Low cholesterol and violent crime

Low cholesterol and violent crime$

Beatrice A. Golomb a,b,c,*, HaÊ kan Stattin d, Sarno� Mednick a

aDepartment of Medicine, University of California, Los Angeles, CA, USAbDepartment of Medicine, University of California, San Diego, CA, USA

cDepartment of Psychology, University of Southern California, Los Angeles, CA, USAdDepartment of Psychology, University of Stockholm, Stockholm, Sweden

Received 30 August 1999; received in revised form 21 February 2000; accepted 13 June 2000

Abstract

Background: Community cohort studies and meta-analyses of randomized trials have shown a relation between low or loweredcholesterol and death by violence (homicide, suicide, accident); in primates, cholesterol reduction has been linked to increasedbehavioral acts of aggression (Kaplan J, Manuck S. The e�ects of fat and cholesterol on aggressive behaviour in monkeys. Psy-chosom. Med 1990;52:226±7; Kaplan J, Shively C, Fontenot D, Morgan T, Howell S, Manuck S et al. Demonstration of an asso-

ciation among dietary cholesterol, central serotonergic activity, and social behaviour in monkeys. Psychosom. Med 1994;56:479±84.). In this study we test for the ®rst time whether cholesterol level is related to commission of violent crimes against others in alarge community cohort.

Methods: We merged one-time cholesterol measurements on 79,777 subjects enrolled in a health screening project in Varmland,Sweden with subsequent police records for arrests for violent crimes in men and women aged 24±70 at enrollment; and with infor-mation on covariates. We performed a nested case control comparison of cholesterol in violent criminals Ð de®ned as those with

two or more crimes of violence against others Ð to cholesterol in nono�enders matched on age, enrollment year, sex, education andalcohol, using variable-ratio matching, with a nonparametric sign test.Results: One hundred individuals met criteria for criminal violence. Low cholesterol (below the median) was strongly associated

with criminal violence in unadjusted analysis (Men: risk ratio 1.94, P=0.002; all subjects risk ratio 2.32, P<0.001). Age emerged as

a strong confounder. Adjusting for covariates using a matching procedure, violent criminals had signi®cantly lower cholesterol thanothers identical in age, sex, alcohol indices and education, using a nonparametric sign test (P=0.012 all subjects; P=0.035 men).Conclusions: Adjusting for other factors, low cholesterol is associated with increased subsequent criminal violence. # 2000

Elsevier Science Ltd. All rights reserved.

Keywords: Cholesterol; Violence; Serotonin; Crime

1. Introduction

This investigation examines the relation between thecommission of violent o�ences and serum cholesterol ina cohort from Varmland, Sweden (Lindberg et al.,1992). Excess risk of violent death and suicide was pre-viously found for men with low cholesterol in a subset

of this population, in the ®rst 7 years of follow-up: a 2.8fold increased risk of violent death (death by suicide,homicide, and accident) (P<.001) and a 4.2 foldincreased risk of death by suicide (P=0.001) were noted(Lindberg et al., 1992). In this paper we extend ®ndingson violence, by examining the in¯uence of low choles-terol on violent crimes against others.Violence is the leading cause of death among those

under age 44 (Holinger and Klemen, 1982; Martinez,1994) and leads to more years of life lost than any medicalcondition, including heart disease (Holinger and Klemer,1982). Surviving victims of violence su�er persistentmedical problems, and are heavy utilizers of medicalresources (Drossman et al., 1995). Enhanced under-standing of biological (as well as societal) risk markersfor violence may facilitate development of strategies to

0022-3956/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.

PI I : S0022-3956(00 )00024-8

Journal of Psychiatric Research 34 (2000) 301±309

www.elsevier.com/locate/jpsychires

$ This work was presented at the Robert Wood Johnson Clinical

Scholars meeting in Ft. Lauderdale, FL, Nov. 1996. It was supported

by the Robert Wood Johnson Foundation and the Harry Frank

Guggenheim Foundation

* Corresponding author at Department of Medicine, University of

California, San Diego School of Medicine, 9500 Gilman Drive, La

Jolla, CA 92093-0995, USA. Tel.: +1-858-558-4950; fax: +1-858-558-

4960.

E-mail address: [email protected] (B.A. Golomb).

reduce the expression of and morbidity from violence;this study extends ®ndings related to one risk marker inparticular, serum cholesterol.A systematic review of the literature has shown that

the relation between cholesterol and violence potentiallymeets criteria for a causal association (Golomb, 1998).The largest cohort studies, with 350,000 (Neaton et al.,1992) and 55,000 subjects (Lindberg et al., 1992), and apooled analysis of 18 smaller studies excluding thosetwo (combined sample size 300,000) (Jacobs et al.,1992) Ð have found signi®cantly more violent deaths inmen with low cholesterol, a ®nding supported by threesmaller studies (Chen et al., 1991; Schuit et al., 1993;Zureik et al., 1996). Analyses of suicide and violence inpsychiatric and criminal populations (Gray et al., 1993;Hillbrand & Foster, 1993; Modai et al., 1994; Spitz etal., 1994; Sullivan et al., 1994; Takei et al., 1994; Gal-lerani et al., 1995; Golier et al., 1995; Hillbrand et al.,1995; Girgin et al., 1996; Mufti et al., 1998) primatestudies (Virkkunen, 1979, 1983; Bramblett et al., 1981;Virkkunen and Penttinen, 1984; Kaplan and Manuck,1990; Kaplan et al., 1991, 1994), and some meta-analy-sis of clinical trials (Muldoon et al., 1990, 1993; DaveySmith and Pekkanen, 1992; Ravnskov, 1992), add sup-port to a relation between lowered cholesterol and vio-lence, though results of meta-analyses vary by studyinclusion criteria and subgroup analyzed (Cummingsand Psaty, 1994; Law et al., 1994). The outcome mea-sure in cohort studies and meta-analyses of RCTs hasbeen death due to violence (accident, homicide, or sui-cide) in the person with low or lowered cholesterol. Yetdeath from violence is correlated with violence duringlife, so if low cholesterol subjects die more from vio-lence, they may, by hypothesis, exhibit more violenceagainst others. Thus, we examined the relation betweenlow cholesterol and violent acts against others, assessedby police-registered repeat violent o�ences. By con®ningthe analysis to those crimes against a person which cul-minated in police involvement, only more serious eventsare counted, reducing sensitivity for adverse behaviors,but increasing speci®city compared to measures of``hostility'' or aggressiveness, which correlate poorlywith violent behaviors (Smith and Frohm, 1985; Leikerand Hailey, 1988). This analysis focuses on subjectswith two or more arrests for violence against others,following established strategies in the violence litera-ture (Brennan et al., 1989), enhancing speci®city byeliminating one-time o�enders. (Such individuals mayin some instances be subjects of false arrest or justi®-able self-defence.) If low cholesterol is linked to vio-lence against others, then victims of such violence reapnegative health e�ects without the compensating car-diac bene®t that occurs in individuals who themselvesreceive treatment.We hypothesized that male and all subjects with low

