Senior Research Fellow
Research Gate: https://www.researchgate.net/profile/David_Mclernon
Methodologically, I am interested in the development of prognostic models using advanced survival analysis methods such as accelerated failure time models, and the validation methods used to assess their predictive ability. My PhD involved the follow-up of patients who have had their initial liver function tests (LFTs) in primary care, and who had no clinically obvious liver disease. Prediction models using accelerated failure time methods were derived and assessed to predict risk of liver disease, liver mortality and all cause mortality at various time points. This involved large observational datasets and record-linkage procedures. Prediction modelling, as a whole, is important because they extend our knowledge in clinical areas using retrospective or prospective data. They are used commonly in primary and secondary care, in combination with clinical knowledge, to inform which patients require an intervention e.g. the Framingham equation.
More recently, I am interested in extensions of standard clinical prediction models such as dynamic prediction and stratified medicine. The latter is used to estimate individualized (i.e. specific to the patient's characteristics) predictions of the absolute increased benefit of treatment. I intend to explore these methods in my CSO funded postdoctoral fellowship (see Current Research for details).
In January 2012 I was awarded a Chief Scientist Office Postdoctoral Fellowship in Health Services and Health of the Public Research supervised by Prof Bhattacharya and Prof Lee. The fellowship aims to develop clinical prediction models that can estimate the probability of pregnancy outcomes in couples attending a fertility clinic. A brief summary follows:
Development and validation of prognostic models for subfertility
Subfertility is defined as the inability to conceive within one year of unprotected vaginal intercourse. Subfertility is a large public health burden, affecting 1 in 5 Scottish couples. Debate exists on what characteristics determine whether and when a subfertile couple should be treated. Treatment for those with a high chance of conceiving naturally may not be cost-effective and result in undue stress. Conversely, those with a low chance of spontaneous conception need immediate treatment. Whilst clinical prediction models exist that estimate the chances of treatment independent pregnancy or treatment dependent pregnancy, none take into account both outcomes. Predicting the chance of pregnancy following IVF treatment is also important since couples with different characteristics e.g. different ages, type of infertility, time spent trying to conceive etc. will have different chances of conception. Some may require IVF sooner while others may need more cycles of IVF treatment. The aims of this study are: 1. To develop and validate prediction models to identify who needs fertility treatment and when (Outcome Prediction In Subfertility (OPIS) Grampian); and 2. To develop clinical prediction models to predict pregnancy outcomes in women considering IVF treatment in the UK (OPIS IVF).
For the OPIS Grampian models, world renowned databases from the Aberdeen Fertility Centre (AFC) and Aberdeen Maternity and Neonatal Databank will be utilised to identify all subfertile couples living in Grampian, who attended the Centre from 1992 to 2011. These databases contain important baseline predictors, treatment information, and outcomes (natural or treatment based pregnancy, live birth). Cox proportional hazards models will be used to estimate the chance of treatment independent pregnancy within different follow-up time periods. A stratified medicine modelling approach will be considered to examine the absolute difference in pregnancy chances between women who undergo treatment and those who undergo expectant management. By calculating this absolute difference for women with the same baseline characteristics, the clinician can decide whether it is beneficial to treat or not. In order to inform clinical decisions regarding when to treat, a dynamic prediction model using landmarking will be used. This involves fitting sequential Cox models predicting treatment independent pregnancy over the following year from incremental time origins from baseline. This enables one to determine the change in the predicted probability of pregnancy the longer treatment is delayed. For OPIS IVF, models predicting treatment related outcomes (including preterm birth) for separate causes of subfertility following IVF will be derived using UK wide data from the Human Fertilisation and Embryology Authority. To adjust for multiple IVF cycles a Poisson regression model will be used considering event time to be the number of cycles until success. This removes the need to account for within-subject correlation because the woman is the unit of observation, rather than the cycle. Predictive ability for all models will be assessed using discrimination and calibration tests. For OPIS Grampian, external validation will be performed using an independent cohort of Dutch subfertile patients. A web-based clinical prediction aid will be created to facilitate clinical management of subfertile couples. Such models will help clinicians and couples make individual decisions about whether fertility treatment would be beneficial and when it should be offered.
