November 17, 2014

triclosan and tumours in mice

A study published in the journal PNAS has examined the effects of the antimicrobial triclosan in mice. The authors report an association between triclosan use and tumour prevalence. This analysis was accompanied by a roundup.

Title, Date of Publication & Journal

The commonly used antimicrobial additive triclosan is a liver tumor promoter

November 2014, PNAS

Study’s main claims – and are they supported by the data?

The paper does not prove the claim that TCS use promotes tumor growth in humans.

The authors of the paper state early in their discussion that their findings highlight a need for further animal experiments, “To date, it is challenging … to translate findings to human health”, and epidemiology studies to determine exposure in humans, “longitudinal epidemiology studies examining the correlation of TCS body concentrations and liver disease should be carried out in the future”

Within the paper there are three experiments reported:

• i) The first compares six (6) control mice to six (6) TCS dosed mice, investigating whether TCS induces liver fibrosis
• ii) The second is a match-comparison, with a sample size of three pairs (so 6 mice), investigating nuclear receptor activation by TCS
• iii) The third compares three groups, each of 30-35 mice, over six months of treatment, investigating TCS as a promoter in liver carcinogenesis in mice given a tumor initiator (diethylnitrosamine, DEN)

Strengths/Limitations

Strengths

– The authors state, there is little current research on the effects of TCS. The experiments are of novel interest at a scientific level to promote further interest in the effects of TCS

Limitations

– As made clear by the authors, there is little evidence on exposure levels in humans, so the dosage regime chosen my not relate to a clinically realistic level (~68.8 mg/kg and ~28.6 mg/kg daily for 6 months in each experiment respectively)

– Single figure for human exposure given as ~0.05 mg/kg daily (for toothpaste usage), no further comment on human exposure levels

– Low sample sizes: twelve (6 control verses 6 treated); a matched sub-experiment on three pairs (6 mice); three group comparison (30-35 mice per group)

– Only a single TCS dose level was investigated, there is no discussion on whether there is a non-linear dose response (does double the exposure lead to double the risk?)

– As for dose levels, the end point of six months is not related to human exposure levels and durations

– The explanation of the sequence of experiments is not entirely clear, specifically what was pre-planned verses ad hoc testing (was the study protocol fully pre-determined)

– Unclear whether the statistical analyses were corrected for multiple testing (e.g. Fig. 2A consists of 11 separate comparisons, only CAR is reported as statistically significant at the 5% level. However, with 11 separate tests we would expect at least one to be significant roughly 4 times out of 10. Combined with small sample sizes the evidence is probably not as strong as stated)

Glossary

Longitudinal – measurements taken on the same individuals over time (e.g. monthly blood pressure)

Epidemiology – study of pattern of disease in a populations

TCS – Triclosan [5-chloro-2-(2,4-dichlorophenoxy)phenol]

Multiple testing – a 5% significance level is incorrect (so-called Type I Error) 5% of the time (i.e. by chance we will see a significant result 1 in 20 times). If we perform twenty independent tests, we would expect at least one of them to be significant by chance. Hence, when performing many tests – so-called multiple testing- as in this paper, we must account for this.

Any specific expertise relevant to studied paper (beyond statistical)?

The reviewer is not an expert in biochemistry, animal studies, or diseases of the liver. The comments are in regard to the statistical evidence provided within the paper for any conclusions made or implied.

Before The Headlines is a service provided to the SMC by volunteer statisticians: members of the Royal Statistical Society (RSS), Statisticians in the Pharmaceutical Industry (PSI) and experienced statisticians in academia and research.  A list of contributors, including affiliations, is available here.