How to deal with missing data? Multiple imputation by chained equations: recommendations and explanations for clinical practice / Comment gérer les données manquantes ? Imputation multiple par équations chaînées : recommandations et explications pour la pratique clinique

The presence of missing data, a constant problem in medical research, has several consequences: systematic loss of power, associated or not with a reduction in the representativeness of the sample analyzed. There are three types of missing data: 1) missing completely at random (MCAR); 2) missing at random (MAR); 3) missing not at random (MNAR). Multiple imputation by chained equations allows for the correct handling of missing data under the MCAR and MAR assumptions. It allows to simulate for each missing data j, a number m of simulated values which seem plausible with regard to the other variables. A random effect is included in this simulation to express the uncertainty. Several data sets are thus created and analyzed individually, in an identical way. Then the estimators of each data set are combined to obtain a global estimator. Multiple imputation increases power, corrects for some biases and has the advantage of being applicable to many types of variables. Complete case analysis should no longer be the norm. The objective of this guide is to help the reader in conducting an analysis with multiple imputed data. We cover the following points: the different types of missing data, the different historical approaches to handling them, and then we detail the multiple imputation method using chained equations. We provide a code example for the mice package of R®.

La présence de données manquantes, problème constant en recherche médicale, a plusieurs conséquences : une perte de puissance systématique, associée ou non à une diminution de la représentativité de l'échantillon analysé. Il existe trois types de données manquantes : 1) manquantes complètement au hasard (MCAR - missing completely at random) ; 2) manquantes au hasard (MAR - missing at random) ; 3) manquantes non au hasard (MNAR - missing not at random). L'imputation multiple par équations chaînées permet de prendre en charge correctement les données manquantes sous les hypothèses MCAR et MAR. Elle permet de simuler pour chaque donnée manquante j, un nombre m de valeurs simulées qui semblent plausibles au regard des autres variables. Un effet aléatoire est inclus dans cette simulation pour exprimer l'incertitude. Plusieurs jeux de données sont ainsi créés et analysés individuellement de façon identique. Ensuite, les estimateurs de chaque jeu de données sont regroupés pour obtenir un estimateur global. L'imputation multiple permet d'augmenter la puissance, de corriger certains biais et a l'avantage d'être applicable à de très nombreux types de variables. L'analyse en cas complets ne devrait plus être la norme. L'objectif de ce guide est d'aider le lecteur dans la réalisation d'une analyse avec des données imputées de manière multiple. Nous traitons ici les différents types de données manquantes, les approches historiques pour les gérer, puis nous détaillons la méthode d'imputation multiple par équations chaînées. Nous fournissons un exemple de code pour le package mice de R®.

From transfer to transition: review and proposition of the SFNDT dialysis Commission / Du transfert à la transition : mise au point et proposition de la Commission de dialyse de la SFNDT

Over the course of their disease, patients with chronic kidney disease (CKD) will be treated by several kidney replacement therapy (KRT) modalities. The transitions between KRT modalities can be experienced as traumatic by patients, and are associated with an increased morbidity and mortality, notably when they are not anticipated. Planning these transition phases could reduce the psychological trauma induced by the transfer, as well as reduce the risk of morbidity and mortality. However, the lack of a clear definition of a transfer, and the lack of criteria enabling the identification of patients at risk of transfer, prevents the anticipation of these transition phases at high risk for patients. We here discuss the various possible causes and risk factors of transfer from peritoneal dialysis (PD) to hemodialysis as well as transfer from hemodialysis to PD. The dialysis Commission of the Société francophone de néphrologie, dialyse et transplantation (SFNDT) makes some proposals to improve transition phases, such as the identification of patients at risk, specific PD programs for unplanned PD start, transition unit and hybrid therapy.

Dans le parcours de soins du patient insuffisant rénal chronique, les différentes modalités de suppléance rénale vont se succéder dans le temps créant ainsi des phases de transition. Ces phases de transition peuvent être vécues comme traumatisantes par les patients, et sont associées à une augmentation de la morbi-mortalité, particulièrement lorsqu'elles ne sont pas suffisamment anticipées. La planification de ces phases de transition par l'équipe de dialyse devrait permettre de diminuer l'expérience du traumatisme psychologique induit par le changement et de réduire le risque de sur-morbi-mortalité. Cependant, l'absence de définition standardisée de la phase de transition entre modalités, le manque de critères et d'outils identifiant les patients à risque de transfert et l'absence d'infrastructures dédiées à ces patients transitionnels sont autant de facteurs limitant l'anticipation de ces phases de transition. Nous abordons ici les différentes causes et possibles facteurs de risque du transfert de la dialyse péritonéale (DP) vers l'hémodialyse ainsi que du transfert de l'hémodialyse vers la DP. Dans cette mise au point, la Commission de dialyse de la Société francophone de néphrologie, dialyse et transplantation (SFNDT) émet certaines propositions pour améliorer la définition et la prise en charge de ces phases de transition, et propose des outils d'identification des sujets « transitionnels ¼ ainsi que des exemples structurels de programmes soutenant la transition, tels que le démarrage en urgence de la DP, l'unité transitionnelle et la dialyse hybride.

