Kellermann et al. analyze gun ownership as a risk factor for homicide in the home (Oct. 7 issue).* A matter related to the methods used in their study has bearing on its validity and, consequently, on the far-reaching conclusions made by the authors. Although the use of case proxies is accepted, a comparable data source must be used for controls.
*Kellermann AL, Rivara FP, Rushforth NB, et al. Gun ownership as a risk factor for homicide in the home. N Engl J Med 1993;329:1084-91.
After the identification of the homicide victims, proxies were identified and asked to provide information on various characteristics of the deceased persons. Controls and control proxies were also identified, but the proxies in this group participated in fewer than 50 percent of the interviews. Surely it is apparent that the data provided by the controls about themselves are scarcely equivalent to the data obtained from case proxies about homicide victims. At the very least, one would expect less misrepresentation by the controls than by the case proxies. To eliminate this threat to the validity of the data, it might have been useful to have interviewed the controls and their proxies separately in order to establish a concordance rate for the information provided. Would the strength of the reported conclusions have been the same if the analysis had been restricted to data from case-proxy and control-proxy pairs?
In addition, a significantly higher number of case subjects than matched controls lived alone. Such a finding compounds the problem noted above. Is the validity of data from an acquaintance or relative living apart from the case subject equivalent to the validity of data from a member of the same household?
Although efforts were made to interview proxies for controls, it is clearly difficult to explain this approach to a person who has just consented to participate in the study. Although practical problems in the implementation of a study design call for practical solutions, I fear that the solution in this case has damaged an important study.
David Litaker, M.D.
Cleveland Clinic Foundation
Cleveland, OH 44195
The fatal flaw in the effort by Kellermann et al. to evaluate the protective value of firearms is that they used only data on criminal homicide. As Kellermann and Reay have previously noted, (1) no study of homicide can evaluate the protective value of firearms, since only 0.1 percent of the over 2 million protective uses of firearms involve mortality, (2,3) and those uses were noncriminal. Kellermann and his colleagues should have realized they were using the wrong methods and data when they found that "controlled security access" to the home was riskier than gun ownership and that the risk of homicide was four times higher in rented homes than in public housing.
Even if we look only at homicide, the case-control approach used by Kellermann et al. tells us nothing about whether the use of alcohol -- found to be riskier than gun possession, although the authors offer no concomitant recommendation of teetotalism -- or possession of firearms is risky for ordinary citizens, since the study looked only at a high-risk population with dramatically different demographic characteristics from those of the general population. A substantial minority of the high-risk population studied may already be prevented from possessing firearms, since 53 percent of the case subjects reported a record of arrest of a household member, and 31 percent reported illicit drug use.
Kellermann et al. reveal little about homicide, since they limited their analysis to homicides that occurred in the home in three metropolitan counties and excluded homicides involving people under the age of 13 years (about 25 percent of which involve firearms (4)). With a few other exclusions, the crude odds ratios were based on 21 percent of the areas' homicides, and the adjusted odds ratios on just 17 percent. The result was an unrepresentative sample of homicides -- over 40 percent involved family members killing family members, although nationally such killings account for just over 10 percent of homicides (4). Their finding that most killings in and around homes involve people who know each other is as newsworthy as the finding that alcohol is involved in barroom slayings.
Paul H. Blackman, Ph.D.
National Rifle Association
Washington, DC 20036
As we discussed the article by Kellermann et al. during our graduate statistics class, several questions arose, which we hope you will address. First, the authors state, "In the light of [other] observations and our present findings, people should be strongly discouraged from keeping guns in their homes," implying that a gun in the home is a causal factor in homicide, suicide, and unintentional death. The study sample comprised people who had been killed in their homes and controls matched according to neighborhood, sex, race, and age. On what basis can one generalize from a sample of people who have been murdered to a population of people who keep guns in their homes?
Second, the authors list alcohol use, domestic violence, and illicit drug use as contributing factors to homicide. Although they state that the use of alcohol was "strongly associated with homicide," did they take into account the possible multicollinearity among the independent variables of living in a rented home, living alone, having a history of domestic violence, having an arrest record, using illicit drugs, and keeping guns in the house?
