SDImpute: A statistical block imputation method based on cell-level and gene-level information for dropouts in single-cell RNA-seq data

The single-cell RNA sequencing (scRNA-seq) technologies obtain gene expression at single-cell resolution and provide a tool for exploring cell heterogeneity and cell types. As the low amount of extracted mRNA copies per cell, scRNA-seq data exhibit a large number of dropouts, which hinders the downstream analysis of the scRNA-seq data. We propose a statistical method, SDImpute (Single-cell RNA-seq Dropout Imputation), to implement block imputation for dropout events in scRNA-seq data. SDImpute automatically identifies the dropout events based on the gene expression levels and the variations of gene expression across similar cells and similar genes, and it implements block imputation for dropouts by utilizing gene expression unaffected by dropouts from similar cells. In the experiments, the results of the simulated datasets and real datasets suggest that SDImpute is an effective tool to recover the data and preserve the heterogeneity of gene expression across cells. Compared with the state-of-the-art imputation methods, SDImpute improves the accuracy of the downstream analysis including clustering, visualization, and differential expression analysis.     

Generalized linear mixture models for handling nonignorable dropouts in longitudinal studies

This paper presents a method for analysing longitudinal data when there are dropouts. In particular, we develop a simple method based on generalized linear mixture models for handling nonignorable dropouts for a variety of discrete and continuous outcomes. Statistical inference for the model parameters is based on a generalized estimating equations (GEE) approach (Liang and Zeger, 1986). The proposed method yields estimates of the model parameters that are valid when nonresponse is nonignorable under a variety of assumptions concerning the dropout process. Furthermore, the proposed method can be implemented using widely available statistical software. Finally, an example using data from a clinical trial of contracepting women is used to illustrate the methodology.     

Legal abortions,socioeconomic status,and measured intelligence in the United States

Abstract

A random national sample of 299 clinic dropouts from the Jamaica National Family Planning Program were interviewed at least five months beyond their first missed clinic appointment. Correlates of early (after 1 or 2 visits) versus late (3+ visits) dropping out, continued contraceptive practice, and pregnancy since visiting the clinic were identified. The Health Belief Model of “compliance” behavior is proposed as an appropriate theoretical basis for future studies of clinic dropouts.     

A note on the effect of ignoring dropout cases on fecundability as estimated from a follow-up study

Fecundability is often overestimated by applying models that do not allow for dropouts to the data that have eliminated the actual dropout cases. This note studies the extent of this overestimation and suggests a simple method of adjustment of data that removes this possibility of over-estimation.     

Allelic dropout from a high-quality DNA source

Allelic dropouts are an important source of genotyping error, particularly in studies using non-invasive sampling techniques. This has important implications for conservation biology, as an increasing number of studies are now using non-invasive techniques to study rare species or endangered populations. Previously, allelic dropout has typically been associated with PCR amplification of low quality/quantity template DNA. However, in this study we recorded high levels of allelic dropout (21–57%) at specific loci amplified from a high quality DNA (63.1 ± 7.8 ng/μl) source in the red fox (Vulpes vulpes). We designed a series of experiments to identify the sources of error. Whilst we were able to show that the best method to identify allelic dropout was the dilution of template DNA prior to PCR amplification, our data also showed two specific patterns: (1) allelic dropouts occurred at specific loci; (2) allelic dropouts occurred at specific pair-wise combinations of alleles. These patterns suggest that mechanisms other than low quantity template DNA are responsible for allelic dropout. Further research on the causes of these patterns in this and other studies would further our understanding of genotyping errors and would aid future studies where allelic dropout may be a serious issue.     

Problems in dealing with missing data and informative censoring in clinical trials

A common problem in clinical trials is the missing data that occurs when patients do not complete the study and drop out without further measurements. Missing data cause the usual statistical analysis of complete or all available data to be subject to bias. There are no universally applicable methods for handling missing data. We recommend the following: (1) Report reasons for dropouts and proportions for each treatment group; (2) Conduct sensitivity analyses to encompass different scenarios of assumptions and discuss consistency or discrepancy among them; (3) Pay attention to minimize the chance of dropouts at the design stage and during trial monitoring; (4) Collect post-dropout data on the primary endpoints, if at all possible; and (5) Consider the dropout event itself an important endpoint in studies with many.     