I was awarded the inaugural Backett Weir Russell Career Development Fellowship in November 2010, an internal competitive award funded by Professor Roy Weir. This Postdoctoral Fellowship provided the opportunity to publish papers from my PhD and to further investigate my methodological interests (focussing on survival modelling techniques in particular). The award also facilitated the development of my external Postdoctoral Fellowship application to the CSO. Without the kind donation by Professor Weir it would have been unlikely that I would have had the protected time to apply for this funding.
I have developed collaborations with several key researchers and clinicians from The Netherlands. Prof Ewout Steyerberg is Professor of Medical Decision Making at the Department of Public Health who is world-renowned for his research in prognostic modelling and is author of the textbook ‘Clinical Prediction Models’. Professor te Velde is an Emeritus Professor from the Department of Public Health who has written multiple papers on prognostic modelling in fertility medicine. Prof Ben Mol is Professor of Obstetrics and Gynaecology at the University of Adelaide. With these eminent Professors (including Prof Bhattacharya) I have recently published an opinion paper on Clinical prediction models to inform individualized decision-making in subfertile couples: a stratified medicine approach.
Chief Scientist Office Postdoctoral Fellowship in Health Services and Health of the Public Research £172,515; Jan 13 to Dec 15
“Development and validation of prognostic models for subfertility”.
David McLernon, Siladitya Bhattacharya (Supervisor), Amanda Lee (Supervisor)
Backett Weir Russell Career Development Fellowship £138,713; Nov 10 to Dec 12
“The use of accelerated failure time models for clinical prediction”.
David McLernon, Marion Campbell (Mentor)
Health Technology Assessment, NHS £249,377; Feb 2005 to Feb 2008
"Development of a decision support tool to facilitate primary care management of patients with abnormal liver function tests without clinically apparent liver disease" (03/38/02)
Donnan PT, Dillon JF, McLernon D, Steinke D, Ryder S, Roderick P, Sullivan F, Rosenberg W.
- Course Co-ordinator and lecturer in the Intermediate Statistics Course for University/external staff on survival analysis
- I currently lecture on t-tests and ANOVA twice a year in the Postgraduate Basic Statistics course
- Co-ordinator for the Introduction to SPSS courses for both staff and BSc Medical Sciences students
- Lecturer in the online PGCert Applied Statistics course.
- Further Info
- Member of the Royal Statistical Society Highlands Local Group Committee.
- Member of the College Ethical Review Board at the University of Aberdeen.
- Organised a ‘Clinical Prediction Modelling’ workshop held on 6th June 2014 which was attended by 30 members of staff across the division of applied health sciences. I invited eminent guest speakers, Prof Ewout Steyerberg and Prof Egbert te Velde from Erasmus Medical Center, Rotterdam to give three lectures on prognostic modelling. The day included presentations from five researchers within the Division (including myself) followed by a discussion hour.
- Member of the College Ethics Review Board (May 2010-).
- Member of the Reproductive Medicine Clinical Studies Group (Oct 2014-)
- Member of the Royal Statistical Society Highlands Local Group Committee (Oct 2010-). I have organised and chaired a talk from Professor Dave Collett, NHS Blood and Transplant, on informed censoring in survival analysis.
- Young Statisicians Group respresentative for the Royal Statistical Society Highlands Local Group Committee (Feb 2014-).
- Member of Health and Data Linkage in North East Scotland (HEADLINES) group (June 2013-).
- Member of Grampian Data Safe Haven Governance subgroup (July 2013-).
- Organised and chaired the Foresterhill Statistics Group meetings at the University of Aberdeen (June 2008-May 2011). Internal and external speakers are invited to talk on various aspects of statistics. Staff and Postgraduate students from the University with an interest in statistics are invited to attend, and our distribution list currently stands at 30 names.
PhD in Health Services Research, University of Dundee, 2008
MPhil in Medical Statistics, Queen's University, Belfast, 2002
BSc Hons in Applied Mathematics, Statistics and Operational Research (First class), Queen's University, Belfast, 1999