Fate of discarded kidney transplants: experience of the transplant university hospital center of Rennes / Devenir des greffons rénaux refusés : expérience du centre de transplantation du CHU de Rennes

Introduction: In a context of tension on the number of available kidney transplants compared to the number needed, the practices of refusal of transplants in the Rennes transplantation center were evaluated. Materials and methods: The donors completely refused by our team (no kidney accepted for any Rennes recipient) between January 1st 2012 and December 31st 2015 were identified from the national CRISTAL registry. The outcome of these refused transplants (possible transplantation in another center), the data of the recipients (from Rennes and other centers) and the data of the donors (refused and then finally accepted) were extracted. The outcome of recipients (from Rennes and other centers) was compared: graft survival (censored on death) and patient survival (not censored on cessation of function). The Kidney Donor Profile Index (KDPI) score was calculated and its usefulness studied. Results: Among the 203 rejected donors, 172 (85 %) were accepted for transplantation in another center; 89% of these grafts were functional at one year. In univariate analysis, Rennes recipients transplanted after a refusal had a better graft survival (censored on death) than recipients transplanted in another center with the refused graft (p < 0.001). The main limitation of this analysis is the non-comparability of the groups. The KDPI score was significantly associated with graft survival (censored on death). Of the 151 Rennes patients who had a refusal, 3% were still on the waiting list at the end of the observation period, the others spent a median additional time on dialysis of 220 days (Q1-Q3 81-483). Conclusion: Rennes recipients transplanted after a first refusal seem to have a better graft survival (censored on death) than recipients from other centers transplanted with refused grafts. This is to be weighed against the additional time on dialysis and even the risk of non-transplantation.

Introduction: Dans un contexte de tension sur le nombre de greffons rénaux disponibles comparé au nombre nécessaire, les pratiques de refus des greffons du centre de transplantation rennais ont été évaluées. Matériels et méthodes: À partir du registre national CRISTAL, les donneurs complètement refusés par notre équipe (aucun rein accepté pour aucun receveur rennais) entre le 1er janvier 2012 et le 31 décembre 2015 ont été identifiés. Le devenir de ces greffons refusés (éventuelle greffe dans un autre centre), les données des receveurs (rennais et des autres centres) et les données des donneurs (refusés puis finalement acceptés) ont été extraits. Le devenir des receveurs (rennais et des autres centres) a été comparé : survie du greffon (censurée sur le décès) et du patient (non censurée sur l'arrêt de fonction). Le score KDPI (Kidney Donor Profile Index) a été calculé et son intérêt étudié. Résultats: Parmi les 203 donneurs refusés, 172 (85 %) ont permis une transplantation dans un autre centre, dont 89 % de greffons fonctionnels à un an. En analyse univariée, les receveurs rennais greffés après un refus avaient une meilleure survie greffon (censurée sur le décès) que les receveurs greffés dans un autre centre avec le greffon refusé (p < 0,001). La principale limite de cette analyse est la non-comparabilité des groupes. Le score KDPI était significativement associé à la survie greffon (censurée sur le décès). Parmi les 151 patients rennais qui ont eu un refus, 3 % étaient toujours sur liste d'attente à l'issue de la période d'observation, les autres passaient un temps médian supplémentaire en dialyse de 220 jours (Q1-Q3 81-483). Conclusion: Les receveurs rennais greffés après un premier refus semblent avoir une meilleure survie du greffon (censurée sur le décès) que les receveurs des autres centres greffés avec les greffons refusés. C'est à mettre en balance avec le temps supplémentaire en dialyse, voire le risque de non-greffe.

Direct Xylella fastidiosa whole genome sequencing from various plant species using targeted enrichment.

A targeted enrichment method was developed to sequence Xylella fastidiosa genomic DNA directly from plant samples. The method was evaluated on various plant species infected with different strains at different levels of contamination. After enrichment, X. fastidiosa genome coverage was above 99.9% for all tested samples.

Suspicions of two bridgehead invasions of Xylella fastidiosa subsp. multiplex in France.

Of American origin, a wide diversity of Xylella fastidiosa strains belonging to different subspecies have been reported in Europe since 2013 and its discovery in Italian olive groves. Strains from the subspecies multiplex (ST6 and ST7) were first identified in France in 2015 in urban and natural areas. To trace back the most probable scenario of introduction in France, the molecular evolution rate of this subspecies was estimated at 3.2165 × 10-7 substitutions per site per year, based on heterochronous genome sequences collected worldwide. This rate allowed the dating of the divergence between French and American strains in 1987 for ST6 and in 1971 for ST7. The development of a new VNTR-13 scheme allowed tracing the spread of the bacterium in France, hypothesizing an American origin. Our results suggest that both sequence types were initially introduced and spread in Provence-Alpes-Côte d'Azur (PACA); then they were introduced in Corsica in two waves from the PACA bridgehead populations.

