Comparing the Harms of Drugs

Drugs (including legal drugs like alcohol and nicotine) are a widespread part of human life and death. In 2019, alcohol was used by an estimated 2.5 billion people over the age of 15, leading to 2.6 million deaths.¹ 1.2 billion people globally use tobacco, and over 7 million people die each year as a result (more than 1.6 million of those deaths are non-smokers exposed to second hand smoke).² In 2023, an estimated 316 million people used other psychoactive drugs, and drug use disorders resulted in close to half a million deaths.³ Mind altering chemicals have always been a part of the human experience, and will likely continue to be for as long as we are human. A clear understanding of the harms associated with drugs is necessary in choosing how we engage with them as a society and as individuals.

How can we compare drug harms?

A simple approach might be to count the deaths caused by the drug like above. There is a clear issue with this: we would ignore harms caused by a drug that aren’t related to mortality. For example, addiction, financial burden, and the development of chronic health issues. To complicate things further, there are harms associated with drug use that aren’t harms to the drug users themselves. Some examples of these harms include crime, damage to the environment, and injuries to others (such as drunk driving accidents). There is no single metric that can adequately capture how “harmful” a drug is, so comparing the harms of drugs requires a more sophisticated approach.

Let’s try something different. First, let’s make a list of all the categories of harm we can think of, whether they are harms to the user or to others. Then let’s give each drug a score from 0 to 100 for each harm category. The most harmful drug in a given category will get a score of 100. Now we can sum a drug’s scores across all the categories of harm to get an overall harm score. Here is an example for an imaginary drug using only three dimensions of harm to save space.

Harm Category	Score (0-100)
Physical harm to user	72
Addiction	46
Injury to others	85
Total	203

If we go through this process with a panel of experts then hopefully we can get reasonable results without it costing huge amounts of resources and time. Promising! But what if we care about some types of harm more than others? For example, maybe we think that injury to others caused by a drug should be weighted more heavily than addiction potential. We can incorporate these judgements by giving every category of harm a weight based on how much we care about it.

Harm Category	Weight
Physical harm to user	50
Addiction	30
Injury to others	100

Then, we can multiply a drug’s score for a harm by the weight we’ve given that harm.

Harm Category	Score	Weight	Weighted Score (Score × Weight)
Physical harm to user	72	50	3600
Addiction	46	30	1380
Injury to others	85	100	8500
Total	203	180	13,480

The total weighted score has ballooned quite a lot. As a final step, let’s constrain it to the range of 0-100 by dividing each weighted harm score by the sum of the weights (180 in this case).

Harm Category	Score	Weight	Weighted Score (Score × Weight) / 180
Physical harm to user	72	50	20.0
Addiction	46	30	7.7
Injury to others	85	100	47.2
Total	203	180	74.9

This is called a Multi-Criteria Decision Analysis (MCDA). In 2010, psychopharmacologist David Nutt and his colleagues on the Independent Scientific Committee on Drugs (ISCD) went through this process for 20 drugs that were common in the UK. They scored each drug on 16 criteria (nine harms to users and seven harms to others), and arrived at the results shown in the interactive chart above. Their paper and underlying data can be downloaded here.

An advantage of this approach is that the process used to arrive at a given ranking is structured and transparent. Thanks to this, the assumptions and methods can be examined, corrected, and refined. This also allows us to do sensitivity analysis. By tweaking harm weights we can see how much has to change for the end ranking to shift significantly, and quantify how robust our end ranking is. In reference to the data above:

Extensive sensitivity analyses on the weights showed that this model is very stable; large changes, or combinations of modest changes, are needed to drive substantial shifts in the overall rankings of the drugs. ⁴

Since 2010, similar analyses have been performed in the European Union, Australia, New Zealand, and Canada, and have arrived at similar results.

Limitations of this technique

The connection between culture and harm

To understand the data shown above it’s important to note that harms to the user are estimated on an individual scale, while harms to society are estimated population wide. Because of this, drugs that are more widespread (particularly legal drugs) will generally have inflated scores for harms to others compared to drugs with fewer users. In the chart above, see how alcohol in particular is lifted to the top by its large “Harms to Society” scores. This inflation of the scores of more widespread drugs should not be interpreted as something fundamental about the drug, but rather should be understood as a manifestation of the drug’s cultural context.

