It has be en exactly 40 years since the seminal paper introducing multiple imputation to handle the problem of missing data was published. Since then, this appr oach has led the way in developing principled statistical methodology to ad dress this problem. The approach has also proved to be widely appealing, wi th uptake popular in many disciplines. In this session we aim to bring toge ther state of the art developments in multiple imputation research in the a rea of medical statistics. In particular, we will look at the use of Multip le Imputation beyond in RCTs. The session speakers comprise some of th e world leaders and experts in the field.

**15:00 – 15:40**

**Speaker: **Ian White, University College London

**Title:** Multiple imputation: the universal panacea
, and its limitations

**Abstract: Ian will review some of the developme
nts in theory and software that led to multiple imputation being seen by so
me people as a universal solution to missing data problems. He will then ex
plain why it isn’t, discussing alternatives to multiple imputation, difficu
lties of imputing multilevel data, and avoiding the untestable missing at r
andom assumption.**

** **

**15:40 – 16:20**

**<
strong>Speaker:** Tra Pham, University College London

**
Title:** Population-calibrated multiple imputation for a binary/cate
gorical covariate in categorical regression models

Abstract: Multiple imputation (MI) has become popular for analyses with missing data in medic al research. The standard implementation of MI is based on the assumption o f data being missing at random (MAR). However, for missing data generated b y missing not at random (MNAR) mechanisms, MI performed assuming MAR might not be satisfactory. For an incomplete variable in a given data set, its co rresponding population marginal distribution might also be available in an external data source. We show how this information can be utilised in the i mputation model to calibrate inference to the population by incorporating a n appropriately calculated offset termed the "calibrated-? adjustment". We describe the derivation of this offset from the population distribution of the incomplete variable and show how, in applications, it can be used to cl osely (and often exactly) match the post-imputation distribution to the pop ulation level. Through analytic and simulation studies of a binary/categori cal covariate in categorical regression models, we show that our proposed c alibrated-? adjustment MI method can give the same inference as standard MI when data are MAR, and can produce more accurate inference under two gener al MNAR mechanisms. The method is used to impute missing ethnicity data in a type 2 diabetes prevalence case study using UK primary care electronic he alth records. Calibrated-? adjustment MI represents a pragmatic approach fo r utilising available population-level information in a sensitivity analysi s to explore potential departures from the MAR assumption.

**16:20 – 16:50:** Break for refreshments

< p>

**Speaker**: James Carp
enter, London School of Hygiene and Tropical Medicine

**16:50 –
17:30**

**Speaker:** James Carpenter, London Sch
ool of Hygiene and Tropical Medicine**Title:** Sensitivi
ty analysis for missing trial outcomes: what can it do for you?

Abstr
act:** **James Carpenter (1,2) and Suzie Cro (3)James Car
penter (1,2) and Suzie Cro (3)

1. Department of Medical Statistics, L
ondon School of Hygiene & Tropical Medicine

2. MRC Clinical Trials
Unit at UCL

3. Imperial Clinical Trials Unit, Imperial College London

Missing outcome data are almost inevitable in clinical trials, for e
xample due to inter-current events such as treatment withdrawal, treatment
switching or loss to follow-up. In such settings, the analysis can only pro
ceed on the basis of an untestable assumption about the missing outcome dat
a. In applications, it is therefore important to understand the robustness
of conclusions to a range of plausible assumptions about the distribution o
f the missing outcomes.

In this talk, we outline two approaches for th
is: (a) reference based imputation, where missing outcomes are imputed by r
eference to other patient groups [1], and (b) eliciting expert opinion on t
he distribution of missing values and incorporating this into the analysis
[2]. For each approach, we discuss the assumptions made about the missing d
ata, implementation using multiple imputation, and give an illustrative app
lication. We conclude with a discussion of the pros and cons of each approa
ch, and how they may be used to address some of the challenges raised by th
e ICH-E9 addendum on estimands.

**References:**

[1]
Cro, S., Carpenter, J. R. and Kenward, M. G. (2018) Information-anchored s
ensitivity analysis: theory and application. Journal of the Royal Statistic
al Society, Series A. https://rss.onlinelibrary.wiley.com/doi/epdf/10.111
1/rssa.12423

[2] Mason, A. J., Gomes M., Grieve, M. Ulug, P., Powe
ll, J. T. and Carpenter J. R. (2017). Development of a practical approach t
o expert elicitation for trials with missing health outcomes: application t
o the IMPROVE trial. Clinical Trials, 14, 357-367. https://doi.org/10.1177/1740774517711442

**17:30 – 18:00**, Panel discussion

Tea and coffee wi ll be available from 2.30pm

Registration is required - the event is f ree to RSS Fellows with a £25 registration fee for non-Fellows

X-EXTRAINFO:The meeting will be preceded by the Annual General Meeting of the Medical S ection\n DTSTAMP:20190626T213402Z DTSTART;TZID=Europe/London:20181204T150000 DTEND;TZID=Europe/London:20181204T180000 SEQUENCE:0 TRANSP:OPAQUE END:VEVENT END:VCALENDAR