cholesterol would exhibit increased criminal violence

against others. We chose to perform an analysis includ-ing women, just as women have been included in ana-lyses of cholesterol reduction and mortality (ScandinavianSimvastatin Survival StudyGroup, 1994), though no studyhas separately shown cardiac death or overall mortalitybene®t separately in this subgroup or has been poweredto do so. (Similarly our sample was expected to beinadequate to evaluate the relationship of cholesterol tocriminal violence independently in women, who exhibitfar lower rates of violence.)

2. Method

2.1. Subjects and measurements

Subjects were drawn from a general health surveycohort in Varmland county, Sweden conducted in 1962±1965, in which men and women over age 24 received, atenrollment, a one-time total cholesterol measurement bya modi®ed Liebermann±Burchard method (Zak et al.,1954). Of the eligible population, 76% participated(Lindberg et al., 1991); 99.5% of participants had cho-lesterol measured, leading ultimately to 79,777 subjectsunder age 70 at entry for whom cholesterol measure-ments were obtained.

2.2. Cholesterol

Cholesterol was retained as a continuous variable forthe main analysis. For purposes of dichotomization,``low'' cholesterol was de®ned to be less than thispopulation's median of 250 mg/dl, and ``high'' choles-terol as greater than this median. The terms ``low'' and``high'' are purely intended to re¯ect values relative tothe median and do not represent judgments aboutvalues with respect to desirable cholesterol level.

2.3. Covariates

Covariates obtained included age at cholesterolscreening, sex (from Varmland database), and educationlevel (from the Utbildningsregistret), which re¯ectssocioeconomic status. No alcohol information wasobtained at enrollment. Number of hospital admissionsin which an alcohol-related diagnosis was noted(obtained from the Patientsregistret) served as theavailable alcohol measure.

2.4. Violence

Arrests for violent crimes (number of arrests andpenal code violation) were obtained from computerizedSwedish police records for the period 1966±1994.Information on date of crime was not available. Violentcrimes were de®ned by Chapter 3 of the Swedish Penal

302 B.A. Golomb et al. / Journal of Psychiatric Research 34 (2000) 301±309

Code (National Council for Crime Prevention (Sweden),1990), ``Crimes Against Life and Health'', and includedmurder, attempted murder, assault, aggravated assault,manslaughter, infanticide, and causing (or creatingdanger of) bodily injury or illness. ``Violent criminals''as de®ned for this study were those committing two ormore such crimes during followup Ð here termed``criminal violence.'' ``Nono�enders'' were those withno registration for arrests during this period. Since theoutcome measure did not require subject participation,and few Varmland subjects emigrated from Sweden,subject loss following cholesterol measurement wasminimal.

2.5. Mortality

Mortality was ascertained from the Swedish mortalityregister (DoÂ dsorsaksregistret). Mortality data permittedassessment of years of follow-up.

2.6. Merging of data

Data related to covariates and outcomes from themultiple records were merged by subject identi®cation;identifying information was removed in Sweden, priorto analysis, to protect subject con®dentiality.

2.7. Analysis

Bivariate analyses of the relations of cholesterol toviolence; and of covariates to cholesterol and to vio-lence were performed. Mortality data were examined bycholesterol, age, and criminal violence status to assessthe hypothesis that low cholesterol subjects may commitmore violent crimes simply through enhanced longevity(with more time-opportunity for violence). An adapta-tion of a nested case control design comprised the mainanalysis, in which cholesterol from all identi®ed cases(n=100 violent criminals) was compared to cholesterolfrom nono�enders identically matched on sex, age atcholesterol measurement, alcohol measure, enrollmentyear (to match age-at-start-of-crime-follow-up), andeducation. Where more than one optimal-match wasavailable, the control cholesterol was determined as themedian cholesterol of all best-matches. A sign test (two-tailed) assessed whether cholesterol of violent criminalswas signi®cantly more often lower (or higher) than thatof matched nono�enders. This nonparametric ``match-ing'' analysis was chosen for the primary analysis,because of the highly skewed nature of the violent crimevariable. Logistic regression was performed using datafrom all 79,777 subjects, as a subsidiary analysis. Infor-mation on date of crimes was not available, precludinguse of survival analysis in this cohort.Although these data pertain to low and not lowered

cholesterol, questions may arise regarding implications

of this study to lipid-lowering treatment. Since choles-terol lowering treatment is typically directed at indivi-duals (and more commonly men) over age 40, whileviolent crime is typically perpetrated by individuals Ðusually men Ð under age 40, we separately examined(posthoc analysis) whether the relation between lowcholesterol and violence was upheld in men over age 40,the group of chief relevance to cholesterol-loweringtreatment.

