Abstract
An index of fatal toxicity for new psychoactive substances has been developed based solely on information provided on death certificates. An updated index of fatal toxicity (Tf), as first described in 2010, was calculated based on the ratio of deaths to prevalence and seizures for the original five substances (amphetamine, cannabis, cocaine/crack, heroin and 3,4-methylenedioxymethylamphetamine)*. These correlated well with the 2010 index. Deaths were then examined for cases both where the substance was and was not found in association with other substances. This ratio (sole to all mentions; S/A) was then calculated for deaths in the period 1993 to 2016. This new measure of fatal toxicity, expressed by S/A, was well-correlated with the index Ln (Tf) of the original reference compounds. The calculation of S/A was then extended to a group of new psychoactive substances where insufficient prevalence or seizure data were available to directly determine a value of Tf by interpolation of a graph of Tf versus S/A. Benzodiazepine analogues had particularly low values of S/A and hence Tf. By contrast, γ-hydroxybutyrate/γ-butyrolactone, α-methyltryptamine, synthetic cannabinoid receptor agonists and benzofurans had a higher fatal toxicity.
Introduction
New psychoactive substances (NPSs) have been associated with a range of serious harms in Europe, including deaths (EMCDDA, 2017). The number of fatalities associated with NPSs appears to be increasing across the United Kingdom (NISRA, 2016; NRS, 2017; ONS, 2017a). This raises the question of how we might determine their lethal toxicity compared with better-known drugs. The intrinsic fatal toxicity of a substance cannot simply be equated to the number of deaths associated with that substance; a suitable denominator of the extent of use or availability is usually required. With licensed medicines, it was found over 30 years ago that an index of fatal toxicity (T) could be created by relating the number of deaths associated with a certain substance to its availability measured by the number of prescriptions issued (King and Moffat, 1981, 1983). Since that method is not appropriate for illicit substances, a different approach is needed. Measures of availability such as law enforcement seizures, estimates of market size and prevalence, as judged by household surveys, were used to provide an alternative index of fatal toxicity (King and Corkery, 2010). However, for many NPSs there is almost no information on prevalence. There are many reasons for this, including: users can purchase them through internet websites and largely avoid interception by law enforcement agencies; new substances are not immediately included in household or other surveys of drug use in the population, in some cases because their prevalence is so low; and many users of NPSs may not know exactly what it is that they are consuming.
During the collection of data for that original index of fatal toxicity (T) based on deaths related to prescriptions (King and Moffat, 1983) it was noted that: ‘…the mortality statistics provide a second, albeit crude, measure of drug toxicity. The index T is found to be related to the proportion of deaths involving a given drug where that drug was not in association with other drugs. For example, 72 per cent of deaths associated with pentobarbitone (T = 65.8) were not associated with other drugs whereas only 23 per cent of deaths associated with diazepam (T = 1.8) did not involve other drugs.’
If we imagine a substance that has an extremely low fatal toxicity, then amongst a population of suspected fatal poisonings it will occur rarely and will mostly be associated with other more toxic substances that are the direct cause of death. If we determine the number of cases where it is the sole mention (S) on death certificates and those where it is mentioned alongside other substances (A), then the ratio S/A will be low. On the other hand, if we have a substance that is highly toxic, then its mere presence will often be associated with the direct cause of death, and the ratio S/A will be higher. Thus S/A should be a measure of the intrinsic fatal toxicity of a substance.
It could be argued that the detection of a substance in post mortem tissues is subject to many distorting factors. This is particularly true with new drugs where a toxicology laboratory may not be able to detect certain substances, perhaps because there are no suitable pure reference standards. Secondly, the toxicological examination may not be exhaustive, but allows analysis to be halted once one or more likely candidates have been found. However, by taking the ratio S/A, such selection effects should be reduced. The mention of a drug in the ‘cause of death’ field on a death certificate does not necessarily indicate that that substance was a direct cause of death; it may have been a contributory factor, or it may have been a combination of substances that led to death. Some of the fatal cases recorded by the Office for National Statistics (ONS) may not have been poisonings in the normal sense because their definition includes cases of intoxication, dependence, drug withdrawal, et cetera. (ONS, 2017a). In other words, in this paper, lethal toxicity needs to be interpreted in a broad sense. This is particularly true for cannabis, which has only rarely been the direct cause of any death, but nevertheless features in a number of deaths where no other substance was mentioned (Ghodse et al., 2007). Furthermore, these figures do not include cases where drugs may have been involved in other ways, such as impairing judgement or perception of risk (e.g. drowning, fall from height, hyperthermia, road traffic accidents), or causing psychiatric problems (e.g. suicide using mechanical means), or even homicide. The numbers of deaths attributed to specific substances are likely to be under-estimates since about 10% of UK drug poisoning deaths are recorded in ambiguous terms such as ‘multiple drug overdose’ without specifying the substances actually implicated in or contributing to death. A further difficulty, experienced in respect of NPSs, is that the name of the specific molecule may not be properly communicated to the ONS (see Table 1).
