economics, insurance, sports, etc. Because, in practice, some of the covariates involved in ZIBP regression modeling often have missing values, we propose methods for estimating the parameters of the ZIBP regression model with missing at random covariates. Assuming that the selection probability is unknown and estimated non parametrically (by a kernel estimator), we propose inverse probability weighting (IPW) and multiple imputation (MI) methods for estimating the parameters of the ZIBP regression model with missing at random covariates. Asymptotic properties of the proposed estimators are studied under certain regularity conditions. A simulation study is conducted to evaluate the performance of the proposed methods. Finally, the practical application of the proposed methodology is illustrated with data from the health-care field.

High-Dimensional Imputation for the Social Sciences: A Comparison of State-of-The-Art Methods

Including a large number of predictors in the imputation model underlying a multiple imputation (MI) procedure is one of the most challenging tasks imputers face. A variety of high-dimensional MI techniques can help, but there has been limited research on their relative performance. In this study, we investigated a wide range of extant high-dimensional MI techniques that can handle a large number of predictors in the imputation models and general missing data patterns. We assessed the relative performance of seven high-dimensional MI methods with a Monte Carlo simulation study and a resampling study based on real survey data. The performance of the methods was defined by the degree to which they facilitate unbiased and confidence-valid estimates of the parameters of complete data analysis models. We found that using lasso penalty or forward selection to select the predictors used in the MI model and using principal component analysis to reduce the dimensionality of auxiliary data produce the best results.

Hospital volume and outcomes following bronchoscopy in patients with interstitial lung disease: A retrospective observational study using a national inpatient database in Japan

BackgroundInterstitial lung diseases (ILDs) are a group of diffuse parenchymal lung disorders that cause inflammation and fibrosis in the interstitium of the lungs. Histopathological examination is pivotal to accurately diagnose the type of ILD, and bronchoscopy (BS) is often performed to collect lung tissue. This study aimed to determine the relationship between hospital volume and outcomes following BS in patients with ILD. MethodsInpatient data on patients with ILD who underwent BS between July 1, 2010 and March 31, 2021 were extracted from the Japanese Diagnosis Procedure Combination database. The annual hospital volume of BS was categorized into four (very low- [≤15 cases/year], low- [16–29 cases/year], high- [30–54 cases/year], and very high- [≥55 cases/year] volume) groups. The primary outcome was all-cause 14-day mortality after BS. Multiple imputation methods followed by multivariable logistic regression analyses fitted with generalized estimating equations were used to estimate the association between hospital volume and 14-day mortality after BS. ResultsA total of 89,454 patients with ILD from 1002 hospitals underwent BS. The all-cause mortality within 14 days after BS was 0.77%. An inverse trend was observed between mortality and hospital volume. Compared with the very low-hospital volume group, the very high-hospital volume group was significantly associated with a lower mortality (adjusted odds ratio = 0.63, 95% confidence interval: 0.48–0.85, p = 0.002). ConclusionsHospital volume was inversely associated with all-cause mortality within 14 days after BS for hospitalized patients with ILD.

Improvement of an External Predictive Model Based on New Information Using a Synthetic Data Approach: Application to CADASIL.

Cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) is the most frequent hereditary cerebral small vessel disease. It is caused by mutations of the NOTCH3 gene. The disease evolves progressively over decades leading to stroke, disability, cognitive decline, and functional dependency. The course and clinical severity of CADASIL seem heterogeneous. Predictive models are thus needed to improve prognostic evaluation and inform future clinical trials. A predictive model of the 3-year variation in the Mattis Dementia Rating Scale (MDRS), which reflects the global cognitive performance of patients with CADASIL, was previously proposed. This model made predictions based on demographic, clinical, and MRI data. We aimed to improve this existing predictive model by integrating a new potential factor, the location of the genetic mutation in the different epidermal growth factor (EGFr) domains of the NOTCH3 gene, dichotomized into EGFr domains 1 to 6 or 7 to 34. We used a new synthetic data approach to improve the initial predictive model by incorporating additional genetic information. This method combined the predicted outcomes from the previous model and 5 "synthetic" data sets with the observed outcome in a new data set. We then applied a multiple imputation method for missing data on the mutation location. The new data set included 367 patients who were followed up for 30 to 42 months. In the multivariable model with synthetic data, patients with NOTCH3 mutations in EGFr domains 7 to 34 had an additional average decrease of -1.4 points (standard error 0.67, p = 0.035) in their MDRS score variation over 3 years compared with patients with mutations located in EGFr domains 1 to 6. Cross-validation results highlighted the improved predictive performance of the enhanced model. Moreover, the model estimation was found to be more robust than fitting a model without synthetic data. The use of synthetic data improved the predictive model of MDRS change over 3 years in CADASIL. The predictive performance and estimation robustness of the predictive model were enhanced using this approach, whether genetic information was used. A statistically significant association between the location of the mutation in the NOTCH3 gene and the 3-year MDRS score variation was detected.

Predictive analytics in smart healthcare for child mortality prediction using a machine learning approach.

In developing countries, child health and restraining under-five child mortality are one of the fundamental concerns. UNICEF adopted sustainable development goal 3 (SDG3) to reduce the under-five child mortality rate globally to 25 deaths per 1,000 live births. The under-five mortality rate is 69 deaths per 1,000 live child-births in Pakistan as reported by the Demographic and Health Survey (2018). Predictive analytics has the power to transform the healthcare industry, personalizing care for every individual. Pakistan Demographic Health Survey (2017-2018), the publicly available dataset, is used in this study and multiple imputation methods are adopted for the treatment of missing values. The information gain, a feature selection method, ranked the information-rich features and examine their impact on child mortality prediction. The synthetic minority over-sampling method (SMOTE) balanced the training dataset, and four supervised machine learning classifiers have been used, namely the decision tree classifier, random forest classifier, naive Bayes classifier, and extreme gradient boosting classifier. For comparative analysis, accuracy, precision, recall, and F1-score have been used. Eventually, a predictive analytics framework is built that predicts whether the child is alive or dead. The number under-five children in a household, preceding birth interval, family members, mother age, age of mother at first birth, antenatal care visits, breastfeeding, child size at birth, and place of delivery were found to be critical risk factors for child mortality. The random forest classifier performed efficiently and predicted under-five child mortality with accuracy (93.8%), precision (0.964), recall (0.971), and F1-score (0.967). The findings could greatly assist child health intervention programs in decision-making.

O-131 BEYOND: a randomised controlled trial comparing efficacy and safety of individualised follitropin delta dosing in GnRH agonist versus antagonist protocols for first ovarian stimulation cycle

