Providing evidence based information on drugs
Click me to open the menu
Click me to close the menu

Lethal toxicity of new psychoactive substances

Les King      

31st July 2018

Research just published in the Journal of Psychopharmacology (REF 1) describes a novel method for assessing the fatal toxicity of new psychoactive substances (NPS). Although various methods have previously been described for determining the relative harms of drugs, they rely on knowing not just how many deaths are attributable to a substance, but some measure of population exposure, such as number of prescriptions for medicinal products or number of users, number of law enforcement seizures or market size. However, for NPS (formerly and misleadingly known as ‘legal highs’) there is almost no information on the user population, and nor is there suitable information on what constitutes recreational and lethal doses meaning that a therapeutic index cannot be calculated. The approach now described relies only on the information provided on death certificates.

Drug-related deaths in England and Wales are published annually by the Office for National Statistics (ONS; REF 2). They are divided into those where the substance in question was the sole mention on the death certificate and those where it was found at post-mortem to be associated with other substances.

Over 30 years ago, a chance observation noted that drugs with a known high toxicity were often found not to be associated with other substances at post mortem (REF 3). On the other hand, drugs with a known low toxicity often occurred in the presence of other substances. This suggested the following hypothesis:

”Substances with a low fatal toxicity may appear on ‘all mention’ death certificates (A), but will rarely be ‘sole mentions’ (S). Thus, a low ratio S/A will indicate a low fatal toxicity. Substances with a high fatal toxicity will often appear alone, thus S and A will be similar, and S/A will approach unity.”

The purpose of the present paper was to test this hypothesis. The validity of the ratio S/A was first compared with Tf, a previously-determined measure of lethal toxicity (REF 4). Using data from ONS for the period 1993 to 2015, the index Tf ,was calculated for a group of reference substances as the ratio of deaths to prevalence for those substances where the number of users and number of seizures were known. Because of the wide spread of values of Tf, it was found convenient to use a logarithmic transform (Ln) as a scale-compression device. Figure 1 shows the relationship between S/A and Ln(Tf) for seven substances (amphetamine, cannabis, cocaine/crack, heroin, ketamine, MDMA and mephedrone).

Figure 1. The relationship between Ln(Tf) and S/A for seven reference substances.

The correlation coefficient (r = 0.92) between S/A and Ln(Tf) is statistically significant; the probability (P) that this arose by chance is less than 1%. A further check on the validity of S/A as a measure of fatal toxicity, was made by examining S/A values for established medicinal drugs. Again, using data from ONS, it was found that the highest values occurred with barbiturates followed by tricyclic antidepressants (e.g. amitriptyline, dothiepin), then a broad group of narcotic analgesics (e.g. fentanyl, methadone), followed by a miscellaneous group (e.g. zopiclone, pregabalin). Diazepam had the lowest value of S/A. This order would come as little surprise to toxicologists, and therefore supports the use of S/A.

Figure 1 can now be interpolated using values of S/A for NPS where no prevalence data are available. Figure 2 shows the results for a group of NPS where Ln(Tf) has been calculated from S/A. For comparison, the substances shown in Figure 1 have also been included for reference.

Figure 2. The index of fatal toxicity Ln(Tf) for a group of NPS and reference substances. AMT = α-methyltryptamine; SCRAs = synthetic cannabinoid receptor agonists; GHB/GBL = γ-hydroxybutyrate/γ-butyrolactone; PMA = para-methoxyamphetamine; PMMA = para-methoxy-N-methylamphetamine; MPA = methiopropamine.

Because of the relatively few numbers of deaths associated with NPS, Figure 2 should not be over-interpreted; the difference between adjacent substances is probably not significant. However, it is useful to comment on the most and least fatally toxic substances. Thus, the appearance of GHB/GBL and PMA/PMMA near the top is consistent with what we know about these substances. The high score for SCRAs is a much more original finding, but it fits with the many anecdotal reports and publications that have described the dangerous properties of these cannabis substitutes, particularly those most recently encountered. It is therefore ironic that cannabis itself scores so low. Benzodiazepine analogues represent a further group with low scores. That finding is consistent with the well-known low toxicity of established benzodiazepines such as diazepam.

This analysis is subject to several limiting factors associated with the mortality statistics. For example, the ONS data rely on the accuracy and completeness of the information provided by toxicologists to the coroners. This is particularly true with new drugs where a laboratory may not be able to detect certain substances because there are no suitable pure reference standards. Furthermore, 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. Some of the fatal cases recorded by ONS may not have been poisonings in the normal sense because their definition includes cases of intoxication, dependence and drug withdrawal. 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 and 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 implicated. A further difficulty, experienced in respect of NPS, is that the name of the specific molecule may not be properly communicated to ONS. In other words, in this paper, lethal toxicity needs to be interpreted in a broad sense.

Despite these caveats, it is expected that the correlation between S/A and Ln(Tf) shown in Figure 1 will improve as more data become available. Despite public concern, the number of deaths associated with NPS is still low, especially when compared with more established drugs of misuse. It is hoped that this analysis can be repeated using data from other countries.


[1] L. A. King and J. M. Corkery (2018) An index of fatal toxicity for new psychoactive substances. J. Psychopharmacol. 32(7): 793-801. e-publication ahead of print:

[2] ONS (2017) Deaths related to drug poisoning in England and Wales: 2016 registrations, 2 August. Office for National Statistics, Newport, Gwent. Available at: See also:

[3] L. A. King LA and A. C. Moffat (1983) A possible index of fatal drug toxicity inhumans. Med. Sci. Law 23(3): 193-197.

[4] L. A. King and J. M. Corkery (2010) An index of fatal toxicity for drugs of misuse. Hum. Psychopharmacol Clin Exp. 25: 162–166.

  • DrugScience
  • 2 Langley Lane
  • London
  • SW8 1GB
  • Registered Charity: 1150449