Workshop 4 Breath odor trial design and statistics

About 25 attendees participated in this workshop. There was a wide range of views expressed – and as other convenors mentioned – more questions raised than answers found. Perhaps some of these could serve as future topics for discussion at ISBOR conferences. Any errors or omissions reflect entirely on the convenor – please do feel free to suggest additions or amendments as appropriate. Preselection of panelists & the ADA approval guidelines  There was much discussion on the question of preselecting panelists (subjects) on the basis of malodour score. The ADA Acceptance Program Guidelines indeed requires ‘the population should be at least 2 ± 0.5(mild malodor) on an intensity scale of 0–5, or a similar level on other appropriate scales’. Thus for compliance with the guidelines, some degree of screening is probably necessary. Some suggested that ‘a product to solve a problem, needs to have a problem to solve’; others preferred the ‘a product for everyman, everyday’ approach. Furthermore, the ADA guidelines also require ‘In the clinical evaluations, 80% of the subjects shall demonstrate a reduction to a 0 or 1 rating based on a 0–5 organoleptic intensity rating...’ Since companies will be keen to gain ADA Acceptance, there is the possibility that study designs could be biased toward recruiting panellists with a minimum acceptable score (2 ± 0.5), with a view to complying with the reduction to 1 or below, post‐treatment. For example, recruitment of a panellist with a resting score of 3 would increase the risk of a product failing to gain acceptance. Some concern was expressed that since the ADA guidelines simply state ‘Depending on the claims being made, oral malodor measurements should be performed at a minimum of two appropriate time periods after baseline during the three‐week test period’ that potentially almost any product could work – if very short time intervals were selected. Some participants suggested that such concerns could be addressed by ‘levelling the playing field’– even if this was achieved ‘artificially’– for example by cysteine rinses etc. Some commented that sulphur compounds do not represent ‘all’ of oral malodour; others commented that only VSC measurements consistently correlated with organoleptically‐perceived malodour. This then is a question that clearly calls for further debate. General statistical issues  A crossover trial design – where each individual acts as their own control, reduces the possibility of inter‐individual variations influencing study outcome (as could be the case in a parallel design). As it is well known in oral malodour research (as well as in most other such clinical studies), that inter‐ subject variation is much greater than intra‐ subject variation, a crossover study design is inherently more powerful than a parallel design for the same number of subject × sample combinations. Crossover designs do, however, require an appropriate washout period (see below). Pretrial procedures  What should be the oral hygiene procedures preceding panellists taking part in a trial? How long a gap should there be prior to trials since the last oral hygiene routines? What restrictions on diet and other oral‐related habits should be imposed? Someone suggested that screening should establish a baseline malodour score over time. Are there known circadian rhythms in malodour? Breath odour scoring on the 0–5 scale  Noted that a score of 0, or of 5 is extremely rare (<0.5% of the population). Likewise, scores of 4 are probably <5%. So for the most part the 0–5 scale is actually a 3‐point scale – 1, 2 or 3. Some workers have thus used ½‐point scales, as a useful and pragmatic solution to the conflicting objectives of encompassing a full odour intensity range in a 0–5 scale, whilst maintaining sufficient sensitivity to distinguish between samples. Other workers stick to the integer‐scale. The issues around statistical treatment of the data were discussed. Some suggested that a ‘pseudo‐continuous’½‐point scale would allow statistical methods such as ANOVA to come into play. Others preferred to stick to medians, and then to use non‐parametric statistical methods. There is an entire literature on sensory perceptions of odour/malodour – maybe this should be something that ISBOR and other interested societies could pursue further. Appropriate washout periods  Participants debated the requirement for appropriate washout periods (and products) to be used in efficacy trials, i.e., to standardize usage of other oral care products between use of products within a trial. However, the appropriate washout period for products would very much depend on their efficacy and/or their mode of action. For a regular oral care product such as a toothpaste, 1 week might be an appropriate timescale; a strong antimicrobial product might require a longer washout period. In the case of probiotics, where the potential for activity might stretch almost indefinitely – what washout period would be appropriate? Some of these issues can be addressed by various statistical methods – the study design can ensure that all products are tested equally often in each relevant time period (to avoid issues such as changes in weather, for example). This is often achieved via a Latin‐Square type design. Carry‐over effects can be balanced by use of a Williams Latin square design, which balances the carryover between successive products by ensuring every product follows every other product with equal frequency. Panellist numbers  Some participants wanted to know the numbers of panellists that would be appropriate for trials. This clearly would depend on a number of factors, especially the statistical design of the trial. Other factors which would be important include (i) the relative efficacy of the products tested (compared with controls), (ii) the initial malodour scores of panellists, (iii) the magnitude of any differences which the study aims to detect. Other participants reported that their typical trials have used 10–15, or 20–30 panellists.