Office for National Statistics definitions of NPS categories.
There was uncertainty in some cases as to which specific molecule was meant as the ONS did not receive toxicological information and relied solely on the text given in the cause of death field on the death certificate or in any supplementary information given by the coroner, for example in their findings or ‘verdict’.
‘Aminopropyl benzofuran’ was stated in the majority of cases, but some just stated APB. The specific isomer was unknown.
‘APDB’ was all that was mentioned on the death certificate. The specific isomer was unknown.
‘MDDA’ was all that was written on the death certificate. This is presumed to mean MDDM aka MDDMA.
The term ‘piperazine’ almost certainly referred to a substituted piperazine.
‘5F-ADB’ aka ‘5F-MDMB-PINACA’.
In all cases, all that was on the death certificate was ‘5F-PB-22’. This was presumed to refer to the indole analogue.
5-EAPB: 1-(1-benzofuran-5-yl)-N-ethylpropan-2-amine; 5F-AKB-48: N-(1-adamantyl)-1-(5-fluoropentyl)-1H-indazole-3–carboxamide; 5F-PB-22: quinolin-8-yl-1-(5-fluoro pentyl)-1H-indole-3-carboxylate; AB-CHMINACA: N-[(1S)-1-(aminocarbonyl)-2-methylpropyl]-1-(cyclohexylmethyl)-1H-indazole-3-carboxamide; AH-7921: 3,4-Dichloro-N-[[1-(dimethylamino)cyclohexyl]methyl]benzamide; APB: aminopropylbenzofuran, also known as 1-(1-benzofuran-2-yl)propan-2-amine; APDB: 5-(2-aminopropyl)-2,3-dihydrobenzofuran; BZP: 1-benzylpiperazine; MDDA: 3,4-methylenedioxy-N,N-dimethylamphetamine; MDMB-CHMICA: N-[[1-(cyclohexylmethyl)-1H-indol-3-yl]carbonyl]-3-methyl-valine, methyl ester; NBOM: N-(methoxybenzyl) phenethylamine; NPS: new psychoactive substance; ONS: Office for National Statistics; TFMPP: 1-(3-trifluoromethylphenyl)piperazine; U-47700: 3,4-dichloro-N-[2-(dimethylamino)cyclohexyl]-N-methylbenzamide.
Source: personal communication to JMC from Mortality Team, ONS, 14 and 31 August 2017.
The advantage of using the ratio S/A is that it relies solely on the information on death certificates, and does not require any knowledge of availability. It could therefore be useful for evaluating the fatal toxicity of NPSs. The purpose of this paper is to test that hypothesis.