3. Results

3.1. Characteristics of the population

Descriptive characteristics of the population are pre-sented in Table 1. Median cholesterol was 250 mg/dl.There were 2.8/1000 male violent criminals. Only 90men and 10 women in the cohort met the criterion forviolent o�ending.

3.2. Unadjusted analysis: violent o�ending as a functionof cholesterol level

Fig. 1 presents the unadjusted relationship betweencholesterol level (in quartiles) and rate of criminal vio-lence. As can be seen, as cholesterol level decreases, therate of violent o�ending increases (w2(3)=15.36,P=0.002); the odds ratio for being a violent o�enderfor those with ``low'' cholesterol (below this popula-tion's median of 250 mg/dl) was 2.26 (w2(1)=15.5,P<0.001) for all subjects and 2.03 (w2(1)=9.61,P=0.002) for men. However these values are unad-justed for potential confounders. Of note, cholesterolvalues for those with one violent o�ence fell betweenthose of nono�enders and violent criminals. (Indeed,

Table 1

Characteristics of study population (males and females): 79,777 obser-

vations

Variable Mean S.D.

Mean cholesterol (mg/dl) 253 41.5

Mean age at enrollment (years) 46.9 12.1

Higher educationa 0.362 0.481

Mean education levelb 1.92 1.41

51 Alcohol admissionc 0.016 0.127

Mean # of alcohol admissions 0.053 0.764

Mean years of follow-up 22.9 7.69

a Coded as 1 if at least nine years of education were completed,

otherwise 0.b Rated on a Swedish registry scale of 1±7 scale, where 1 corre-

sponds to fewer than 9 years of education, and 7 to research level

education.c Coded as 1 if subject had at least one hospital admission with an

alcohol diagnosis recorded, otherwise 0.

B.A. Golomb et al. / Journal of Psychiatric Research 34 (2000) 301±309 303

individuals who had committed a single violent crimewere intermediate between violent criminals and non-o�enders on all measured variables Ð data not shown.)

3.3. Bivariate relationships

Table 2 shows the results of bivariate relations ofvariables to cholesterol and to violence. Young age isassociated both with low cholesterol and with criminalviolence. Low education is associated with high choles-terol and with a (trend toward) increased violence.Alcohol admissions were strongly related to increasedviolence but were unrelated to cholesterol. The in¯uenceof cholesterol and of violence on survival was examined;low cholesterol was associated with increased survival,but criminal violence was strongly associated withdecreased survival; respective odds ratios for survival tostudy completion, adjusted for year of age, are: 1.06,

P<0.001 (for cholesterol <250 mg/dl); and 0.40,P<0.001 (criminal violence), precluding an explanationby which low cholesterol leads to increased criminalviolence through increased longevity, with more time tocommit violent crimes.

3.4. Main analysis

Table 3 shows the results of the primary analysis.There were 100 violent o�enders. Cholesterol of violento�enders was lower than cholesterol of nono�endercontrols 1.7 times more often than it was higher (signtest; two-tailed P=0.012). The result is independentlysigni®cant for males (two-tailed P=0.03). This resultsuggest that low cholesterol is prospectively related toincreased criminal violence independent of age, sex,alcohol, and education.Logistic regression, using criminal violence as the

outcome variable, qualitatively corroborated the ®nd-ings of this analysis, controlling for all factors exceptage at start of crime follow-up (which was excluded dueto collinearity with age at cholesterol measurement).Risk ratio for violence with low cholesterol, adjusted forall covariates, was 1.7, a result of borderline signi®cance(P=0.07).

3.5. Violent o�ending as a function of cholesterol level:men over age 40

For each year of age over 40, normative cholesterolfor that age exceeded the Varmland cohort's mediancholesterol of 250 mg/dl. However, most violent crim-inals over age 40 (2.15 times as many as not) had cho-lesterol under 250 mg/dl. The odds ratio for violentcrime with ``low'' cholesterol (dichotomized at thispopulation's median of 250 mg/dl) was 2.34 (P=0.033)for those over age 40 (de®ned by age at start of follow-up).

4. Discussion

The ®ndings of this study support the hypothesis thatlow cholesterol level is associated with increased risk ofpolice registration for violent o�enses against others inSweden.Use of data from Sweden has several advantages. The

Scandinavian police registers are viewed as the mostcomprehensive and accurate registers in the world(Wolfgang, 1977), and the Swedish death register isviewed as one of the most complete mortality registers(Bolander, 1981). The Varmland population is a ruralpopulation with low mobility; in addition the SwedishPolice Register is national; thus losses from movementoutside the geographical area of follow-up are minimal.Periods of incarceration might be considered to reducethe e�ective duration of follow-up for crimes. However,

Fig. 1. As cholesterol decreases (across descending cholesterol quar-

tiles), the rate of violent criminals per 1000 nono�enders increases.

The di�erences are statistically signi®cant (P<0.001); however, these

data are unadjusted for age and other covariates.