Abstract Study question How does a long GnRH agonist versus a GnRH antagonist protocol affect ovarian response when using individualised fixed daily follitropin delta dose for ovarian stimulation? Summary answer BEYOND trial data demonstrates that individualised follitropin delta dosing is safe and effective in a long GnRH agonist protocol in women with AMH ≤35 pmol/L. What is known already Efficacy and safety of a fixed daily dose of follitropin delta, individualised based on bodyweight and anti-Müllerian (AMH), have been established in RCTs using a GnRH antagonist protocol. Preliminary study data shows that individualised follitropin delta is efficacious when used with a long GnRH agonist protocol (RAINBOW trial). There is no comparative data for efficacy and safety of individualised follitropin delta in a long GnRH agonist versus an antagonist protocol. Study design, size, duration This was a randomised, controlled, open-label, multi-centre trial comparing efficacy and safety of individualised follitropin delta dosing in a long GnRH agonist versus an antagonist protocol in participants undergoing their first ovarian stimulation for IVF/ICSI conducted between May 2019 and February 2022. The primary endpoint was the number of oocytes retrieved. Important secondary endpoints included ongoing pregnancy rates and adverse drug reactions, with a special focus on OHSS. A total of 437 participants were randomised. Participants/materials, setting, methods Participants, 18–40 years old with AMH ≤35 pmol/L, were enrolled at specialist reproductive health clinics in Austria, Denmark, Israel, Italy, the Netherlands, Norway and Switzerland. The mean number of oocytes retrieved was compared between the agonist and antagonist protocols using a negative binomial regression model with age and AMH at screening as factors. The analyses were based on all randomised subjects, using a multiple imputation method for randomised subjects withdrawing before start of stimulation. Main results and the role of chance Of the 437 randomised subjects, 202 and 204 initiated ovarian stimulation with follitropin delta in the agonist and antagonist groups, respectively. Baseline mean data were: age, 32.3±4.3 years; AMH, 16.6±7.8 pmol/L, respectively. The number of oocytes retrieved was statistically significantly higher in the agonist (11.1±5.9) versus antagonist (9.6±5.5) group, with an estimated mean difference of 1.31 oocytes (95% CI: 0.22; 2.40, p = 0.0185). The difference in number of oocytes retrieved was influenced by the patients’ age and ovarian reserve, with a greater difference observed in patients <35 years and patients with high ovarian reserve (AMH >15 pmol/L). Cycle cancellations (2.0% versus 3.4%) and transfer cancellations (13.7% versus 14.7%) were similar in both groups. Ongoing pregnancy rate (36.9% vs 29.1%; estimated difference: 7.74% [95% CI: –1.49; 16.97, p = 0.1002]) and other pregnancy outcomes were higher in the agonist versus antagonist protocol group, but the differences were not statistically significant. The most commonly reported adverse events (≥1% in either group) were similar in both groups (headache, OHSS, nausea, pelvic pain or discomfort, and abdominal pain). Incidence of early moderate/severe OHSS was low (1.5% versus 2.5%). Early pregnancy loss was similar between the groups (20% vs 24% of subjects with positive beta-hCG test, respectively). Limitations, reasons for caution Subjects with AMH >35 pmol/L were not enrolled. Clinicians should remain cautious when using a GnRH agonist protocol in patients with AMH >35 pmol/L (greater OHSS risk). OHSS incidence in the GnRH antagonist group may have been lower if a GnRH agonist trigger had been allowed. Wider implications of the findings In women with AMH ≤35 pmol/L, a fixed daily dose of follitropin delta (individualised according to bodyweight and AMH) showed similar efficacy when used in a long GnRH agonist protocol with no additional safety signals observed and no additional risk of OHSS versus follitropin delta used in an antagonist protocol. Trial registration number ClinicalTrials.gov identifier: NCT03809429; EudraCT number: 2017-002783-40

Comparison of Imputation Strategies for Incomplete Longitudinal Data in Life-Course Epidemiology.

Incomplete longitudinal data are common in life-course epidemiology and may induce bias leading to incorrect inference. Multiple imputation (MI) is increasingly preferred for handling missing data, but few studies explore MI-method performance and feasibility in real-data settings. We compared 3 MI methods using real data under 9 missing-data scenarios, representing combinations of 10%, 20%, and 30% missingness and missing completely at random, at random, and not at random. Using data from Health and Retirement Study (HRS) participants, we introduced record-level missingness to a sample of participants with complete data on depressive symptoms (1998-2008), mortality (2008-2018), and relevant covariates. We then imputed missing data using 3 MI methods (normal linear regression, predictive mean matching, variable-tailored specification), and fitted Cox proportional hazards models to estimate effects of 4 operationalizations of longitudinal depressive symptoms on mortality. We compared bias in hazard ratios, root mean square error, and computation time for each method. Bias was similar across MI methods, and results were consistent across operationalizations of the longitudinal exposure variable. However, our results suggest that predictive mean matching may be an appealing strategy for imputing life-course exposure data, given consistently low root mean square error, competitive computation times, and few implementation challenges.

Computing within-study covariances, data visualization, and missing data solutions for multivariate meta-analysis with metavcov.