Methods
Data related to deaths registered in England and Wales between 1993 and 2016 were either extracted from the mortality statistics published by ONS (2017a) or extracted specifically for this study (ONS, 2015, 2016a,b,c, 2017b), where the selected substances were implicated in the cause of death. The principal substances examined were: amphetamine; cannabis; cocaine/crack; ecstasy; heroin; ketamine; and mephedrone. Table 1 lists those drugs that come under the ONS definition of NPS. Table 2 lists the number of deaths classified as ‘sole mentions’ (S), while Table 3 lists the ‘any mentions’ (A). In respect of deaths involving ‘any mention’, other substances present may or may not have been psychoactive. In both Tables 2 and 3, substances are included where sufficient data were available (minimum of 10 deaths). Amongst NPSs, this included AMT (α-methyltryptamine), benzodiazepine analogues, benzofurans, GHB/GBL (γ-hydroxybutyric acid/γ-butyrolactone), MPA (methiopropamine), novel amphetamines, piperazines, PMA (para-methoxyamphetamine; 4-MA; 4-methoxyamphetamine)/PMMA (para-methoxy-N-methylamphetamine), synthetic cathinones (other than mephedrone) and synthetic cannabinoids. There were insufficient data to examine the fatal toxicity of specific synthetic cannabinoids. The term ‘amphetamines’ mostly relates to amphetamine (α-methylbenzeneethanamine) itself. Although the word ‘ecstasy’ may be used rather broadly, it is taken here to mean 3,4-methylenedioxymethylamphetamine (MDMA). The mortality statistics refer to ‘heroin and morphine’ as a category. Since heroin is rapidly metabolised to morphine, post mortem analysis will rarely differentiate the two. Nevertheless, most of these deaths will have followed use of (illicit) heroin rather than (pharmaceutical) morphine, so in the present paper this is listed as ‘heroin’. Mephedrone and other synthetic cathinones as a wider class are both examined. Deaths associated with ketamine or GHB/GBL are believed to be almost entirely a consequence of the use of illicit preparations rather than prescription medicines. Finally, crack cocaine is not usually distinguished from powder cocaine on death certificates, so figures for deaths comprise both substances. For the period in question, the number of users of crack cocaine was less than 10% of all cocaine users (Broadfield, 2017).
Deaths registered in England and Wales for the period 1993–2016 where selected substances were implicated in the cause of death (sole mentions).
AMT: α-methyltryptamine; GHB/GBL: γ-hydroxybutyrate/γ-butyrolactone; MDMA: 3,4-methylenedioxymethamphetamine; MPA: methiopropamine; n/a: not available; PMA/PMMA: para-methoxyamphetamine/para-methoxy-N-methylamphetamine.
Deaths registered in England and Wales for the period 1993–2016 where selected substances were implicated in the cause of death (any mentions).
AMT: α-methyltryptamine; GHB/GBL: γ-hydroxybutyrate/γ-butyrolactone; MDMA: 3,4-methylenedioxymethamphetamine; MPA: methiopropamine; n/a: not available; PMA/PMMA: para-methoxyamphetamine/para-methoxy-N-methylamphetamine.
Table 4 shows the estimated number of last year users (aged 16–59 years) of nine substances as derived from the British Crime Survey Drug Misuse Declared (BCS), now known as the Crime Survey for England and Wales (CSEW; Broadfield, 2017). The mean estimated number of amphetamine users excludes the minor contribution of methylamphetamine, which was only recorded since 2008. In the case of ketamine, although mortality data were available for the period 1993–2016, prevalence data only started in 2006. For this reason, analysis was based on the 11 years from 2006. With mephedrone and synthetic cathinones more generally, mortality and prevalence estimates only related to the seven years 2010–2016. For piperazines and GHB/GBL, prevalence data were available only for one or two years (2010/2011 and 2011/2012; and 2011/2012 respectively).
Estimated number (thousands) of individuals aged 16–59 years who had used selected substances in the last year in England and Wales, 2001/2002–2016/2017.
n/a: not available; no earlier data available for this age-range.
GHB/GBL: γ-hydroxybutyrate/γ-butyrolactone; MDMA: 3,4-methylenedioxymethamphetamine; n/a: not available.
Sources: Broadfield, 2017; Home Office, 2012.
Seizures of selected substances made by law enforcement agencies in England and Wales for the period 2000–2015/2016 (Hargreaves and Smith, 2016a,b) are given in Table 5. Most of the data are derived from published sources, but some were specially extracted for this study. It should be noted that the statistics moved from a calendar year to a financial year in 2006/2007, hence the change in date format.
Total seizures of selected substances made by police forces and Customs & Excise/Border Force in England and Wales for the period 2000–2015/2016.
GHB: γ-hydroxybutyrate; MDMA: 3,4-methylenedioxymethamphetamine; n/a: not controlled.
Sources: Hargreaves and Smith (2016a,b); Smith (2008a,b); personal communication to JMC from J Hargreaves, Home Office Crime and Policing Analysis Unit, 24 November 2016.