Table 2

Bivariate relations of variables to cholesterol and to violence

Variable E�ect on violence Odds ratioa

Low cholesterol (<250 mg/dl) Increase 2.32*

Low age ( # by 1 year) Increase 1.09*

Low education (49 years) Increase 1.31

Alcohol admissions (51) Increase 36.3*

Variable E�ect on cholesterol Odds ratioa

Low age ( # by 1 year) Decrease 0.96*

Low education (4 9 years) Increase 1.27*

Alcohol admissions (51) No change 1.06

a *P<0.001 (P>0.2 where no asterisk is present).


jail periods were short by American standards, withonly 6 of the 100 violent criminals jailed for more than 2years (incarceration periods, summed over all o�ences,were 2.2, 3.2, 4.1, 4,4, 8.9 and 10.25 years for these six).Violent o�enders might have had time to commit moreo�ences had incarceration not supervened, but thiswould not have in¯uenced their categorization. Themedian cholesterol for this population Ð 250 mg/dl Ðis high by American standards. This is an advantage forthe present analysis, since it allows evaluation of di�er-ences in risk of violence within the range in which cho-lesterol-lowering treatment might be undertaken. Again,the term ``low'' cholesterol refers exclusively to therelation to this population's median, and is not intendedto imply a judgment regarding cholesterol level in rela-tion to ``desirable'' cholesterol values.Young age, low socioeconomic status, and use of

alcohol have all been suggested as possible causal con-founders in the relation between low cholesterol andviolence; it has been speculated that each of these vari-ables may be related both to low cholesterol and toviolence, and thus may produce the appearance of arelationship between these variables. Indeed, all three(using low education as a proxy for low socioeconomicstatus) are related to increased violence in bivariateanalysis, a ®nding that is signi®cant for young age anduse of alcohol. However, only young age was related tolow cholesterol; our measure of alcohol was unrelatedto cholesterol (despite tapping more signi®cant alcoholuse); and low education was related to high rather thanlow cholesterol. Thus, only age emerged as a plausiblecausal confounder (a factor that would distort theapparent relation between cholesterol and violence inunadjusted analysis), suggesting age in particular mustbe carefully adjusted for in any analysis; thus, matchingby year of age (rather than accepting broader age bins)and regression adjusted by year of age were performedto mitigate confounding by age.Since most violent crime is perpetrated by men under

40, but most concern regarding cholesterol reduction isdirected at individuals Ð and predominantly males Ðover age 40, we undertook a subsidiary analysis toevaluate whether the relationship between low choles-terol and violence could be discounted in men over age

40, the population of greatest clinical concern withregard to cholesterol. The relation between low choles-terol and criminal violence was upheld in this popula-tion: the odds ratio for criminal violence, dichotomizingcholesterol at the Varmland population's median of 250mg/dl, exceeded two, and the result was signi®cant.Natively low cholesterol must be distinguished from

experimentally lowered (but perhaps still high) choles-terol. We studied the relation of natively ``low'' choles-terol (compared to the population median, not to idealcholesterol) and violent o�ending; it remains quite pos-sible that low cholesterol re¯ects baseline metaboliccharacteristics of violent o�enders, with no causal rela-tionship to violence. However, since evidence from bothobservational (low) and experimental (lowered) studiessuggest a relation between cholesterol and violence, thepossibility remains that a low cholesterol±violence rela-tionship could extend to lowered cholesterol, perhapsexclusively in a susceptible subset. The connectionbetween low and lowered cholesterol is particularlypertinent in the present study, since ``low'' cholesterol(below this population's median) was de®ned as under250 mg/dl, which is viewed as high by guidelines, andmany individuals are being treated to bring cholesterolbelow this value.A plausible mechanism for an association between

low or lowered cholesterol and violence has been pro-posed involving reduction of central serotonin activity(Kaplan et al., 1997; Golomb, 1998). Observationalstudies in humans have shown a trend (Ringo, 1994) ore�ect (Steegmans et al., 1996) of lower serotonin inthose with low cholesterol. Experimental assignment tocholesterol reduction produces low serotonin activity inanimal models (monkeys) (Muldoon et al., 1992;Kaplan et al., 1994). In animals (Olivier et al., 1984;Higley et al., 1992; Mehlman et al., 1994; Raleigh andMcGuire, 1994; Saudou et al., 1994; Brunner and Hen,1997; Higley and Linnoila, 1997; Sanchez, 1997) andhumans ( AÊ sberg et al., 1976; Traskman et al., 1981;Linnoila et al., 1983; Lidberg et al., 1985; van Praag etal., 1987; Coccaro, 1989; Kruesi, 1989; Virkkunen et al.,1989; Brown and Linnoila, 1990; Asberg, 1997; Cleareand Bond, 1997; Cologer-Cli�ord et al., 1997; Lucki,1998; Manuck et al., 1998; Moeller et al., 1998), nativelylow and experimentally lowered serotonin have beenlinked to increased aggression, violence and suicide.Restoring or augmenting serotonin has been found toattenuate violent behaviors (Coccaro, 1989) in humans(Morand et al., 1983; Gedye, 1991, Ratey et al., 1991,Coccaro and Kavoussi, 1997; Beshay and Pumariega,1998; Kant et al., 1998; McDougle et al., 1998) as wellas animals (Berzsenyi et al., 1983; Olivier et al., 1984;Molina et al., 1986; McMillen et al., 1988; Mitchell andRedfern, 1997; Abe et al., 1998; Moechars et al., 1998).Together these suggest that low or lowered cholesterolmay be linked to reduced central serotonin activity, and

Table 3

Adjusted relation of cholesterol to criminal violence: cholesterol in

violent criminal compared to matched nono�ender

No. of Cases

with cholesterol

lower in

criminal than in

matched control

No. of cases

with cholesterol

higher in criminal

than matched

control

Ratio Sign-test:

two-sided

P-value

All subjects 62 37 1.7 0.017

Males 55 34 1.6 0.0349


may thereby promote violence. Other suggestedmechanisms for a relation of cholesterol to violence(with or without a relation to serotonin) have describeda central role for fatty acids, particularly docosahex-aenoic acid (Hibbeln and Salem, 1995, 1996; Hibbeln etal., 1997, 1998a, b) (however a connection of this fattyacid to total serum cholesterol has not been demon-strated); or the metabolic response to carbohydrates(with hypoglycemic response to glucose challenge ser-ving as a marker for risk) (Virkkunen et al., 1994). Ineach case the proximal neurobiological mechanismremains to be identi®ed.This study is limited by use of a one-time cholesterol