Multivariate meta-analysis (MMA) is a powerful statistical technique that can provide more reliable and informative results than traditional univariate meta-analysis, which allows for comparisons across outcomes with increased statistical power. However, implementing appropriate statistical methods for MMA can be challenging due to the requirement of various specific tasks in data preparation. The metavcov package aims for model preparation, data visualization, and missing data solutions to provide tools for different methods that cannot be found in accessible software. It provides sufficient constructs for estimating coefficients from other well-established packages. For model preparation, users can compute both effect sizes of various types and their variance-covariance matrices, including correlation coefficients, standardized mean difference, mean difference, log odds ratio, log risk ratio, and risk difference. The package provides a tool to plot the confidence intervals for the primary studies and the overall estimates. When specific effect sizes are missing, single imputation is available in the model preparation stage; a multiple imputation method is also available for pooling the results in a statistically principled manner from models of users' choice. The package is demonstrated in two real data applications and a simulation study to assess methods for handling missing data.

893-P: Relation between Combination Therapy with SGLT2 Inhibitors and GLP-1 Receptor Agonists and Renal Functions among Japanese Diabetics

Background/Objective: SGLT2 inhibitors (SGLT2i) and GLP1 receptor agonists (GLP1Ra) have already shown evidence of cardiovascular and renal outcomes; therefore, combination therapy with these drugs is now of interest in clinical practice for the management of type 2 diabetes mellitus (T2DM). However, there are few reports on the effect of combination therapy on renal functions. This study aimed to investigate the effect of combination therapy on renal functions. Methods: We retrospectively extracted data of patients with T2DM who had received both SGLT2i and GLP1Ra treatment for at least 1-year at 18 medical facilities in Japan. Finally, 341 patients in the GLP1Ra-preceding group and 319 in the SGLT2i-preceding group were analyzed. The multiple imputation method was adopted for missing values, and propensity score (PS) matching with the algorithm of 1:1, neighbor match, and no replacement, was performed for comparison between the two groups. Result: In each group, 152 patients were matched. On the matched model, mean observational period was 54 months. Regarding the annual change in eGFR (mL/min/m2/year) before and after combination treatment, the SGLT2i-preceding group showed a significant increase from -3.76±10.48 to 0.04±6.29 (p=0.002), whereas the GLP1Ra-preceding group showed no significant change from -2.09±13.44 to -1.97±6.16. The logarithmic value of urine albumin-to-creatinine ratio at the time of the initiation of either drug, at the time of the initiation of combination treatment, and after the combination treatment, were 3.77±1.63, 3.83±2.21, 3.73±1.80 in GLP1Ra-preceding group, and 3.65±20.4, 3.56±1.29, 3.75±0.91 in SGLT2i-preceding group, respectively, and there was no significant difference among two groups. Conclusion: The addition of GLP1Ra to SGLT2i-treated patients significantly improved the annual change in eGFR. The renal benefits by combination therapy with SGLT2i and GLP1Ra may be affected by preceding drug. Disclosure M.Toyoda: None. K.Kobayashi: None.

892-P: Influence of Preceding Medications on the Renal Outcomes of Japanese Diabetics Treated with SGLT2 Inhibitors and GLP-1 Receptor Agonists