Although the earlier index (King and Corkery, 2010) made use of other measures of availability, in the present study it was found that prevalence in the form of household surveys, as summarised in the BCS/CSEW, and law enforcement seizures were the best measures. Furthermore, there are no recent estimates of market size for most of the drugs examined in this study; the estimates used here can be found in King and Corkery (2010).
The index of fatal toxicity as defined by King and Corkery (2010), here named as Tf, was recalculated for each substance where data were available. The Tf was defined as the mean number of deaths per annum (sole mentions) divided by the availability of that drug, where availability could be the mean number of users, the mean number of seizures per annum, or the market size.
The calculation of Tf proceeds as follows:
If Tf (H) is the fatal toxicity of heroin, which will be defined as 1000, then:
where D(H) is the number of deaths associated with heroin (sole mentions), A(H) is the availability of heroin, measured either by the number of users, the number of seizures or some other metric, and F is a numerical factor that is unique to the particular type of availability measure.
The fatal toxicity of some other substance (Z) is then given by:
where D(Z) is the number of deaths associated with substance Z (sole mentions) and A(Z) is the appropriate availability measure of Z.
Prevalence data in the BCS/CSEW were only available from 2001/2002, so values of Tf were calculated using the average number of deaths in the 16 year period 2001–2016 or, where appropriate, for shorter periods as noted above. As before, the index was then normalised such that heroin = 1000. A Napierian logarithmic (Ln) transform was then used. This was necessary for graphical purposes because of the wide range in values of the index Tf. However, this was quite arbitrary since using the cube root of the index, for example, produced broadly similar scale compression. Tf values for seizures and market size were calculated in a similar way. For combined parameters Tf values were calculated by obtaining the mean of the indices for the individual parameters (King and Corkery, 2010), then the Naperian logarithmic transforms were applied.
Results and discussion
Table 6(a)–(e) shows the index Tf for the original five substances (amphetamine, cannabis, cocaine/crack, heroin and MDMA) based on: (a) prevalence; (b) number of seizures; (c) a combination of prevalence and number of seizures; (d) market size and, finally, (e) a combination of prevalence, number of seizures and market size. When compared with the original (King and Corkery, 2010) values of LnTf, the Pearson product moment correlation coefficients (r) for each of these five measures of the index (as LnTf) were all greater than 0.96, and values of p, the probability of the null hypothesis, were less than 0.01 in all cases. For all subsequent analyses, values of Tf were based on deaths related to the combination of prevalence and number of seizures.
Overall index (Tf) of fatal toxicity for the period 2001/2002–2016.
The overall index is the mean of the ‘sole’ and ‘any’ mention indices.
GHB/GBL: γ-hydroxybutyrate/γ-butyrolactone; MDMA: 3,4-methylenedioxymethamphetamine.
Table 7 lists values of sole (S) and any (A) deaths as well as the ratio S/A for the period 1993–2016, or, as otherwise indicated, for the reference compounds and NPS or groups of NPSs (AMT, benzodiazepine analogues, benzofurans, GHB/GBL, ketamine, mephedrone, methiopropamine, novel amphetamines, para-methoxyamphetamine/para-methoxy-N-methylamphetamine, piperazines, synthetic cathinones (other than mephedrone) and synthetic cannabinoids).
Mean values of sole (S) and any (A) mentions, and S/A where selected substances were implicated in the cause of death registered in England and Wales for the period 1993–2016 or as otherwise shown.
Values of LnTf have been interpolated from Figure 1 for those substances where LnTf was unavailable directly because of lack of or limited prevalence/seizure data (i.e. GHB/GBL and piperazines). Substances with high interpolated values of LnTf (>4) are shown emboldened, whereas substances with low values (<2) are shown in italics.
AMT: α-methyltryptamine; GHB/GBL: γ-hydroxybutyrate/γ-butyrolactone; MDMA: 3,4-methylenedioxymethamphetamine; MPA: methiopropamine; PMA/PMMA: para-methoxyamphetamine/para-methoxy-N-methylamphetamine
Figure 1 shows the relationship between the updated index Tf (based on a combination of prevalence and number of seizures; see Table 6(c)) and values of S/A (see Table 7) (r = 0.922; p < 0.01) for the seven reference compounds (amphetamine, cannabis, cocaine, heroin, ketamine, MDMA and mephedrone). The straight line is the least squares fit [LnTf = 11.12(S/A) – 0.48].
