Context.-Forensic pathologists regularly investigate the deaths of individuals with a history of drug abuse. Autopsy, including toxicology testing, reveals no cause for death in a subset of this cohort.

Objective.-To determine whether deaths with an undetermined cause and manner of death are associated with a history of drug abuse.

Design.-Retrospective matched case-control study of 52 decedents whose cause of death remained undetermined following autopsy, matched 1:2 to a control group of living patients admitted for cholecystectomy according to age and date of death or procedure.

Results.-Individuals whose cause of death was undetermined were 5.3 times (95% confidence interval, 1.9- 14.5) more likely to have a history of drug abuse than were patients with cholecystitis.

Conclusions.-Decedents with a history of chronic drug abuse appear to be at an increased risk of dying by their chronic drug abuse, even in the absence of any anatomical or toxicologic finding at autopsy to account for death.

(Arch Pathol Lab Med. 2008;132:1903-1906)

Occasionally a young adult dies suddenly and unexpectedly and investigation by the medical examiner's office reveals neither anatomical nor toxicologic cause for death. Often the only information suggestive of a possible cause of death is a history of drug abuse. We hypothesize that the individuals in these undetermined deaths, where neither anatomical nor toxicologic cause for death is found at autopsy, share a common history of drug abuse because drug abuse induces some change that persists after the drug is no longer detectable in the body. If our hypothesis is true, then this change, which is presumably at the molecular level, would increase the likelihood of sudden death. In an earlier study, we compared deaths that were undetermined in cause and manner with other deaths that were accidental.1 In that study we showed that evidence of drug abuse was 4.2 times more common in the undetermined deaths than in the accidental deaths.We continue to investigate this hypothesis by conducting a retrospective case-control study comparing the case findings and history of drug abuse in deaths examined in our office that had an undetermined cause of death with a control group of living patients who presented to a hospital for cholecystectomy. Our purpose in this study was to evaluate the reproducibility of our previous study by determining whether drug abuse is more prevalent in individuals who died of undetermined causes than in a separate and distinct control group, namely, a sample of living patients evaluated at an indigent care hospital.

MATERIALS AND METHODS

We conducted a retrospective matched case-control study of deaths investigated by the Jefferson County Coroner/Medical Examiner Office, Birmingham, Alabama between 1986 and 2003. During this time the medical examiner's office was operating under a consistent medical examiner statute, and all deaths falling under its jurisdiction were certified by 1 of 5 forensic pathologists. The statute charges the office with the responsibility of investigating all sudden and unexpected deaths that have occurred in Jefferson County, Alabama. The nature of deaths investigated by our office and the toxicologic screening for drugs of abuse remained constant throughout the course of this study. Cases for this study were identified by a computer search of our office database for all individuals between 10 and 70 years of age whose cause and manner of death remained undetermined following an autopsy and toxicologic analysis for ethanol and drugs of abuse. (Drug screens are performed on urine, bile, or an extract of blood or other tissue using an enzyme multiplied immunoassay technique that tests for amphetamines, barbiturates, benzodiazepines, Cocaine metabolite, Opiates, propoxyphene, tricyclic antidepressants, and methadone. Positive screens are confirmed with gas chromatography/mass spectrometry.) Cases in which the cause of death was undetermined because of the advanced state of decomposition were excluded from further study (14 cases deleted). Note that undetermined deaths in this office and in this study specifically excluded cases with heart disease sufficient to account for sudden death because in any case in which heart disease sufficient to account for death was present, then the cause of death would be ascribed to that specific form of heart disease, and thus the cause would not be listed as ''undetermined.'' Following this algorithm, we produced 52 cases.

The control group was chosen from living patients who un derwent cholecystectomy at the indigent care hospital serving Jefferson County, Alabama, because the patient population at this hospital is most similar to the individuals who die in circumstances that fall under the jurisdiction of the medical examiner office (ie, many of the decedents evaluated at the medical examiner office are indigent and received health care during life from the indigent care hospital). Patients with cholecystitis were chosen to serve as the control group because the pathogenesis of cholecystitis is generally accepted to be unrelated to drug use, thus avoiding the bias that would occur if cholecystitis and drug abuse were related.2 A control group of patients was identified by a computerized search of hospital medical records to identify all patients presenting for cholecystectomy between 1988 and 2003. Two controls were matched to each study group member within 5 years of the age of the study decedent and within 2 calendar years of the date of death of the study decedent (to keep social trends comparable). The first suitable controls found for a study case were paired with the study cases and then removed from further consideration for matching. If a patient chart was unavailable for review, the study case was rematched with another patient chart using the same parameters.

