Dealing with Missing Outcomes: Lessons from a Randomized Trial of a Prenatal Intervention to Prevent Early Childhood Caries
Kamila Plutzera, *, Gloria C Mejiaa, b, A. John Spencera, Marc J.N.C Keirsec
a Australian Research Centre for Population Oral Health, School of Dentistry, The University of Adelaide, Adelaide, South Australia
b Department of Preventive and Restorative Dental Sciences, School of Dentistry, The University of California, San Francisco, USA
c Department of Obstetrics, Gynaecology and Reproductive Medicine, Flinders University, Adelaide, South Australia
Severe early childhood caries (S-ECC) affects 17% of 2-3 year old children in South Australia impacting on their general health and well-being. S-ECC is largely preventable by providing mothers with anticipatory guidance. Randomised controlled trials (RCTs) are the most decisive way to test this, but that approach suffers from near inevitable loss to follow-up that occurs with preventative strategies and distant outcome assessment.
We re-examined the results of an RCT to prevent S-ECC using sensitivity analyses and multiple imputation to test different assumptions about violation of random allocation (1%) and major loss to follow-up (32%). Irrespective of any assumptions about missing outcomes, providing expectant mothers with anticipatory guidance during pregnancy and in the child’s first year of life, significantly reduced the incidence of S-ECC at 20 months of age. However, the relative risk of S-ECC varied from 0.18 (95% confidence interval (CI): 0.06 – 0.52) to 0.70 (95% CI: 0.56 – 0.88). Also the ‘number needed to treat’ (NNT) to prevent one case of S-ECC varied 2.5-fold: from 8 to 20 women given anticipatory guidance. Multiple imputation provided a best estimate of 0.25 (95% CI: 0.11 – 0.56) for the relative risk and of 14 (95% CI: 10 – 33) for the number needed to treat.
Avoiding loss to follow-up is crucial in any RCT, but is difficult with preventative health care strategies. Instead of abandoning randomisation in such circumstances, sensitivity analyses and multiple imputation can consolidate the findings of an RCT and add extra value to the conclusions derived from it.
Key Words: Health promotion, early childhood caries, randomized controlled trial, multiple imputation, intention-to-treat, number needed to treat, sensitivity analysis, pregnancy, Zelen design.
open-access license: This is an open access article licensed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0/) which permits unrestricted, non-commercial use, distribution and reproduction in any medium, provided the work is properly cited.
* Address correspondence to this author at the Australian Research Centre for Population Oral Health, School of Dentistry, The University of Adelaide, SA 5005, Australia; Tel: +61 8 8303 3292; Fax: +61 8 8303 4858; E-mail: email@example.com