measurement, potentially remote from the time of theviolent crime. This misrepresents subjects' ``true'' serumcholesterol due to cholesterol variability (Patterson etal., 1993), leading to ``measurement error'' with ensuingloss of precision. However, the expected consequence isbias toward the null leading to attenuation of a truerelationship rather than production of a spurious one.Additionally, this was a relatively nonviolent olderpopulation; use of a predominantly older populationhas the advantage of enhancing possible clinical rele-vance by more closely re¯ecting the population likely toreceive cholesterol-lowering intervention, although theabsolute excess of violent crimes seen may understatethat of a younger or more violent population. Thoughyounger individuals are less commonly placed on cho-lesterol lowering drugs, many may be in¯uenced bypopulation-based e�orts to lower cholesterol; moreover,recent studies showing reduction in cardiovascularevents with statin (lipid-lowering) treatment in healthypersons with ``normal'' cholesterol and LDL have beentouted as favoring revision (liberalization) of cholesteroltreatment guidelines (the slight, nonsigni®cant excess ofoverall mortality in the treated group was not addres-sed) (Downs et al., 1997), and a concurring accom-panying editorial (the thrust of which is that everyoneprobably bene®ts from statins, the question is only inwhom to deny treatment on grounds of cost) actually®nished by stating that the issue now is ``to help theclinician answer the increasingly di�cult question ofwhom not to treat'' (Pearson, 1998). Expansion of lipidlowering therapy to include younger healthy subjectswith ``normal'' cholesterol values, as is proposed, mayincrease the public health impact of an associationbetween low cholesterol and violence, if one exists. Ourstudy used a speci®c but insensitive measure of violence;more sensitive (but similarly speci®c) measures shouldbe sought.Does low cholesterol cause violent o�ending? This

study cannot establish the answer. The magnitude of(unintentional) drop in cholesterol has been linked tosubsequent suicide (Zureik et al., 1996) and to the riskof postpartum depression (Ploeckinger et al., 1996).Experimental studies in our nonhuman relatives ®nd

excess aggressive behavior in those assigned or rando-mized to cholesterol reduction (Kaplan et al., 1991,1994); and meta-analyses of clinical trials in humanshave found an excess of violent deaths in men (withoutheart disease) randomized to lipid lowering treatment(Muldoon et al., 1990, 1993; Davey Smith et al., 1992;Ravnskov, 1992). Those ®ndings suggest a causal con-nection; moreover in the present study, the measure ofcholesterol preceded the violent acts (consistent with butnot necessarily indicative of causality). Thus, a causalconnection between low cholesterol and perpetration ofviolence cannot be excluded. Low cholesterol is unlikelyto serve as a useful predictor of future violence whenconsidered in isolation; nonetheless, if a relation of lowcholesterol to violence is con®rmed to be causal, it ispossible that manipulations of lipids or serotoninthrough diet or drugs could be used to prevent violence;or that risk factors that amplify the behavioral impactof low cholesterol could be assessed, and where possible,addressed.Further research is needed to more con®dently estab-

lish or reject a causal association between low and low-ered cholesterol and violent crime; and, if theassociation is con®rmed, to identify who is at greatestrisk (and whether factors de®ning risk are remediable);and to ascertain whether elevation of low cholesterolwill reverse the propensity to violence associated withlow cholesterol, or whether augmentation of tryptophanor serotonin will do so. Such research will help informpublic health policy regarding interpretation of choles-terol levels and associated risks; and will enable moreeducated targeting of cholesterol screening, treatment,and population based e�orts at cholesterol reduction.While present evidence should not controvert currentrecommendations to undertake cholesterol reduction inthose for whom de®nite bene®t to overall mortality hasbeen shown Ð middle aged men with pre-existing cor-onary disease or otherwise at markedly elevated cardio-vascular risk Ð this evidence does add to concernsregarding cholesterol reduction in other groups forwhom proof of overall mortality bene®t is lacking. Ifthe relation between low cholesterol and violent crimesis con®rmed to be causal, careful consideration must begiven not only to clinical implications of possibleincreased violent morbidity in the treated individual,but to potentially explosive ethical implications ofshifting the distribution of violence and death to onegroup of individuals through treatment of cholesterol inanother.

Acknowledgements

We thank Dr. Gunnar Eklund for assisting withaccess to the data; Dr. Naihua Duan for generous sta-tistical advice; Susan Stack and Brittany Mache for


administrative assistance; Shannon Alma Bush forassistance with references; Dr. Samuel Bozzette and Dr.Robert Brook for comments on the manuscript; and Dr.Paul Shekelle, Dr. Hal Morgenstern, and particularlyDr. Robert Brook for advice throughout the study. Thiswork was supported by the Robert Wood JohnsonFoundation and the Harry Frank Guggenheim Foun-dation.

References

Abe M, Nakai H, Tabata R, Saito K, Egawa M. E�ect of 5-[3-[((2S)-

1,4-benzodioxan-2-ylmethyl)amino]propoxy]-1,3-benzodioxole HCl

(MKC-242), a novel 5-HT1A-receptor agonist, on aggressive beha-

vior and marble burying behavior in mice. Japanese Journal of

Pharmacology 1998;76:297±304.

Asberg M. Neurotransmitters and suicidal behavior. The evidence

from cerebrospinal ¯uid studies. Annals of the New York Academy

of Sciences 1997;836:158±81.

AÊ sberg M, Traskmen L, Thoren P. 5-HIAA in the cerebrospinal ¯uid:

A biochemical suicide predictor? Arch Gen Psychiatr 1976;33:1193±

7.

Berzsenyi P, Galateo E, Valzelli L. Fluoxetine activity of muricidal

aggression induced in rats by p-chlorophenylalanine. Aggressive

Behavior 1983;9:333±8.