The charts of all cases were reviewed for the circumstances surrounding death and for history or signs of drug abuse. In cases in which the cause of death remained undetermined following an autopsy, no physical finding that would reasonably account for death in light of the circumstances surrounding death was found on review, that is, in none of these cases did the decedent have sufficient disease, such as cardiac hypertrophy, anomalous coronary artery circulation, severe coronary artery atherosclerosis, myocarditis, pneumonia, cirrhosis, and intracranial hemorrhage, to explain death. All toxicology results were also recorded.

The charts of the control group were reviewed for history or signs of drug abuse as documented in the physicians' history and physical examination. Blood pressure and history of hypertension were also recorded. The charts of the living control group were reviewed under the approval of and in accordance with the hospital institutional review board. For some patients the chart contained an explicit statement that the patient had a positive history of drug abuse. In other patients the chart explicitly stated that the decedent had no history of drug abuse. In still other cases the chart stated that the patient had no history of substance abuse or was a social drinker, and lack of drug abuse was inferred from the lack of risk factors for infection with human immunodefi- ciency virus, as documented by the surgeon in the preoperative note.

A history of drug abuse was considered positive in this study if either one of the following instances was documented in an individual's chart:

1. A history of drug abuse was elicited by questioning (of relatives, acquaintances, or physicians of the decedents by the death investigator or of the patient by the surgeon);

2. Signs on external physical examination that revealed compelling evidence of drug abuse, that is, needle track marks or nasal septum perforation (found either by the pathologist examining the decedent or the surgeon examining the patient).

The risk odds ratio was calculated as the measure of association for this study.

RESULTS

This study consisted of 52 decedents in the study group matched 1:2 with living controls for a total of 104 controls. The results are shown in the Table. Within the study group the male-female ratio was 38:14, and in the control group the ratio was 19:85. Evidence of hypertension was present in 15 cases in the control group and in 5 cases in the study group.

Comparison of the study group to the control group shows a risk odds ratio of 6.6 (95% confidence interval, 2.9-15.4; P < .001) for a history of drug abuse. In other words, in this study an individual with an undetermined cause and manner of death is 6.6 times more likely to have evidence of a history of drug abuse compared with a living patient who presents for cholecystectomy.

Conditional logistic regression showed that 3 factors differed significantly between the study group and the control group. One factor was sex; males were more likely to have died from an undetermined cause than females (risk odds ratio, 8.6; 95% confidence interval, 3.4-22.1). Individuals with hypertension were less likely to have died from an undetermined cause (risk odds ratio, 0.44; 95% confi- dence interval, 0.10-2.0). After adjusting for sex and hypertension, those individuals who died of an undetermined cause were still 5.3 times more likely to have a history of drug abuse than were those individuals with cholecystitis (95% confidence interval, 1.9-14.5).

COMMENT

We compared the history of drug use in a group of individuals who died suddenly with an undetermined cause of death to the history of drug abuse in a control group of living patients presenting for cholecystectomy. Our study shows that an individual dying of an undetermined cause of death is 5.3 times more likely to have a history of drug abuse than is a patient presenting for cholecystectomy. There are 4 possible causes for such a large risk odds ratio-chance, bias, a confounding factor, or a valid relationship suggesting cause and effect. Chance is an unlikely explanation for the results of this study as evidenced by the small P value (ie, P < .001). Confounding was addressed by the use of matching and statistical adjustment, allowing us to see the association of a history of drug abuse with sudden death independent of other potential, measured characteristics. Of course, residual confounding is always possible and would occur if we overlooked a characteristic that could explain the observed association.

Bias, also known as systematic error, can falsely alter the risk odds ratio, and it is the job of experimenters to design their study so that bias is eliminated, or at least minimized. The control group selected for a study is a common source of bias. For this reason we chose cholecystitis as the disease present in the control group, because cholecystitis seems to develop independent of drug abuse.2 Matching by year of death eliminated bias by keeping the prevailing trends of drug use similar for both the study group and the control group, so that changes in the local availability of drugs of abuse and of their form (cocaine hydrochloride vs Crack cocaine) were similar.

We did not attempt to match for sex, and analysis of our data shows that males were associated with having an undetermined cause of death. This association may be due to the selection of patients with cholecystitis as the control group, because cholecystitis is more common in women. It is also possible that this association between being male and dying of an undetermined cause is true. Our data also indicate that individuals with hypertension were less likely to have died of an undetermined cause of death. As we have shown in a previous study, this lack of a history of hypertension in individuals with an undetermined cause of death is the result of selection bias; most decedents with a history of hypertension who are examined in a medical examiner office have a cause of death from some form of heart disease visible at autopsy. A sample of individuals with an undetermined cause of death will include few with a history of hypertension in any medical examiner population.3