Methods to Maximize Retention in Weight Loss Studies

GREENWAY, FRANK L., GEORGE A. BRAY, AND ROBERT L. MARLIN. Methods to maximize retention in weight loss studies. Obes Res. Objective: Dropouts from clinical trials decrease quality and increase costs. Free participation, paid participation, and contingency contracting are three study retention methods. Contingency contracting, or depositing a fee to be refunded contingent upon attendance in a clinical trial, has been reported to decrease dropouts without affecting weight loss. These three methods of retention were compared with a commercial weight loss clinic's practice of charging non-refundable fees. Methods and Procedures: Dropouts were compared in two studies testing mazindol, with one study using free care and the other using contingency contracting; two studies testing phenylpropanolamine, one using free care and the other using contingency contracting; and in studies with phenylpropanolamine on file with Thompson Medical Company using free care, paid participation, and contingency contracting. Results: The dropout rate was 50% at 8 weeks in a trial of mazindol with free care vs. 7% for contingency contracting (p <0. 001). The two phenylpropanolamine studies gave the same weight losses, but the dropouts were 37% at 8 weeks for free care vs. 11 % for contingency contracting (p<0. 001). The studies of phenylpropanolamine on file at the Thompson Medical Company had 28% dropouts at 8 weeks using free care vs. 19% for paid participation (p<0. 001), and 11% for contingency contracting (p<0.005). Dropouts with contingency contracting (11%) were not different from the commercial weight loss program (13%). Discussion: Contingency contracting can decrease dropouts, improve quality, and decrease costs without affecting weight loss in clinical trials for obesity.     

Predicting Power for Longitudinal Studies with Attrition

Power analyses play an integral role in designing biomedical studies and are customary in biomedical research proposals, for example, submitted to federal regulatory and medical research agencies. An underpowered study potentially leads to inconclusive inferences and consequently misspends valuable time and financial resources allocated to the study. The occurrence of attrition or dropout further increases the risk of conducting an underpowered study. In some applications, it is desirable to provide additional assurance against insufficient statistical power by producing conservative power predictions. We propose two new methods for predicting power when expecting attrition, and both methods employ a simple probability model for the number of dropouts. One approach allows conservative power predictions that reduce the chance of underpowering studies. The second procedure gives the minimum mean squared error predictor of conditional power, given dependence on a random number of unobserved values or dropouts. We illustrate our methods for predicting power using a longitudinal study of depression. (© 2004 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

Quantifying genotyping errors in noninvasive population genetics

The use of noninvasively collected samples greatly expands the range of ecological issues that may be investigated through population genetics. Furthermore, the difficulty of obtaining reliable genotypes with samples containing low quantities of amplifiable DNA may be overcome by designing optimal genotyping schemes. Such protocols are mainly determined by the rates of genotyping errors caused by false alleles and allelic dropouts. These errors may not be avoided through laboratory procedure and hence must be quantified. However, the definition of genotyping error rates remains elusive and various estimation methods have been reported in the literature. In this paper we proposed accurate codification for the frequencies of false alleles and allelic dropouts. We then reviewed other estimation methods employed in hair- or faeces-based population genetics studies and modelled the bias associated with erroneous methods. It is emphasized that error rates may be substantially underestimated when using an erroneous approach. Genotyping error rates may be important determinants of the outcome of noninvasive studies and hence should be carefully computed and reported.     

Addressing an Idiosyncrasy in Estimating Survival Curves Using Double Sampling in the Presence of Self-Selected Right Censoring

We investigate the use of follow-up samples of individuals to estimate survival curves from studies that are subject to right censoring from two sources: (i) early termination of the study, namely, administrative censoring, or (ii) censoring due to lost data prior to administrative censoring, so-called dropout. We assume that, for the full cohort of individuals, administrative censoring times are independent of the subjects' inherent characteristics, including survival time. To address the loss to censoring due to dropout, which we allow to be possibly selective, we consider an intensive second phase of the study where a representative sample of the originally lost subjects is subsequently followed and their data recorded. As with double-sampling designs in survey methodology, the objective is to provide data on a representative subset of the dropouts. Despite assumed full response from the follow-up sample, we show that, in general in our setting, administrative censoring times are not independent of survival times within the two subgroups, nondropouts and sampled dropouts. As a result, the stratified Kaplan-Meier estimator is not appropriate for the cohort survival curve. Moreover, using the concept of potential outcomes, as opposed to observed outcomes, and thereby explicitly formulating the problem as a missing data problem, reveals and addresses these complications. We present an estimation method based on the likelihood of an easily observed subset of the data and study its properties analytically for large samples. We evaluate our method in a realistic situation by simulating data that match published margins on survival and dropout from an actual hip-replacement study. Limitations and extensions of our design and analytic method are discussed.     