Beshay H, Pumariega A. Sertraline treatment of mood disorder asso-

ciated with prednisone: a case report. Journal of Child and Adoles-

cent Psychopharmacology 1998;8:187±93.

Bolander A. Mortality statistics in Sweden and its neighbouring

countries. Advantages and hazards inherent in systems and materi-

als. In: Bostrom H, Ljungstedt N, editors. Medical aspects of mor-

tality statistics. Stockholm: Almqvist and Wiksell, 1981.

Bramblett C, Coelho A, Mott G. Behavior and serum cholesterol in a

social group of Cercopithecus aethiops. Primates 1981;22:96±102.

Brennan P, Mednick S, John R. Specialization in violence: evidence of

a criminal subgroup. Criminology 1989;27:437±53.

Brown G, Linnoila M. CSF serotonin metabolite (5-HIAA) studies in

depression, impulsivity, and violence. J Clin Psychiat 1990;51:31±41.

Brunner D, Hen R. Insights into the neurobiology of impulsive beha-

vior from serotonin receptor knockout mice. Annals of the New

York Academy of Sciences 1997;836:81±105.

Chen Z, Peto R, Collins R, MacMahon S, Lu J, Li W. Serum choles-

terol concentration and coronary heart disease in population with

low cholesterol concentrations. BMJ 1991;303:276±82.

Cleare A, Bond A. Does central serotonergic function correlate inver-

sely with aggression A study using D-fen¯uramine in healthy sub-

jects? Psychiatry Research 1997;69:89±95.

Coccaro E. Central serotonin and impulsive aggression. Br J Psy-

chiatry 1989;155:52±62.

Coccaro E, Kavoussi R. Fluoxetine and impulsive aggressive behavior

in personality-disordered subjects. Arch Gen Psychiat 1997;54:1081±

8.

Cologer-Cli�ord A, Ng S, Lu S, Smoluk S. Serotonin agonist-induced

decreases in intermale aggression are dependent on brain region and

receptor subtype. Pharmacology, Biochemistry and Behavior

1997;58:425±30.

Cummings P, Psaty B. The association between cholesterol and death

from injury. Ann Intern Med 1994;120:848±55.

Davey Smith G, Pekkanen J. Should there be a moratorium on the use

of cholesterol-lowering drugs. Br Med J 1992;304:431±4.

Downs JR, Beere PA, Whitney E, Clear®eld M, Weis S, Rochen J, et

al. Design & rationale of the Air Force/Texas Coronary Athero-

sclerosis Prevention Study (AFCAPS/TexCAPS). Am J Cardiol

1997;80:287±93.

Drossman D, Talley N, Leserman J, Olden K, Barreiro M. Sexual and

physical abuse and gastrointestinal illness: review and recommen-

dations. Ann Intern Med 1995;123:782±94.

Gallerani M, Manfredini R, Caracciolo S, Scapoli C, Molinari S,

Fersini C. Serum cholesterol concentrations in parasuicide. BMJ

1995;310:1632±6.

Gedye A. Buspirone alone or with serotonergic diet reduced aggres-

sion in a developmentally disabled adult. Biological Psychiatry

1991;30:88±91.

Girgin FK, Sozmen EY, Hasegeli S, Hanci H, Ersoz B. Serum cho-

lesterol and aggression. In: XIII. National Congress of Biochem-

istry. Antalya: Turkey, 1996.

Golier J, Marzuk P, Leon A, Weiner C, Tardi� K. Low serum cho-

lesterol and attempted suicide. Am J Psychiatry 1995;152:3.

Golomb B. Cholesterol and violence: Is there a connection? Annals of

Internal Medicine 1998;128:478±87.

Gray R, Corrigan F, Strathdee A, Skinner E, van Rhijn A, Horrobin

D. Cholesterol metabolism and violence: a study of individuals

convicted of violent crimes. Neuroreport 1993;4:754±6.

Hibbeln J, Linnoila M, Umhau J, Rawlings R, George D, Salem NJ.

Essential fatty acids predict metabolites of serotonin and dopamine

in cerebrospinal ¯uid among healthy control subjects, and early-

and late-onset alcoholics. Biological Psychiatry 1998a;44:235±42.

Hibbeln J, Salem NJ. Dietary polyunsaturated fatty acids and depres-

sion: when cholesterol does not satisfy. Am J Clin Nutr 1995;62:1±9.

Hibbeln J, Salem NJ. Cholesterol lowering drugs alter polyunsaturated

fatty acid levels. Biol Psychiatry 1996;40:686±7.

Hibbeln J, Umhau J, George D, Salem N. Do plasma polyunsaturates

predict hostility and depression? Metabolic and Behavioral Aspects

in Health and Disease 1997;82:175±86.

Hibbeln J, Umhau J, Linnoila M, George D, Ragan P, Shoaf S, et al.

A replication study of violent and nonviolent subjects: cerebrospinal

¯uid metabolites of serotonin and dopamine are predicted by

plasma essential fatty acids. Biological Psychiatry 1998b;44:243±9.

Higley J, Linnoila M. Low central nervous system serotonergic activ-

ity is traitlike and correlates with impulsive behavior. A nonhuman

primate model investigating genetic and environmental in¯uences

on neurotransmission. Annals of the New York Academy of Sci-

ences 1997;836:39±56.

Higley J, Mehlman P, Taub D, Higley S, Suomi S, Linnoila M, et al.

Cerebrospinal ¯uid monoamine and adrenal correlates of aggression

in free-ranging rhesus monkeys. Arch Gen Psychiatry 1992;49:436±

41.

Hillbrand M, Foster H. Serum cholesterol levels and severity of

aggression. Psychological Reports 1993;72:270.

Hillbrand M, Spitz R, Foster H. Serum cholesterol and aggression in

hospitalized male forensic patients. J Behav Med 1995;18:33±43.

