In the introduction of his letter, Buchanan refers to psychopathy as a ‘pejorative term’ but later categorises it as a risk assessment instrument. It is neither. Psychopathy as a psychiatric syndrome was first described by a general psychiatrist ^{Reference Cleckley1} and further developed into a diagnostic construct operationalised with the PCL-R. ^{Reference Hare2} It is retained within dissocial personality disorder in ICD-10 and as an alternative model of antisocial personality disorder in DSM-5. The PCL-R is recognised internationally as the gold standard for assessment of psychopathy. Proficiency in its use should be a core competency for clinicians who work with offenders. Sadly, many are not adequately trained and struggle to comprehend why their treatments usually fail with these individuals and sometimes make their behaviour worse.

Buchanan may have misunderstood Seto’s ^{Reference Seto3} method. The instruments were applied simultaneously, not sequentially. However, he is right that sequential screening does not improve accuracy. We would suggest a better reference for an explanation. ^{Reference Leon4} We would also emphasise that risk assessment instruments are no more than screening instruments. Most importantly, there is currently no evidence base to demonstrate that routine clinical use of these screens can prevent violence, despite mandatory use in some UK services.

With regard to the ‘glass ceiling’ effect that we have previously investigated, ^{Reference Coid, Yang, Ullrich, Zhang, Sizmur and Farrington5,Reference Liu, Yang, Ramsay, Li and Coid6} reducing missing data will achieve little to break through this. Trigger factors precede many violent events. They may occur in the context of static and dynamic risk factors which have predictive efficacy. But trigger factors are causal, can occur within seconds to trigger violence and, most importantly, are not predictable.

Finally: the wager. There is no purpose in doing this if psychopathy is a personality construct. Furthermore, we have previously shown that few PCL-R items are predictive. ^{Reference Coid, Yang, Ullrich, Zhang, Sizmur and Farrington5} But we did rise to the challenge of Buchanan and tested the predictive accuracy of the VRAG, OGRS and HCR-20 in high-risk groups defined by these instruments. Using 32 as the HCR-20 cut-off and 27 for VRAG to be as close as possible to Buchanan’s 5.7%, we estimated AUCs for VRAG and OGRS in the same HCR-20 high-risk group, and AUCs for HCR-20 and OGRS in the corresponding VRAG high-risk group. In the VRAG high-risk group, the HCR-20 showed a low AUC of 0.44 (95% CI 0.28-0.61). The OGRS was a more respectable 0.69 (95% CI 0.53-0.85). In the HCR-20 high-risk group, AUC for VRAG was 0.67 (95% CI 0.54-0.81) and OGRS 0.68 (95% CI 0.64-0.81).

Perhaps there would be mileage in squeezing the fruit again in Buchanan’s next study?