There is, in fact, no way for such a study to rule out other factors for an attempted suicide.
So what does the study really show? It is based on 15,645 people born in the 1990s in the Avon Longitudinal Study of Parents and Children. "Tracked from birth" is an exaggeration, but it does include survey evidence about whether they had ever attempted suicide at the ages of 16, 24 and 25, and it has survey evidence showing their
PGSI (problem gambling) scores at the age of 20 and 24. A PGSI score of 8 or more out of 27 is the conventional threshold for problem gambling.
The idea behind the study is to see whether suicide attempts are preceded by problem gambling, thereby strengthening the correlation and reducing the risk of erroneous reverse causation (i.e. that suicidal feelings lead to problem gambling). The researchers also control for other factors that could lead to suicide attempts, such as alcoholism.
To cut a long story short, they find associations. They find that people who are problem gamblers aged 24 are 15% more likely to have made a suicide attempt in the last twelve months at the age of 25. They also find that people who are problem gamblers aged 20 are 20% more likely to have made
a suicide attempt in the last twelve months at the age of 24. The latter finding is a relative risk of 1.20 and is after controlling for other factors.
Although it is not made clear in the study, the press release says that...
The future suicidality link was most stark among 20-year-olds, where
there was a 20% increase in suicide attempts for every increment on the
PGSI meaning that those scoring eight and above experienced quadruple
the rate of suicide attempts four years later.
This seems to be saying that someone who scores 1 out of 27 in
the PGSI test is 20% more likely to kill themselves. That sounds unlikely. To get 1 out of 27, all you have to have done is "bet more than you can really afford to lose" or "gone back to try to win back money you had lost" in the last twelve months. It is hard to see how this could drive anyone to suicide. The researchers have then multiplied 20% by something to get a figure of 400% for people who are actually problem gamblers (i.e. score 8 or more). That suggests 20% x 20, but it is unclear where the 20 comes from. 20% x 8 seems more logical, but that only produces 160%. I confess that I am rather baffled by how they came up with the claim that "problem gambling quadruples the risk of suicide".
None of this stuff is in the study, so we can only guess where it came from. Did the authors find an association between attempted suicide and low PGSI scores and draw a linear line upwards or did they find an association with high PGSI scores and draw a linear line backwards? One thing is for sure: there were not enough attempted suicides for them to confidently draw conclusions about people with low, moderate and high PGSI scores.
The study seems to be 95% maths and 5% data, and the most important data is not presented. A lot of data simply doesn't exist and had to be "imputed" (i.e. guesstimated). Of the 15,645 people born into the study, 10,528 did not answer any questions about suicide attempts. That is 67%. Of the 4,331 people who answered the suicide questions, 2,616 (60%) did not answer the gambling questions. There were only 729 people who answered all the relevant questions and only 11 of them had attempted suicide in the past year. These are very small numbers of self-selecting people to be drawing bald conclusions about.
Moreover, the authors chose not to control for several co-morbidities that seem relevant. They controlled for alcohol use disorders, socio-economic status and hyperactivity (all of which are strongly correlated with suicide attempts), but not illicit drug use or - surprisingly - mental health disorders. Although they give reasons for this, they acknowledge that it may be a weakness:
Our selection of covariates was parsimonious. Other studies with greater
power should include additional covariates such as depression, anxiety
and illicit drug use to investigate whether these influence the
relationship between PGSI and suicidality.
The study itself is quite cautiously written and heavily caveated. At the end, the authors say:
In conclusion, the present study found evidence of positive associations
between harmful gambling and current and future suicide attempts among
young UK adults, which was unlikely to be confounded by pre-existing
suicidality.
That is a reasonable summary, as is the statement in the abstract that "this association may be more complex long-term, and increases in harmful
gambling during adulthood may not be an important factor", but it is in contrast to the press release which claims the study was able to "rule
out alternative explanations" (implying that all alternative explanations were ruled out, not just pre-existing
suicidality). No such claim can be reasonably made. It is perfectly possible for someone who was neither a problem gambler nor suicidal at the age of 20 to have gone down a dark path by the age of 24 in which symptoms of problem gambling feature but are not terribly important.
It is fairly obvious that suicidal people typically have a number of different problems in their lives.
A study published a few weeks ago found that problem gamblers in a Swedish hospital had a wide range of co-morbidities and were more likely than the average person to kill themselves, but that problem gambling
per se was not a risk factor for suicide. Despite the grandiose claim that the people in the new study have been tracked since birth, the authors do not have enough knowledge about their circumstances to be able to claim a causal link between "harmful gambling" and attempted suicide and, to be fair, they do not make such a claim explicitly in the study.
The press release is a different beast, however. The press release seems designed to get the factoid that problem gamblers are four times more likely to attempt suicide into circulation. It also includes the red flag of a call to political action....
Dr Newall said: “Gambling is a part of the government’s suicide
prevention strategy, and these results help underscore the need for
additional population-wide measures to prevent gambling harms, such as
meaningful restrictions on gambling advertising.”
The study has nothing to do with advertising or any other "population-wide measure". This is pure editorialising.
Incidentally, if we want to know what effect an advertising ban would have, we can look at what the problem gambling prevalence was in the first major gambling survey in 1999 - 8 years before gambling ads were legalised. The figure was 0.6%, statistically indistinguishable from the 0.4% reported in
the most recent survey of this kind.
The claim that problem gambling is a major cause of suicide is central to the argument that gambling is a "public health" issue. The aforementioned Swedish study should put an end to the zombie claim that "up to" 496 suicides are linked to gambling every year in Britain (since it was an earlier version of the Swedish study that Public Health England used to make its
foolish extrapolation), but there is a palpable urge on the part of anti-gambling campaigners and 'public health' academics to quantify something that probably cannot be quantified. Now that the gambling research field is
awash with cash, expect to see much more of this.