Holinger P, Klemen E. Violent deaths in the United States, 1990±75:

relationships between suicide, homicide and accidental deaths. Soc

Sci Med 1982;16:1929±38.

Jacobs D, Blackburn H, Higgins M, Reed D, Iso H, McMillan G, et

al. Report of the Conference on Low Blood Cholesterol: Mortality

associations. Circulation 1992;86:1046±60.

Kant R, Smith-Seemiller L, Zeiler D. Treatment of aggression and

irritability after head injury. Brain Injury 1998;12:661±6.

Kaplan J, Manuck S. The e�ects of fat and cholesterol on aggressive

behavior in monkeys. Psychosomatic Med 1990;52:226±7.

Kaplan J, Muldoon M, Manuck S, Mann J. Assessing the observed

relationship between low cholesterol and violence-related mortality.

Implications for suicide risk. Annals of the New York Academy of

Sciences 1997;836:57±80.

Kaplan J, Shively C, Fontenot D, Morgan T, Howell S, Manuck S,

Muldoon M, Mann J. Demonstration of an association among

dietary cholesterol, central serotonergic activity, and social behavior

in monkeys. Psychosom Med 1994;56:479±84.

Kaplan JR, Manuck SB, Shively C. The e�ects of fat and cholesterol

on social behavior in monkeys. Psychosom Med 1991;53:634±42.


Kruesi M. Cruelty to animals and CSF 5-HIAA. Psychiatry Res

1989;28:115±6.

Law M, Thompson S, Wald N. Assessing possible hazards of reducing

serum cholesterol. BMJ 1994;308:373±9.

Leiker M, Hailey B. A link between hostility and disease: poor health

habits? Behavioral Med Fall:129±33

Lidberg L, Tuck J. AÊ sberg, M., Scalia-Tomba, G., Bertilsson, L.

Homicide, suicide and CSF 5HIAA. Acta Psychiatrica Scandinavica

1985;71:230±6.

Lindberg G, Eklund G, Gullberg B, RaÊ stam L. Serum sialic acid con-

centration and cardiovascular mortality. BMJ 1991;302:143±6.

Lindberg G, Rastam L, Gullberg B, Eklund G. Low serum cholesterol

concentration and short term mortality from injuries in men and

women. BMJ 1992;305:277±9.

LinnoilaM, VirkkunenM, ScheininM, Nuutila A, RimonR, Goodwin

F. Low cerebrospinal ¯uid 5-hydroxyindoleacetic acid concentration

di�erentiates impulsive from nonimpulsive violent behavior. Life

Sciences 1983;33:2609±14.

Lucki I. The spectrum of behaviors in¯uenced by serotonin. Biological

Psychiatry 1998;44:151±62.

Manuck S, Flory J, McCa�ery J, Matthews K, Mann J, Muldoon M.

Aggression, impulsivity, and central nervous system serotonergic

responsivity in a nonpatient sample. Neuropsychopharmacology

1998;19:287±99.

Martinez R. Injury prevention: a new perspective [editorial; comment].

JAMA 1994;272:1541±2.

McDougle C, Brodkin E, Naylor S, Carlson D, Cohen D, Price L.

Sertraline in adults with pervasive developmental disorders: a pro-

spective open-label investigation. Journal of Clinical Psycho-

pharmacology 1998;18:62±6.

McMillen B, Ghamberlain J, DaVanzo J. E�ects of housing and

muricidal behavior on serotonergic receptors and interactions wtih

novel anxiolytic drug. J Neurotransmission 1988;71:123±32.

Mehlman P, Higley J, Faucher I, Lilly A, Taub D, Vickers J, Suomi S,

Linnoila M. Low CSF 5-HIAA concentrations and severe aggres-

sion and impaired impulse control in nonhuman primates. Am J

Psychiatry 1994;151:1485±91.

Mitchell P, Redfern P. Potentiation of the time-dependent, anti-

depressant-induced changes in the agonistic behavior of resident

rats by the 5-HT-sub(1A) receptor antagonist, WAY-100635. Beha-

vioural Pharmacology 1997;8:585±606.

Modai I, Valevski A, Dror S, Weizman A. Serum cholesterol levels

and suicidal tendencies in psychiatric inpatients. J Clin Psychiatry

1994;55:252±4.

Moechars D, Gilis M, Kuiperi C, Laenen I, Van Leuven F. Aggressive

behaviour in transgenic mice expressing APP is alleviated by ser-

otonergic drugs. Neuroreport 1998;9:3561±4.

Moeller F, Allen T, Cherek D, Dougherty D, Lane S, Swann A. Ipsa-

pirone neuroendocrine challenge: relationship to aggression as mea-

sured in the human laboratory. Psychiatry Research 1998;81:31±8.

Molina V, Gobaille S, Mandel P. E�ects of serotonin-mimetic drugs

on mouse-killing behavior. Aggressive Behavior 1986;12:201±11.

Morand C, Young S, Ervin F. Clinical response of aggressive schizo-

phrenics to oral tryptophan. Biol Pyschiatr 1983;18:575±8.

Mufti R, Balon R, Arfken C. Low cholesterol and violence. Psychia-

tric Services 1998;49:221±4.

Muldoon M, Manuck S, Matthews K. Lowering cholesterol con-

centrations and mortality: a review of primary prevention trials. Br

Med J 1990;301:309±14.

Muldoon M, Kaplan J, Manuck S, Mann J. E�ects of a low-fat diet

on brain serotonergic responsivity in cynomolgus monkeys. Biol

Psychiatry 1992;31:739±42.

Muldoon M, Rossouw J, Manuck S, Gluech C, Kaplan J, Kaufmann

P. Low or lowered cholesterol and risk of death from suicide and

trauma. Metabolism 1993;42:45±56.

National Council for Crime Prevention (Sweden). The Swedish penal

code. Stockholm, 1990.