Finally, the authors also state, "One or more guns were reportedly kept in 45.4 percent of the homes of the case subjects." This implies that no guns were kept in 54.6 percent of the homes of the case subjects. In how many of the homicides was the victim killed with a gun that was kept in the house rather than a gun that was brought to the house by the perpetrator?
The students of Dr. Mark Ferris's Mathematical Statistics 460
Saint Louis University
St. Louis, MO 63108
Kellermann et al. tar gun ownership with an unnecessarily wide brush. According to the univariate analysis of the risk of homicide, the odds ratio was less than 1 for ownership of a shotgun or rifle, 1.9 for ownership of a handgun, 1.6 for any gun kept in the home, 2.1 for any gun kept unlocked, and 2.7 for any gun kept loaded. However, the one number that will be widely quoted from this article is the odds ratio for "gun or guns kept in the home," in the logistic-regression model (Table 4). If the raw data permitted it, an analysis of handguns, unlocked handguns, or best of all, loaded and unlocked handguns in the logistic-regression model might be much more pertinent than an analysis of any "gun or guns kept in the home." On the basis of the univariate analysis, the odds ratio for loaded, unlocked handguns would probably be much more specific and meaningful than the odds ratio of 2.7 for all guns.
The subjects with an unlocked, loaded handgun in the home might well represent a group with tendencies toward violence or feelings of victimization, as well as a group that blatantly ignores the clear rules of gun safety.
Jerry E. LeClaire, M.D.
Spokane Eye Clinic
Spokane, WA 99204
As so often happens in the health care arena, the article by Kellermann et al. raises some new questions. The term "risk factor" means different things to different people, (1,2) and the authors' use of the term is open to various interpretations. Their data indicate that "keeping a gun in the home was strongly and independently associated with an increased risk of homicide" but do not tell us more than that. "Associated with" and "causally related to" represent two different concepts. Does the presence of firearms in the home predispose the residents to their use, which suggests that if guns were not available, the risk of domestic homicide would be greatly reduced? Is gun ownership a marker for having one's thoughts focused on violence and thus being more likely than others to act violently toward people in their environment, in which case getting rid of one type of weapon might simply lead to the substitution of another? Or is a combination of cause-and-effect relations at work here?
Robert D. Gillette, M.D.
St. Elizabeth Family Health Center
Youngstown, OH 44501-1790
The study by Kellermann et al. suffers from many flaws. Who or what were the controls? Nonvictimized households? We are told that 35.8 percent of the households of the controls had a gun. We are also told that the households of the case subjects were more likely to have a loaded or unlocked gun and were more likely to have a member with a record of a criminal arrest. Meaningful controls? I doubt it.
The authors' interpretation of their results is an example of "data torturing" (1). Specifically, Kellermann and his colleagues are guilty of Procrustean data torturing, which is defined as "deciding on the hypothesis to be proved [in this case, owning a gun increases the risk of homicide] and making the data fit the hypothesis." Never mind that there were more users of illicit drugs, alcoholics, and persons with a history of violence in the households of the case subjects than in the households of the controls or that, by the authors' own admission, 11 of the case subjects were killed by private citizens acting legally in self-defense. In other words, some instances of gun ownership prevented the owner or family members from becoming victims -- indeed, may have even saved their lives.
What the article did show is that illicit drug use, alcoholism, and a pattern of violent behavior are risk factors for homicide involving firearms. What the article failed to address is that gun ownership by responsible people is not a risk factor. In other words, it is not the gun (an inanimate object) that is the problem but its inappropriate use.
What is equally disturbing is your editorial in the same issue, (2) in which you criticize the National Rifle Association (NRA). Clearly, you have focused on the NRA's lobbying and not on the fact that the NRA promotes responsible, safe handling of firearms for appropriate activities, such as hunting, collecting, competitive shooting, practicing shooting to maintain proficiency, and using a gun as a last-resort means of personal defense. (It is interesting that the study by Kellermann et al. showed that home-security measures, such as deadbolts, window bars, security doors, and alarms, had little effect on the risk of homicide in the home.) Instead of criticizing the NRA, why not use that organization and its millions of members, many of whom are physicians and other health professionals, as a resource in our schools to teach our children how to use firearms in a safe and responsible manner?