The relationship between the updated index LnTf (based on a combination of prevalence and number of seizures; see Table 6(c)) and values of S/A (see Table 7) (r = 0.923; p < 0.01) for the seven reference compounds (amphetamine, cannabis, cocaine, heroin, ketamine, MDMA and mephedrone). The straight line is the least squares fit [LnTf = 11.22(S/A) – 0.48].
Since sufficient data had been available to calculate values of Tf for ketamine and mephedrone, their inclusion in this group provides a cross-check on the indirect calculation of Tf from their S/A values using the above linear relationship. Thus, for ketamine the value of LnTf calculated from deaths/(prevalence + seizures) was 3.47 whereas the value interpolated from Figure 1 was 4.13. For mephedrone, the value of LnTf calculated from deaths/(prevalence + seizures) was 3.52 whereas the value interpolated from Figure 1 was 1.73. As will be seen from Figure 1, the high value of LnTf for mephedrone appears anomalous, whereas the value interpolated from S/A seems more consistent with other substances, and particularly with synthetic cathinones (other than mephedrone) where the interpolated value of LnTf was 1.95 (Table 7).
The results shown in Figure 1 confirm the original hypothesis that the ratio S/A is a measure of the fatal toxicity of a substance. Because of limitations in the data, as discussed below, values of S/A, as shown in Table 7, should not be over-interpreted. Many NPSs are based on what might be termed ‘amphetamine-type drugs’. It was found that substituted phenylisopropylamines had broadly similar fatal toxicities (i.e. S/A) compared with MDMA and amphetamine, but the fatal toxicity of ‘novel amphetamines’ was low. However, small differences in the estimated fatal toxicity of this group may not be significant. However, benzodiazepine analogues had particularly low values of S/A. By contrast GHB/GBL and synthetic cannabinoid receptor agonists had higher toxicities.
There are a number of limitations in the original data. For example, it is believed that prevalence, as assessed by BCS/CSEW, may underestimate the number of users of heroin and crack cocaine as well as those who do not live in normal domestic premises. For GHB/GBL and piperazine derivatives, prevalence data were only available for one and two years respectively, but as more data accumulate a future analysis should enable estimates of Tf to be made for these substances. In respect of NPSs, and despite the public concern about their toxicity, there have still been relatively few fatal poisonings, especially for ‘sole mentions’, when compared with more established drugs of misuse. The role of alcohol as a contributory factor in some of the deaths is unknown, and has not been taken into account here. (The ONS only recently started publishing data for deaths involving alcohol but no other substance (ONS, 2017a)). Finally, the fatality index treats all users of a drug as an equivalent point. That is to say, the index relates to an ‘average’ user. Those who regularly consume more of a drug on each occasion compared with another user are probably at greater risk of a fatal outcome, but there is no reliable information on what constitutes occasional/regular or heavy/light use for most of the substances considered here.
Conclusion
This analysis supports the hypothesis that the ratio S/A is a meaningful measure of fatal toxicity. Bearing in mind the limited mortality and prevalence data available for some new substances, it is suggested that the fatal toxicity of benzodiazepine analogues was particularly low. By contrast GHB/GBL, synthetic cannabinoid receptor agonists, benzofurans and AMT had higher toxicities. It is expected that the analysis shown here could be improved as more mortality and prevalence data become available.
Footnotes
Acknowledgements
We would like to thank Keith Sullivan, Senior Research Fellow in the Centre for Health Services and Clinical Research at the University of Hertfordshire, for statistical advice. Thanks are also due to the Mortality Team at the ONS for providing data specially extracted for this research project. The Home Office Crime and Policing Analysis Unit also provided unpublished data on piperazines which were extracted for this study.
Declaration of conflicting interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: LAK is a former head of the Home Office Forensic Science Drugs Intelligence Unit (1992–2001), and a former member of the Home Office Advisory Council on the Misuse of Drugs (ACMD). JMC was responsible for producing drug statistics for the Home Office (1994–2002), acted as the UK Focal Point on Drugs’ expert on drug-related deaths and mortality related to drug use (2000–2015), was a member of the ACMD’s Working Groups on Drug-related deaths (1999–2000 and 2016–2017), and is currently a co-opted member of the Technical Committee (2016 to date) and NPS Committee (2009 to date).
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.