Joint inference for nonlinear mixed-effects models and time to event at the presence of missing data

In many longitudinal studies, the individual characteristics associated with the repeated measures may be possible covariates of the time to an event of interest, and thus, it is desirable to model the time-to-event process and the longitudinal process jointly. Statistical analyses may be further complicated in such studies with missing data such as informative dropouts. This article considers a nonlinear mixed-effects model for the longitudinal process and the Cox proportional hazards model for the time-to-event process. We provide a method for simultaneous likelihood inference on the 2 models and allow for nonignorable data missing. The approach is illustrated with a recent AIDS study by jointly modeling HIV viral dynamics and time to viral rebound.     

Nonspecific Adverse Events in Knee Osteoarthritis Clinical Trials: A Systematic Review

Background

Adverse events (AEs) derived from nonspecific activity of treatments can impair the validity of trials, and even make it difficult to identify specific AEs associated with treatments. To better understand these nonspecific AEs, we investigated the AEs in placebo groups by using knee osteoarthritis clinical trials.

Methods

Randomized, placebo-controlled, knee osteoarthritis trials were identified by searching electronic databases. We determined the rate of patients with AEs and the rate of dropouts caused by AEs in the active and placebo groups. Furthermore, we calculated the rate of patients for individual AEs in the placebo groups. Finally, we performed secondary analyses to identify the factors associated with these rates.

Results

Overall, 272 papers reporting 281 trials were included in the analysis. The rates of patients with AEs were 31.8% in the active groups and 27.4% in the placebo groups. The rate of the placebo groups accounted for 86.2% of the rate of the active groups. The rates of dropouts caused by AEs were 5.2% in the active groups and 4.8% in the placebo groups. The rate of the placebo groups accounted for 92.3% of the rate of the active groups. AEs in the placebo groups included a number of clinical conditions, with elevated alanine aminotransferase (0.59%; 95% CI: 0.46 to 0.77) being the most common objective outcome and headache (4.48%; 95% CI: 4.20 to 4.79) being the most frequent subjective outcome. The rate of patients with AEs and the rate of dropouts caused by AEs were associated with the treatment type, delivery route, and study design.

Conclusions

The nonspecific AEs substantially accounted for the development of AEs in the active groups and included conditions involving the entire body.     

A Model for the Censoring Process in Incomplete Repeated Measurement Experiments

Premature terminations or dropouts occur often in repeated measurement experiments. A number of methods have been proposed to analyze such data but most of them assume that the censoring mechanism is, within each group, unaffected by the mechanism generating the response variables. In this paper, we propose a model for the censoring mechanism that generates dropouts. We then show how this model can be used to check whether the censoring mechanism is affected by the response variables and other covariates. Finally, the methods of the paper are applied to the “Halothane” data set.     

Prediction of early termination from the self-regulatory treatment of chronic headache

The ability of demographic, psychological testing and history information to predict which patients will terminate early from nonpharmacological treatment of headache (relaxation and biofeedback) was tested. Information from each of these areas was initially examined for differences between dropouts and treatment completers using univariate analyses. These analyses were followed by a canonical discriminate function analysis that predicted whether patients would complete treatment or drop out. Information from the three predictor sets combined resulted in 77.4% of the patients being correctly classified.     

Overcoming Barriers to Skills Training in Borderline Personality Disorder: A Qualitative Interview Study