Switch from Hemodialysis to Peritoneal Dialysis: Does the Time Spent on Hemodialysis Impact Outcomes in Peritoneal Dialysis?

INTRODUCTION: Switch from hemodialysis (HD) to peritoneal dialysis (PD) is unfrequent, but incentive strategies to perform PD can lead to an increase of these transitions. However, data on transitioning from HD to PD are scarce. We hypothesized that time spent on HD before transfer to PD would impact PD outcomes. METHODS: This registry-based, nationwide study analyzed patients transferred from HD to PD. Patients who began HD between January 2008 and December 2016 were included. Cox and Fine and Gray regression models were used to explore the relationship between time spent on HD before PD and outcomes in PD: PD cessation for death or retransfer to HD (composite endpoint); for death; and for retransfer to HD. RESULTS: Over the study period, 1,985 of the 77,587 HD starters (3%) were transferred to PD. The median time spent on HD before transfer to PD was 1.94 months (interquartile range [IQR] 1.02-4.01). The median survival time on PD after this transition was 20 months (IQR 18-21). Time spent on HD before PD was associated with increased risk of death or retransfer to HD (cause-specific hazard ratio [cs-HR] 1.01, 95% confidence interval [CI]: 1-1.02 for a 1-month increase) and death (cs-HR 1.02, 95% CI: 1.01-1.03) but not with retransfer to HD censored on death (cs-HR 1.00, 95% CI: 0.99-1.01). The results were similar when considering competing events. DISCUSSION/CONCLUSION: Switch from HD to PD is rare in France. Time spent on HD before transfer is associated with patient survival but not with retransfer to HD.

Comparison of real-time PCR and droplet digital PCR for the detection of Xylella fastidiosa in plants.

Xylella fastidiosa (Xf) is a quarantine plant pathogen bacterium originating from the Americas and that has emerged in Europe in 2013. Xf can be detected directly on plant macerate using molecular methods such as real-time PCR, which is a sensitive technique. However, some plants may contain components that can act as PCR reaction inhibitors, which can lead to false negative results or an underestimation of the bacterial concentration present in the analyzed plant sample. Droplet digital PCR (ddPCR) is an innovative tool based on the partitioning of the PCR reagents and the DNA sample into thousands of droplets, allowing the quantification of the absolute number of target DNA molecules present in a reaction mixture, or an increase of the detection sensitivity. In this study, a real-time PCR protocol, already used for Xf detection in the framework of official surveys in the European Union, was transferred and optimized for Xf detection using ddPCR. This new assay was evaluated and compared to the initial real-time PCR on five plant matrices artificially inoculated and on naturally infected plants. In our conditions, this new ddPCR enabled the detection of Xf on all artificially inoculated plant macerates with a similar limit of detection, or a slight benefit for Quercus ilex. Moreover, ddPCR improved diagnostic sensitivity as it enabled detection of Xf in samples of Polygala myrtifolia or Q. ilex that were categorized as negative or close to the limit of detection using the real-time PCR. Here, we report for the first time a ddPCR assay for the detection of the bacterium Xf.

New Coffee Plant-Infecting Xylella fastidiosa Variants Derived via Homologous Recombination.

Xylella fastidiosa is a xylem-limited phytopathogenic bacterium endemic to the Americas that has recently emerged in Asia and Europe. Although this bacterium is classified as a quarantine organism in the European Union, importation of plant material from contaminated areas and latent infection in asymptomatic plants have engendered its inevitable introduction. In 2012, four coffee plants (Coffea arabica and Coffea canephora) with leaf scorch symptoms growing in a confined greenhouse were detected and intercepted in France. After identification of the causal agent, this outbreak was eradicated. Three X. fastidiosa strains were isolated from these plants, confirming a preliminary identification based on immunology. The strains were characterized by multiplex PCR and by multilocus sequence analysis/typing (MLSA-MLST) based on seven housekeeping genes. One strain, CFBP 8073, isolated from C. canephora imported from Mexico, was assigned to X. fastidiosa subsp. fastidiosa/X. fastidiosa subsp. sandyi. This strain harbors a novel sequence type (ST) with novel alleles at two loci. The two other strains, CFBP 8072 and CFBP 8074, isolated from Coffea arabica imported from Ecuador, were allocated to X. fastidiosa subsp. pauca. These two strains shared a novel ST with novel alleles at two loci. These MLST profiles showed evidence of recombination events. We provide genome sequences for CFBP 8072 and CFBP 8073 strains. Comparative genomic analyses of these two genome sequences with publicly available X. fastidiosa genomes, including the Italian strain CoDiRO, confirmed these phylogenetic positions and provided candidate alleles for coffee plant adaptation. This study demonstrates the global diversity of X. fastidiosa and highlights the diversity of strains isolated from coffee plants.