Neaton J, Blackburn H, Jacobs D, Kuller L, Lee D, Sherwin R, et al.

Serum cholesterol level and mortality ®ndings for men screened in

the multiple risk factor intervention trial. Arch Intern Med

1992;152:1490±500.

Olivier B, Van Aken H, Jaarsma I, van Oorschot R, Zethof T, Brad-

ford D. Behavioural e�ects of psychoactive drugs on agonistic

behaviour of male territorial rats (resident-intruder model). In: Oli-

vier B, Miczek K, Kruk M, editors. Ethopharmacological aggres-

sion research. New York: Alan R Liss, 1984. p. 137±156.

Patterson S, Gottdiener J, Hecht G, Vargot S, Krantz D. E�ects of

acute mental stress on serum lipids: mediating e�ects of plasma

volume. Psychosomatic Med 1993;55:525±32.

Pearson T. Lipid-lowering therapy in low-risk patients. JAMA

1998;279:1659±61.

Ploeckinger B, Dantendorfer K, Ulm M, Baischer W, Der¯er K,

Musalek M, Dadek C. Rapid decrease of serum cholesterol and

postpartum depression. BMJ 1996;313:664±79.

Raleigh M, McGuire M. Serotonin, aggression, and violence in vervet

monkeys. In: Masters R, McGuire M, editors. The neurotransmitter

revolution. Carbondale: Southern Illinois University Press, 1994. p.

129±45.

Ratey J, Sovner R, Parks A, Rogentine K. Buspirone treatment of

aggression and anxiety in mentally retarded patients: a multiple-

baseline, placebo lead-in study. J Clinical Psychiatry 1991;52:159±

62.

Ravnskov U. Cholesterol lowering trials in coronary heart disease:

frequency of citation and outcome. BMJ 1992;305:15±19.

Ringo D, Lindley S, Faull K, Faustman W. Cholesterol and serotonin:

Seeking a possible link between blood cholesterol and CSF 5-HIAA.

Biol Psychiatry 1994;35:957±9.

Sanchez C. Interaction studies of 5-HT1A receptor antagonists and

selective 5-HT reuptake inhibitors in isolated aggressive mice. Eur-

opean Journal of Pharmacology 1997;334:127±32.

Saudou R, Amara D, Dierich A, LeMeur M, Ramboz S, Segu L, et al.

Enhanced aggressive behavior in mice lacking 5-HT1B receptor.

Science 1994;265:1875±8.

Scandinavian Simvastatin Survival Study Group. Randomised trial of

cholesterol lowering in patients with coronary heart disease: the

Scandinavian Simvastatin survival study (4S). Lancet

1994;344:1383±9.

Schuit A, Dekker J, Schouten E, Kok F. Low serum cholesterol and

death due to accidents, violence, or suicide [letter] [published erra-

tum appears in Lancet 1993 May 1;341(8853):1150]. Lancet 1993;

341:827.

Smith T, Frohm K. What's so unhealthy about hostility construct

validity and psychosocial correlates of the Cook and Medley Ho

scale? Health Psychol 1985;4:503±20.

Spitz R, Hillbrand M, Foster HJ. Serum cholesterol levels and fre-

quency of aggression. Psychol Rep 1994;74:622.

Steegmans P, Fekkes D, Hoes A, Bak A, van der Does E, Grobbee D.

Low serum cholesterol concentration and serotonin metabolism in

men. BMJ 1996;312:221.

Sullivan P, Joyce P, Bulik C, Mulder R, Oakley-Browne M. Total

cholesterol and suicidality in depression. Biol Psychiatry

1994;36:472±7.

Takei N, Kunugi H, Nanko S, Aoki H, Iyo R, Kazamatsuri H. Low

serum cholesterol and suicide attempts [letter]. Brit J Psychiatry

1994;164:702±3.

Traskman L. AÊ sberg, M., Bertilsson, B., & Sjostrand, L. Monoamine

metabolites in CSF and suicidal behavior. Arch Gen Psychiatr

1981;38:631±6.

van Praag H, Kahn R, Asnis G, Wetzler S, Brown S, Bleich A, et al.

Denosologization of biological psychiatry or the speci®city of 5-HT

disturbances in psychiatric disorders. J A�ective Disorders

1987;13:1±8.

Virkkunen M. Serum cholesterol in antisocial personality. Neu-

ropsychobiology 1979;5:27±30.


Virkkunen M. Insulin secretion during the glucose tolerance test in

antisocial personality. Brit J Psychiat 1983;142:598±604.

Virkkunen M, Penttinen H. Serum cholesterol in aggressive conduct

disorder: a preliminary study. Biol Psychiat 1984;19:435±9.

Virkkunen M, De Jong J, Bartko J, Linnoila M. Psychobiological

concomitants of history of suicide attempts among violent o�enders

and impulsive ®re setters. Arch Gen Psychiatry 1989;46:604±6.

Virkkunen M, Rawlings R, Tokola R, Poland R, Guidotti A, Nemer-

o� C, et al. CSF biochemistries, glucose metabolism, and diurnal

activity rhythms in alcoholic, violent o�enders, ®re setters, and

healthy volunteers. Arch Gen Psychiatry 1994;51:20±7.

Wolfgang M. Foreward. In: Mednick S, Christiansen K, editors. Bioso-

cial bases of criminal behavior. New York: Gardner, 1977. pp. v±vi.

Zak B, Dickenman R, White E, Burnett H, Cherney P. Rapid estima-

tion of free and total cholesterol. Am J Clin Pathol 1954;24:1307±15.

Zureik M, Courbon D, Ducimetiere P. Serum cholesterol concentra-

tion and death from suicide in men: Paris Prospective Study I. BMJ

1996;313:649±50.


Documents

Low cholesterol and violent crime