Pat Baranello, M.D.
200 S. Ostrander Rd.
East Aurora, NY 14052
I was as disturbed by the rhetoric and misunderstanding of the problem in your editorial on household guns as I have been by writings on the subject by the NRA.
Those who advocate sweeping gun restrictions and those who advocate no restrictions are both missing the point and wasting energy that could be used to deal realistically and appropriately with the problem. I think everyone would agree with some basic ideas. Accidental or deliberate deaths or injuries involving firearms are outcomes that we should all actively work to avoid. Those who wish to hunt and shoot for sport should be allowed to do so in a free society, so long as in doing so they do no harm to others.
Automobiles are tools that become deadly when used in a dangerous manner. Therefore, we have driver's licenses, which are obtained by demonstrating knowledge about the safe and expert use of an automobile. Similarly, firearms are tools. Society should require that those who wish to own and use firearms demonstrate their knowledge of how to use them safely and skillfully. States should issue firearm user's licenses with stringent requirements. This would not infringe on our rights, since the right to own and use firearms includes the responsibility to use them safely and appropriately.
Hunting and sport shooting are not the same as military shooting. The former involves the use of rifles, pistols, or shotguns firing projectiles that differ considerably from those shot from military firearms designed for rapid fire against a protected human enemy. Military weapons and ammunition should be strictly reserved for use by the military and the police. Honestly, shooting a duck with an antitank missile will not do much for the dinner table. . . .
Charles R.B. Beckmann, M.D.
University of Wisconsin-Madison Medical School
Milwaukee, WI 53201
Once again, you use your position as editor to expound on your desire for government confiscation of firearms. At the same time, you proclaim the imminent doom of the NRA, which you accuse of wasting huge amounts of money to oppose gun legislation that you claim is actually supported by the NRA's rank-and-file members. You fail to mention that the NRA, which you portray as fragmented and failing, has grown from 2.4 million to 3.2 million members in the past 18 months, making it one of the largest, fastest-growing organizations in the United States.
You also fail to mention that there are already more than 20,000 gun-control laws in the United States at the local, state, and national levels, and that more than two thirds of Americans are already subject to a waiting period similar to the one in the Brady bill. Furthermore, it has yet to be shown that a waiting period for the purchase of handguns such as that mandated by the Brady bill reduces crime. Semiautomatic "assault" weapons (a misnomer, since assault weapons are, by definition, fully automatic) have been used widely for sport for more than 100 years, yet according to the Federal Bureau of Investigation's Uniform Crime Report of 1992, such weapons account for less than 1 percent of guns used in crimes.
The NRA has called for a response to violent crime in our country, including mandatory prison sentences, truth in sentencing, a comprehensive reform of juvenile justice, and protection against liability for people who exercise reasonable defensive force.* Yet you mention none of this in your rush to blame private ownership of guns, in general, and the men and women of the NRA, in particular, for crime in America. The present fad of gun-control rhetoric is a vain attempt to find an easy solution to a complex problem rather than address the real causes of crime in our society: the glorification of violence by the media, the failure of the criminal-justice system, and the disintegration of the family unit in inner-city America.
*Fighting crime the NRA way. American Hunter 1993;21(10):44-5.
J. Marc Pipas, M.D.
Dartmouth-Hitchcock Medical Center
Lebanon, NH 03756
To the Editor:
Large-scale cohort studies are usually the best way to explore the relation between a potential risk factor and an outcome of interest. Since this approach was not feasible, we conducted a case-control study. By ascertaining the rate of gun ownership in households where a homicide had occurred and comparing this rate with that noted in a random sample of neighboring households that contained a person of the same age group, sex, and race as the victim, we obtained a good approximation of relative risk (1). This is the same research technique that was used to identify the link between cigarette smoking and lung cancer (2).