Background/Objective: Evidence of cardiovascular and renal outcomes has already been shown for SGLT2 inhibitors (SGLT2i) and GLP1 receptor agonists (GLP1Ras); therefore, combination therapy with these drugs is of clinical interest. This study aimed to evaluate the effect of the preceding drug for patients with the combination treatment of SGLT2i and GLP1Ra on renal outcomes. Methods: We retrospectively extracted data of patients with type 2 diabetes mellitus who had received both SGLT2i and GLP1Ra treatment for at least 1-year at 18 medical facilities in Japan. Finally, 341 patients in the GLP1Ra-preceding group and 319 in the SGLT2i-preceding group were analyzed. The renal composite outcome was progression of the albuminuria stage, ≥30% decrease in eGFR, or both. The multiple imputation method for missing values and an inverse probability weighting method using the propensity score were used for the comparison of outcome. Results: The median observation period and the incidence of the renal composite outcome in the SGLT2i-preceding group and the GLP1Ra-preceding group was 55.1months, 26.9% and 58.7months, 25.6%, respectively, with an odds ratio (OR) [95% confidence interval] of 1.50 [0.89, 2.53] (p=0.13). The ORs for the ≥30% decrease in eGFR and the progression of albuminuria stage were 0.96 [0.47, 2.00] (p=0.93) and 1.76 [0.97, 3.69] (p=0.06), respectively. Compared to the GLP1Ra-preceding group, the annual change in eGFR and change in the logarithmic value of urine albumin-to-creatinine ratio (LnACR) was 0.27mL/min/1.73 m2/year [-0.60, 1.14], and 0.18 [-0.10, 0.43] respectively. LnACR at baseline, pioglitazone use, and baseline mean arterial blood pressure were independent factors for renal composite outcome with ORs [95% CI] of 1.15 [1.01, 1.31], 0.47 [0.23, 0.96], and 1.03 [1.01, 1.05], respectively. Conclusion: In combination therapy with SGLT2i and GLP1Ra, the difference in the preceding drug did not affect the renal composite outcome. Disclosure K.Kobayashi: None. M.Toyoda: None.

Multiple imputation of missing data under missing at random: compatible imputation models are not sufficient to avoid bias if they are mis-specified

ObjectivesEpidemiological studies often have missing data, which are commonly handled by multiple imputation (MI). Standard (default) MI procedures use simple linear covariate functions in the imputation model. We examine the bias that may be caused by acceptance of this default option and evaluate methods to identify problematic imputation models, providing practical guidance for researchers. Study Design and SettingUsing simulation and real data analysis, we investigated how imputation model mis-specification affected MI performance, comparing results with complete records analysis (CRA). We considered scenarios in which imputation model mis-specification occurred because (i) the analysis model was mis-specified or (ii) the relationship between exposure and confounder was mis-specified. ResultsMis-specification of the relationship between outcome and exposure, or between exposure and confounder, can cause biased CRA and MI estimates (in addition to any bias in the full-data estimate due to analysis model mis-specification). MI by predictive mean matching can mitigate model mis-specification. Methods for examining model mis-specification were effective in identifying mis-specified relationships. ConclusionWhen using MI methods that assume data are MAR, compatibility between the analysis and imputation models is necessary, but not sufficient to avoid bias. We propose a step-by-step procedure for identifying and correcting mis-specification of imputation models.

Using Multiple Imputation to Account for the Uncertainty Due to Missing Data in the Context of Factor Retention.

Although parallel analysis has been found to be an accurate method for determining the number of factors in many conditions with complete data, its application under missing data is limited. The existing literature recommends that, after using an appropriate multiple imputation method, researchers either apply parallel analysis to every imputed data set and use the number of factors suggested by most of the data copies or average the correlation matrices across all data copies, followed by applying the parallel analysis to the average correlation matrix. Both approaches for pooling the results provide a single suggested number without reflecting the uncertainty introduced by missing values. The present study proposes the use of an alternative approach, which calculates the proportion of imputed data sets that result in k (k = 1, 2, 3 . . .) factors. This approach will inform applied researchers of the degree of uncertainty due to the missingness. Results from a simulation experiment show that the proposed method can more likely suggest the correct number of factors when missingness contributes to a large amount of uncertainty.

Methods for comparative effectiveness based on time to confirmed disability progression with irregular observations in multiple sclerosis.

