Saturday, June 20, 2009

Randomized controlled trial of an intervention to reduce gun-related violence: A Parody

I am incredibly disappointed that the journal that I consider to be the very pinnacle of medical evidence continues to print ideological propaganda without any regard whatever to evidence and logic when it suits the editorial agenda Unadulterated propaganda pieces related to capital punishment, abortion, and gun control are shamelessly and predictably aligned with a singular political stance, and evidence and logic are eschewed entirely in favor of dogmatic and sanctimonious deontology. Without slinging any more mud on my favorite journal, I will demonstrate this in the following parody:

Efficacy of a gun control policy in reducing gun-related violence: A multi-state, multi-center, randomized controlled trial.

Gun related violence results in tens of thousands of deaths (mostly suicides and homicides) each year. Interventions to reduce the toll of gun-related violence are desperately needed.

We used CDC data on gun-related deaths over the last decade to identify populations at risk for gun-related violence. However, our inclusion criteria did not comport with NIH-funding guidelines about inclusion of women and minorities and vulnerable populations such as former prisoners and felons and people with mental disabilities, some of which were over-represented and some of which were under-represented in the at-risk group we identified. Therefore, we dropped inclusion and exclusion criteria altogether, and randomized the entire populations of several states to the intervention (moratorium on firearms ownership defined as a complete ban imposed by state legislatures coupled with Directly Observed Confiscation) versus control (no moratorium or ban). Causes of deaths in each group were tracked and adjudicated by medical examiners in each state.

The two populations were well matched on baseline demographic characteristics. There was no difference in the gun-related fatality rate between the intervention and control groups (20.1 per 100,000 in the intervention group and 20.2 per 100,000 in the control group; P=0.98) based on an intention to treat analysis. There was considerable cross-over between groups and this potentially explains the failure of the intervention to produce the intended result. In subjects who crossed over from the intervention to the control group (hereafter called "criminals"), the odds of gun-related violence increased 1000.42 (p=0.00001). Many criminals were responsible for more than one gun-related death and crossed over multiple times from intervention to control. There was wide variability between the rates of gun related violence on the basis of geography and other factors, with fatality rates 10-100 times higher in Baltimore, MD than in Provo, UT.

An intervention to reduce gun-related violence failed to achieve this goal, largely as a result of cross-over from the intervention to the control group by "criminals". These criminals undermined the efficacy of the intervention. Moreover, the high geographic variability in gun related violence suggests that factors unrelated to the availability of firearms may drive gun-related violence rates. Future studies in limiting gun-related violence should focus on at-risk groups identified through crime statistics, and should not be NIH funded. Moreover, recrudescent crossover in future studies should be limited by incarceration of criminals for life without parole. Future studies might also focus on more traditional ways of preventing recrudescent cross-over (such as capital punishment). The Personalized Healthcare movement might also provide guidance on how to deal with this challenging problem.