This is not the only place that cultural context affects the rankings. For example, prohibition and stigmatization of illegal drugs result in adulteration of the drug supply, direct legal risks for users (such as the risk of getting arrested), and behavioral risks such as using alone. Prohibition also forces the movement of illegal drugs into black markets, leading to growth in organized crime (cartels and mafia), and related violence. The reality is that many dimensions of a drug’s harm cannot be separated from its cultural context, and changing that context will change the harms of the drug. In their rather scathing response to the Nutt study, Caulkins, Reuter, and Coulson give some more examples:

Important harms include drug-related crime, environmental damage and the cost of police and prisons, yet none of those are characteristics of a chemical; they depend as well on legal status and programs implementing laws. Methamphetamines create environmental problems when produced in small, technologically primitive laboratories; this would disappear if they were legal. ⁵

Subjectivity

Another limitation of the MCDA is the subjectivity of the weighting process. At core, the choice of weights is a reflection of the MCDA panel’s values; it’s a philosophical exercise rather than a scientific one. Caulkins makes this explicit:

computing composite ranks is not a value-free exercise the public can delegate to scientific experts without concern about whether the scientists’ values are representative of the electorate’s. ⁵

For this reason, I have included functionality in the visualization allowing users to edit the weights. I encourage you to do so and see how it affects the ranking.

There is also the ever-present possibility of bias in the expert panel when creating the harm scores. Through a lack of data or personal experience they may simply underestimate or overestimate the harms of a drug.

Focusing on harms alone

Ranking drugs in this way gives the illusion that the drugs have been ranked from “worst” to “best”, but an examination of the harms of drugs is only one half of a cost-benefit analysis. Nutt calls attention to the incompleteness of focusing only on harms in the original study, writing:

Limitations of this approach include the fact that we scored only harms. All drugs have some benefits to the user, at least initially, otherwise they would not be used. ⁴

Jean-François Crépault and colleagues write about some of the benefits of drugs in a 2026 Canadian MCDA study:

People who use drugs report benefits including pleasure, relief of physical or psychological pain, enhanced cognitive functioning, social connection, and meaningful spiritual experiences ⁶

And Nutt mentions economic benefits:

Some drugs such as alcohol and tobacco have commercial benefits to society in terms of providing work and tax, which to some extent offset the harms and, although less easy to measure, is also true of production and dealing in illegal drugs. ⁴

An understanding of drugs that only includes their harms is an incomplete understanding of drugs.

Some other limitations

Nutt includes some more caveats himself:

[The harm rankings] do not relate to drugs when used for prescription purposes. Other issues to explore further include building into the model an assessment of polydrug use, and the effect of different routes of ingestion, patterns of use, and context. Finally, we should note that a low score in our assessment does not mean the drug is not harmful, since all drugs can be harmful under specific circumstances. ⁴

Conclusion

If we want our relationship with drugs to be aligned with their realities, then we need an evidence-based understanding of the harms of drugs. Despite their limitations, MCDAs represent a step forward in this effort. We could wait to reach conclusions about the relative harmfulness of drugs until we have ironed out every issue in our models, but this may never happen, and in the meantime poorly informed decisions by both individuals and governments will cost lives, money, and freedoms. Nutt’s 2011 response to Caulkins’ criticism was titled “let not the best be the enemy of the good”. Nutt concludes this response with:

Anyone interested in alcohol and other drugs, from law enforcement to education and from health improvement to international policy, needs some measure that allows them to understand and communicate relative harms and risks. I believe we have provided the best currently available analysis of an extremely complex multifaceted data set. It isn’t perfect but is nevertheless good enough to be useful. ⁷

Global status report on alcohol and health and treatment of substance use disorders ↩
WHO: Tobacco and nicotine ↩
UNODC World Drug Report 2025: Global instability compounding social, economic and security costs of the world drug problem ↩
Nutt DJ, King LA, Phillips LD; Independent Scientific Committee on Drugs. Drug harms in the UK: a multicriteria decision analysis. Lancet. 2010 Nov 6;376(9752):1558-65. doi: 10.1016/S0140-6736(10)61462-6. Epub 2010 Oct 29. PMID: 21036393. Read Here ↩ ↩² ↩³ ↩⁴
Caulkins J. P., Reuter P., Coulson C. (2011). Basing drug scheduling decisions on scientific ranking of harmfulness: false promise from false premises. Addiction 106 1886–1890. 10.1111/j.1360-0443.2011.03461.x Read Here ↩ ↩²
Crépault J-F, Russell C, Asbridge M, et al. Drug harms in Canada: A multi-criteria decision analysis. Journal of Psychopharmacology. 2026;40(2):286-295. doi:10.1177/02698811251409147 Read Here ↩
NUTT, D. (2011), LET NOT THE BEST BE THE ENEMY OF THE GOOD. Addiction, 106: 1892-1893. https://doi.org/10.1111/j.1360-0443.2011.03527.x ↩