Real-world data sources offer opportunities to compare the effectiveness of treatments in practical clinical settings. However, relevant outcomes are often recorded selectively and collected at irregular measurement times. It is therefore common to convert the available visits to a standardized schedule with equally spaced visits. Although more advanced imputation methods exist, they are not designed to recover longitudinal outcome trajectories and typically assume that missingness is non-informative. We, therefore, propose an extension of multilevel multiple imputation methods to facilitate the analysis of real-world outcome data that is collected at irregular observation times. We illustrate multilevel multiple imputation in a case study evaluating two disease-modifying therapies for multiple sclerosis in terms of time to confirmed disability progression. This survival outcome is derived from repeated measurements of the Expanded Disability Status Scale, which is collected when patients come to the healthcare center for a clinical visit and for which longitudinal trajectories can be estimated. Subsequently, we perform a simulation study to compare the performance of multilevel multiple imputation to commonly used single imputation methods. Results indicate that multilevel multiple imputation leads to less biased treatment effect estimates and improves the coverage of confidence intervals, even when outcomes are missing not at random.

Estimating the aquatic risk from exposure to up to twenty-two pesticide active ingredients in waterways discharging to the Great Barrier Reef

Pesticides decrease the quality of water reaching the Great Barrier Reef (GBR), Australia. Up to 86 pesticide active ingredients (PAIs) were monitored between July 2015 to end of June 2018 at 28 sites in waterways that discharge to the GBR. Twenty-two frequently detected PAIs were selected to calculate their combined risk when they co-occur in water samples. Species sensitivity distributions (SSDs) for the 22 PAIs to fresh and marine species were developed. The SSDs, the multi-substance potentially affected fraction (msPAF) method, Independent Action model of joint toxicity and a Multiple Imputation method were combined to convert measured PAI concentration data to estimates of the Total Pesticide Risk for the 22 PAIs (TPR22) expressed as the average percentage of species affected during the wet season (i.e., 182 days). The TPR22 and percent contribution of active ingredients of Photosystem II inhibiting herbicides, Other Herbicides, and Insecticides to the TPR22 were estimated. The TPR22 ranged from <1 % to 42 % of aquatic species being affected. Approximately 85 % of the TPR22 estimates were >1 % — meaning they did not meet the Reef 2050 Water Quality Improvement Plan's pesticide target for waters entering the GBR. There were marked spatial differences in TPR22 estimates — regions dominated by grazing had lower estimates while those with sugar cane tended to have higher estimates. On average, active ingredients of PSII herbicides contributed 39 % of the TPR22, the active ingredients of Other Herbicides contributed ~36 % and of Insecticides contributed ~24 %. Nine PAIs (diuron, imidacloprid, metolachlor, atrazine, MCPA, imazapic, metsulfuron, triclopyr and ametryn) were responsible for >97 % of TPR22 across all the monitored waterways.

Comparison of multiple imputation and other methods for the analysis of imputed genotypes

BackgroundAnalysis of imputed genotypes is an important and routine component of genome-wide association studies and the increasing size of imputation reference panels has facilitated the ability to impute and test low-frequency variants for associations. In the context of genotype imputation, the true genotype is unknown and genotypes are inferred with uncertainty using statistical models. Here, we present a novel method for integrating imputation uncertainty into statistical association tests using a fully conditional multiple imputation (MI) approach which is implemented using the Substantive Model Compatible Fully Conditional Specification (SMCFCS). We compared the performance of this method to an unconditional MI and two additional approaches that have been shown to demonstrate excellent performance: regression with dosages and a mixture of regression models (MRM).ResultsOur simulations considered a range of allele frequencies and imputation qualities based on data from the UK Biobank. We found that the unconditional MI was computationally costly and overly conservative across a wide range of settings. Analyzing data with Dosage, MRM, or MI SMCFCS resulted in greater power, including for low frequency variants, compared to unconditional MI while effectively controlling type I error rates. MRM andl MI SMCFCS are both more computationally intensive then using Dosage.ConclusionsThe unconditional MI approach for association testing is overly conservative and we do not recommend its use in the context of imputed genotypes. Given its performance, speed, and ease of implementation, we recommend using Dosage for imputed genotypes with MAF ge 0.001 and Rsq ge 0.3.