Despite evidence suggesting that skills training is an important mechanism of change in dialectical behaviour therapy, little research exploring facilitators and barriers to this process has been conducted. The study aimed to explore clients’ experiences of barriers to dialectical behaviour therapy skills training and how they felt they overcame these barriers, and to compare experiences between treatment completers and dropouts. In-depth qualitative interviews were conducted with 40 clients with borderline personality disorder who had attended a dialectical behaviour therapy programme. A thematic analysis of participants’ reported experiences found that key barriers to learning the skills were anxiety during the skills groups and difficulty understanding the material. Key barriers to using the skills were overwhelming emotions which left participants feeling unable or unwilling to use them. Key ways in which participants reported overcoming barriers to skills training were by sustaining their commitment to attending therapy and practising the skills, personalising the way they used them, and practising them so often that they became an integral part of their behavioural repertoire. Participants also highlighted a number of key ways in which they were supported with their skills training by other skills group members, the group therapists, their individual therapist, friends and family. Treatment dropouts were more likely than completers to describe anxiety during the skills groups as a barrier to learning, and were less likely to report overcoming barriers to skills training via the key processes outlined above. The findings of this qualitative study require replication, but could be used to generate hypotheses for testing in further research on barriers to skills training, how these relate to dropout, and how they can be overcome. The paper outlines several such suggestions for further research.     

HiCImpute: A Bayesian hierarchical model for identifying structural zeros and enhancing single cell Hi-C data

Single cell Hi-C techniques enable one to study cell to cell variability in chromatin interactions. However, single cell Hi-C (scHi-C) data suffer severely from sparsity, that is, the existence of excess zeros due to insufficient sequencing depth. Complicating the matter further is the fact that not all zeros are created equal: some are due to loci truly not interacting because of the underlying biological mechanism (structural zeros); others are indeed due to insufficient sequencing depth (sampling zeros or dropouts), especially for loci that interact infrequently. Differentiating between structural zeros and dropouts is important since correct inference would improve downstream analyses such as clustering and discovery of subtypes. Nevertheless, distinguishing between these two types of zeros has received little attention in the single cell Hi-C literature, where the issue of sparsity has been addressed mainly as a data quality improvement problem. To fill this gap, in this paper, we propose HiCImpute, a Bayesian hierarchical model that goes beyond data quality improvement by also identifying observed zeros that are in fact structural zeros. HiCImpute takes spatial dependencies of scHi-C 2D data structure into account while also borrowing information from similar single cells and bulk data, when such are available. Through an extensive set of analyses of synthetic and real data, we demonstrate the ability of HiCImpute for identifying structural zeros with high sensitivity, and for accurate imputation of dropout values. Downstream analyses using data improved from HiCImpute yielded much more accurate clustering of cell types compared to using observed data or data improved by several comparison methods. Most significantly, HiCImpute-improved data have led to the identification of subtypes within each of the excitatory neuronal cells of L4 and L5 in the prefrontal cortex.     

Sample-size redetermination for repeated measures studies

Clinical trialists recently have shown interest in two-stage procedures for updating the sample-size calculation at an interim point in a trial. Because many clinical trials involve repeated measures designs, it is desirable to have available practical two-stage procedures for such designs. Shih and Gould (1995, Statistics in Medicine 14, 2239-2248) discuss sample-size redetermination for repeated measures studies but under a highly simplified setup. We develop two-stage procedures under the general mixed linear model, allowing for dropouts and missed visits. We present a range of procedures and compare their Type I error and power by simulation. We find that, in general, the achieved power is brought considerably closer to the required level without inflating the Type I error rate. We also derive an inflation factor that ensures the power requirement is more closely met.     

Exact two-sample inference with missing data

When comparing follow-up measurements from two independent populations, missing records may arise due to censoring by events whose occurrence is associated with baseline covariates. In these situations, inferences based only on the completely followed observations may be biased if the follow-up measurements and the covariates are correlated. This article describes exact inference for a class of modified U-statistics under covariate-dependent dropouts. The method involves weighing each permutation according to the retention probabilities, and thus requires estimation of the missing data mechanism. The proposed procedure is nonparametric in that no distributional assumption is necessary for the outcome variables and the missingness patterns. Monte Carlo approximation by the Gibbs sampler is proposed, and is shown to be fast and accurate via simulation. The method is illustrated in two small data sets for which asymptotic inferential procedures may not be appropriate.     

Prediction of early termination from the self-regulatory treatment of chronic headache

The ability of demographic, psychological testing and history information to predict which patients will terminate early from nonpharmacological treatment of headache (relaxation and biofeedback) was tested. Information from each of these areas was initially examined for differences between dropouts and treatment completers using univariate analyses. These analyses were followed by a canonical discriminate function analysis that predicted whether patients would complete treatment or drop out. Information from the three predictor sets combined resulted in 77.4% of the patients being correctly classified.This research was supported by a grant from the NINCDS, NS15235.