The exaggerated claim that guns are used in self-defense more than a million times a year has not withstood scientific scrutiny (3,4). If a gun in the home affords substantial protection from homicide (whether it is used to injure, kill, or frighten intruders or simply discourage them from entering), we should have found that homes in which a homicide occurred were less likely to contain a gun than similar households in which a homicide did not occur. The opposite was true.
We restricted our study to homicides in the home because the risk or protective benefit of a readily available firearm should be most plausibly demonstrated where it is kept. All such homicides were included, whether or not they involved a person at high risk for violence because of various behavioral factors. Although we noted a degree of association among several behavioral risk factors, each contributed independently to the risk of homicide. Therefore, we took these effects into consideration in our final model.
A comparable ascertainment of exposure is crucial in any case-control study, which is why we based our analysis on interviews rather than on-the-scene reports. Ninety-three percent of the homicides involving firearms occurred in homes where a gun was kept, according to the proxy respondents. In 8 of the other 14 homicides, the investigating officer specifically noted that the gun involved had been kept in the home.
Although we tried to interview a proxy respondent for each control, this was often impossible. However, the rate of gun ownership reported by the proxy respondents was actually higher than the rate reported by the controls themselves.
We are confident that our findings will be corroborated. The early studies of smoking and lung cancer were confirmed by subsequent studies. Nonetheless, some doctors still smoke. Old habits -- and deeply held beliefs -- die hard.
Arthur L. Kellermann, M.D., M.P.H.
Atlanta, GA 30329
Grant Somes, Ph.D.
University of Tennessee, Memphis
Memphis, TN 38103
Frederick P. Rivara, M.D., M.P.H.
University of Washington
Seattle, WA 98195
1. Schlesselman JJ. Case control studies: design, conduct, analysis. New York: Oxford University Press, 1982.
Return to Text
2. Wright JD, Rossi PH, Daly K, Weber-Burdin E. Weapons, crime, and violence in America: a literature review and research agenda. Washington, D.C.: Government Printing Office, 1983:257.
Return to Text
I have never proposed that all guns be confiscated. In my editorial I supported the passage of the Brady bill; recommended new design standards for firearms, registration of all firearms, and licensing of all gun owners; and proposed a ban on bringing guns into schools. I stressed the need for nationwide laws because of the variability of the thousands of local and state ordinances. It is pointless to restrict the sale and possession of handguns in one locale, for example, when they can be purchased across a nearby state line. I also supported a federal ban on the sale of automatic and semiautomatic weapons. I am unmoved by the argument that these weapons account for only a small fraction of deaths. What is their purpose other than to kill people? Why should any civilian own one?
Poverty, frustration, humiliation, hopelessness, drugs, broken families, estrangement from society, and an obsession with violence as entertainment are some of the causes of our plague of gun-related violence, not the guns themselves. Yet guns, especially handguns and automatic weapons, are instruments of death, often of innocent bystanders, as the study by Hutson and his colleagues in this issue of the Journal illustrates (see page 324).
The NRA supports all kinds of penalties for criminals but refuses to give an inch on reasonable gun-control measures. Yes, the NRA has achieved an impressive growth in membership recently, in part through scare tactics aimed at women in its "Refuse to Be a Victim" campaign (1). More and more women have purchased handguns, adding to the dangerousness of society.
I am convinced that the overwhelming majority of NRA members are responsible citizens who use their guns for sport and hunting and who support educational programs on the use of firearms. But I fault these rank-and-file members, 3 million strong, for supporting an organization whose leaders are so recalcitrant (see the Book Reviews, page 373). I fault them for preserving leaders whose solution to the Somalia crisis is to arm all mothers with AK-47s (2).
Fortunately, Congress has finally heard the wail of the people, and the NRA has lost some of its hold on lawmakers. The more the gun lobby spends trying to defeat courageous politicians, the less power they will have to oppose reasonable and effective gun-control legislation.
Jerome P. Kassirer