Evaluation of historic in vivo data to characterise the emetic properties of liquid cleaning products and provide a framework for the development of an in silico predictive algorithm

Highlights

• Nausea and vomiting are frequently reported symptoms of accidental ingestion of xenobiotics.

• Analysis of historic in vivo studies identified emetic properties of liquid cleaning products.

• A high percentage of non-ionic surfactants/high ionic strength, and pH of ~10 are linked to emesis.

• A mechanism based on activation of gut enterochromaffin cells and vagal afferents is presented.

• We discuss using historical data to make in silico models as an alternative (3Rs) to animal testing.

Abstract

Accidental ingestion of household cleaning products frequently results in emesis but the physicochemical properties responsible are not known. To investigate whether data collected during in vivo animal studies performed >30 years ago could provide novel insights into the components responsible, we re-analysed original studies from a total of 74 liquid cleaning formulations. The incidence of emesis was dose-related with ED₅₀ values between 0.012 and 8.4 ml/kg and 57% of formulations having an ED₅₀ ≤ 1 ml/kg. The median latency for emesis was 10.0 min (95% CI, 8–12 min) and number of vomits in 60 min ranged from 1 to 10 (median 2). From the ED₁₀₀, latency and number of vomits we derived a “vomiting index” (VI) for a subset of 15 formulations which revealed an association between a high VI, a high percentage of non-ionic surfactants/high ionic strength, and a pH of ~10 which we propose are causally linked with the possible mechanism(s) discussed. The limitations of using historic data are discussed but analysis of such data has provided novel insights into the emetic characteristics of this class of products and has informed the development of an in silico model to predict the emetic liability of novel formulations without additional in vivo studies.

Introduction

Nausea (the unpleasant aversive sensation associated with the urge to vomit; see Stern et al., 2011 for detailed definitions) and vomiting (the forceful oral expulsion of gastric contents) are components of the body's hierarchically organised defences against toxins accidently ingested with food or fluids (Davis et al., 1986). Although this system evolved to help defend animals against ingested toxins, the pathways activated by such toxins can also be activated by synthetic substances, including therapeutic agents, e.g. the anti-cancer agent cisplatin (Andrews and Rudd, 2015), some of which have no obvious natural counterpart. This illustrates the remarkable adaptive plasticity of this detection system.

Because changing lifestyles bring humans into contact with relatively high concentrations of substances, which are otherwise unlikely to have been encountered in the natural world, manufacturers of such products need to be aware of their potential to produce such a response should unintentional exposure occur, particularly by ingestion. One such example is household cleaning products, which are ranked in the top five classes of substance involved in overall human exposure (7.28%) and in children aged five or under (10.70%) with vomiting being one of the most common symptoms following ingestion (Gummin et al., 2019). While the occurrence of an emetic response to such cleaning products is not new (Snyder et al., 1964; Griffith et al., 1969; Smyth and Calandra, 1969; Weaver and Griffith, 1969), the accidental exposure to liquid detergents, particularly in the form of laundry detergent packets or pacs, has increased in recent years. For example, data from the U.S. National Poison Data System (Valdez et al., 2014; Centers for Disease and Prevention, 2012) and the UK National Poisons Information Service (Williams et al., 2012a, Williams et al., 2012b) have identified ingestion of laundry detergent pacs as a risk in the paediatric population with vomiting being the most common symptom (Do et al., 2015). The majority of cases of accidental oral exposure have been reported in younger children; 96% in age 5 or lower in the UK (Williams et al., 2012a, Williams et al., 2012b) and 73.5% in children under 3 years of age in the USA (Valdez et al., 2014). Nausea is also reported but at a much lower incidence than vomiting, e.g. 4.9% vs. 56% (Valdez et al., 2014). However, this statistic should be treated with caution because nausea is a self-reported sensation so reporting may be problematic in younger children (Richards and Andrews, 2004; Benninga et al., 2016). Whilst the focus has been on children it should not be overlooked that adults with cognitive impairment may also accidently ingest cleaning products (Janeway, 2017) and there are also reports of voluntary ingestion in juveniles as a “challenge” (https://en.wikipedia.org/wiki/Consumption_of_Tide_Pods).

There is limited published data on emesis induced by household cleaning products in humans and preclinical models (Gieseker and Troutman, 1981), in contrast to extensive data on emesis induced by therapeutic agents (Percie du Sert et al., 2011). Limited published data, such as that of Weaver and Griffith (1969) and historical company knowledge (e.g., Procter & Gamble Company), indicates that such studies were being performed particularly prior to 1980 as part of product safety assessment. Prompted by growing concern about the effects of accidental ingestion of detergents, including nausea and vomiting, particularly in the form of laundry detergent pacs, and the need to better understand the emetic characteristics of such formulations to inform risk management strategies, we re-examined historic unpublished data of preclinical (canine) studies undertaken more than thirty years ago that investigated the emetic potential of a range of orally administered liquid cleaning formulations. Canine emesis studies for household product testing were relatively common at one time but are no longer undertaken such that analysis of historic data represents a unique opportunity to gain novel insights into the emetic characteristics of these products which typically contain a mixture of chemical classes such as anionic/cationic/non-ionic surfactants, chelants, solvents and polymers.

The potential contribution of analysis of archival commercial data and the published literature to develop predictive models, obviating the need for further animal studies, was noted in an analysis of the application of the 3Rs (Replacement, Reduction, Refinement) principles to studies of emesis (Holmes et al., 2009). The potential application of this approach to modelling emetic liability for pharmacologically active agents has also been demonstrated (Percie du Sert et al., 2012). To develop in silico models to predict emetic liability, it is desirable to access large data sets covering a range of substances and formulations evaluated under similar experimental conditions. In this regard, a first step is to critically review the data set in an unbiased manner.

This paper reports a detailed analysis of historic archived data from Procter & Gamble studies included in the safety assessment of liquid cleaning products intended for household use. All studies analysed in this paper were conducted over 30 years ago when such studies were the standard of practice to assess the safety of cleaning technologies and whilst the data informed product development at the time, such studies are no longer conducted by P&G in line with their commitment to eliminating the use of animals and because the endpoint (i.e., the potential to cause emesis) was relatively well known for such products.

We hypothesised that detailed analysis of historic in vivo data would enable characterisation of the emetic response to liquid cleaning products allowing identification of key features inducing emesis and potential mechanisms. If correct, this allows us to further hypothesise that historic in vivo data could be used to predict emetogenicity of novel formulations obviating animal use. Data was gathered to test the hypothesis under the following specific aims: (i.) To extract and analyse data from a sample of archived studies to more fully characterise the emetic responses of a diverse range of liquid cleaning formulations; (ii.) Compare the emetic profile of the formulations with data in the published literature; (iii.) Identify common features amongst the formulations which may influence the emetic response; and, (iv.) Assess the potential utility of archival data for development of a model to predict the emetic liability of novel products without the use of animals. The latter is reported in an accompanying paper (Li et al., 2020; in preparation).