Monte Carlo approaches for determining power and sample size in low-prevalence applications

Animal and Plant Health Inspection Service, USDA 2150 B Centre Avenue, Mail Stop 2E6, Fort Collins, CO 80526, USA.
Preventive Veterinary Medicine (Impact Factor: 2.17). 12/2007; 82(1-2):151-8. DOI: 10.1016/j.prevetmed.2007.05.015
Source: PubMed


The prevalence of disease in many populations is often low. For example, the prevalence of tuberculosis, brucellosis, and bovine spongiform encephalopathy range from 1 per 100,000 to less than 1 per 1,000,000 in many countries. When an outbreak occurs, epidemiological investigations often require comparing the prevalence in an exposed population with that of an unexposed population. To determine if the level of disease in the two populations is significantly different, the epidemiologist must consider the test to be used, desired power of the test, and determine the appropriate sample size for both the exposed and unexposed populations. Commonly available software packages provide estimates of the required sample sizes for this application. This study shows that these estimated sample sizes can exceed the necessary number of samples by more than 35% when the prevalence is low. We provide a Monte Carlo-based solution and show that in low-prevalence applications this approach can lead to reductions in the total samples size of more than 10,000 samples.

Download full-text


Available from: Michael S Williams, Sep 29, 2014
  • Source
    • "The minimum, non-zero prevalence in small flocks may be greater than PstarA; e.g. the smallest non-zero prevalence in a flock with 4 animals is 25%. Additionally, small flocks play a special role at low prevalence levels, such as at the time of disease incursion into a population [14]. Therefore, a large proportion of small flocks within a population has an impact on the sensitivity of a SSC. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Globalization and subsequent growth in international trade in animals and animal products has increased the importance of international disease reporting. Efficient and reliable surveillance systems are needed in order to document the disease status of a population at a given time. In this context, passive surveillance plays an important role in early warning systems. However, it is not yet routinely integrated in the assessment of disease surveillance systems because different factors like the disease awareness (DA) of people reporting suspect cases influence the detection performance of passive surveillance. In this paper, we used scenario tree methodology in order to evaluate and compare the quality and benefit of abortion testing (ABT) for Brucella melitensis (Bm) between the disease free situation in Switzerland (CH) and a hypothetical disease free situation in Bosnia and Herzegovina (BH), taking into account DA levels assumed for the current endemic situation in BH. The structure and input parameters of the scenario tree were identical for CH and BH with the exception of population data in small ruminants and the DA in farmers and veterinarians. The sensitivity analysis of the stochastic scenario tree model showed that the small ruminant population structure and the DA of farmers were important influential parameters with regard to the unit sensitivity of ABT in both CH and BH. The DA of both farmers and veterinarians was assumed to be higher in BH than in CH due to the current endemic situation in BH. Although the same DA cannot necessarily be assumed for the modelled hypothetical disease free situation as for the actual endemic situation, it shows the importance of the higher vigilance of people reporting suspect cases on the probability that an average unit processed in the ABT-component would test positive. The actual sensitivity of passive surveillance approaches heavily depends on the context in which they are applied. Scenario tree modelling allows for the evaluation of such passive surveillance system components under assumed disease free situation. Despite data gaps, this is a real opportunity to compare different situations and to explore consequences of changes that could be made.
    BMC Veterinary Research 01/2009; 4(1):52. DOI:10.1186/1746-6148-4-52 · 1.78 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: If the prevalence of bovine spongiform encephalopathy (BSE) varies among cohorts within a population, stratified analysis of BSE surveillance data may allow identification of differences in BSE exposure that are important with respect to the design and evaluation of disease prevention and control measures. In low BSE prevalence populations, however, surveillance at levels that meet or exceed international guidelines may provide insufficient statistical power to distinguish prevalence levels among cohorts. Furthermore, overstratification to account for hypothetical variability in the population may inflate uncertainty in BSE risk estimates.
    Preventive Veterinary Medicine 04/2008; 83(3-4):337-46. DOI:10.1016/j.prevetmed.2007.09.003 · 2.17 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: This paper shows how Monte Carlo simulation can be used for sample size, power or precision calculations when planning medical research studies. Standard study designs can lead to the use of analysis methods for which power formulae do not exist. This may be because complex modelling techniques with optimal statistical properties are used but power formulae have not yet been derived or because analysis models are employed that divert from the population model due to lack of availability of more appropriate analysis tools. Our presentation concentrates on the conceptual steps involved in carrying out power or precision calculations by simulation. We demonstrate these steps in three examples concerned with (i) drop out in longitudinal studies, (ii) measurement error in observational studies and (iii) causal effect estimation in randomised controlled trials with non-compliance. We conclude that the Monte Carlo simulation approach is an important general tool in the methodological arsenal for assessing power and precision.
    Statistical Methods in Medical Research 04/2012; 22(3). DOI:10.1177/0962280212439578 · 4.47 Impact Factor
Show more