A potential bias that would undermine our interpretation of our results is the possibility that a history of drug abuse was sought less diligently by either the pathologists or the surgeons, thus making the 2 groups incomparable. In a separate study we have shown that history alone is inefficient at detecting all cases of possible drug abuse, compared with toxicologic screening for drugs of abuse in all cases.4 Nevertheless, we believe that we have not introduced bias into this study by using history as our primary means of ascertaining a lifestyle of drug abuse for 3 reasons. First, we treated both the study group and the control group identically regarding our attempt to determine a history of drug abuse. If we were unable to ascertain a history of drug abuse in a decedent in the study group, then we listed that decedent as having no history of drug abuse, even if we found talc granulomata in the lungs at autopsy. Because we were unable to assess the living individuals in the control group for pulmonary talc granulomata, it would bias our results to use information available from one group but not from the other. Furthermore, although the goal of forensic pathology differs from that of surgery, both pathologists and surgeons have good cause to seek diligently for evidence of drug abuse and associated pathogens such as viruses that cause hepatitis or immunodeficiency. We have no reason to believe that either the surgeons or the pathologists looked more or less diligently for a history of drug abuse in their patient populations, leading us to think that bias of this sort is unlikely in our study. Finally, we think this bias unlikely because our results are similar to those in our previous study, where we found a risk odds ratio of 4.2 (95% confidence interval, 1.6-11.1).1 In the 26 study cases in the present study in which we found no history of drug abuse we found only 2 cases with toxicology findings that indicated our history failed to reveal the decedent's drug abuse. The inclusion or exclusion of those 2 cases (and their controls) in statistical analysis gave identical risk odds ratios and confidence limits.

Causation is the final explanation for an elevated risk odds ratio, and we believe this is the proper explanation for our results. With respect to our hypothesis that drug abuse induces changes that increase the risk of sudden death without leaving evidence demonstrable at autopsy, several lines of evidence already exist. One line is that such a concept with drugs of abuse is analogous to a chronic alcoholic dying from sequelae of chronic Alcohol abuse without being intoxicated to the time of death, which certainly happens.5 Consistent findings also support an argument for causation, and this study indicates a relationship between sudden death and a history of drug use, just as our previous study did. Moreover, this relationship is strong, and strong relationships are more likely to prove lasting than tenuous relationships. Epidemiologists refer to calculations such as the risk odds ratio as a ''measure of association'' because the risk odds ratio is a measure of how strong the association is between the risk factor and the outcome being studied-the higher the risk odds ratio, the stronger the association.

Yet another line of evidence supporting the hypothesis that chronic drug abuse can cause death even in the absence of acute intoxication is provided by research results published by other researchers that are in keeping with the hypothesis. For example, in human addicts, chronic Cocaine use has been shown to alter endothelial cell function, rendering the addicts susceptible to ischemic vascular accidents by thrombosis of injured vessels, by accelerated atherosclerosis, or by vasoconstriction and vasospasm.6 In dogs, Jones and Tackett7 found that chronic Cocaine treatment enhances the responsiveness of the left anterior descending coronary artery and the femoral artery to vasoactive substances, enhancing peripheral vasoconstriction and cardiac ischemia. Vasoconstriction and vasospasm need not leave visible evidence of occurrence at autopsy. Karch et al8 suggest that drug abuse causes chronic detrimental changes to the heart in their study that compared heart weights of Cocaine users with a nonuser control group. Karch et al found that the degree of myocardial hypertrophy is statistically significantly higher in men who used Cocaine, but because the increase is slight (10%) it may be underappreciated, or even unrecognizable, at autopsy in an individual case. Finally, Mehta et al9 report that human infants with intrauterine Cocaine exposure had a decrease in both vagal and global heart rate variability indices compared with a control group of infants not exposed to cocaine. Mehta et al examined these infants during sleep 2 weeks after birth, when no measurable level of Cocaine remained in the infants. Mehta et al opine that the differences they observed may be the result of altered cardiorespiratory influence of sleep on heart rate variability or may reflect nonacute effects of Cocaine on the autonomic system. Studies like these also provide evidence toward the mechanisms of sudden death that we are describing in our study.

CONCLUSION

This is our second study to show that a history of drug abuse is far more common in decedents with an undetermined cause of death than it is in a control group chosen to represent a random sample of the population. Epidemiologic theory indicates that for rare events, such as the death of an individual with a history of drug abuse, the measure of the association between the risk (drug abuse in this study) and the event (death in this study) is a valid and accurate predictor of the incidence of death due to a given risk factor. In other words, individuals who abuse drugs are at increased risk of dying suddenly because of their habit of using drugs, even if not acutely intoxicated at the time of death. Although more research needs to be done to elucidate the mechanism of these deaths, based on these findings the implication is plain-decedents with a convincing history of drug abuse and no other anatomical or toxicologic findings at autopsy have been placed at increased risk of dying by their chronic drug abuse. We think it appropriate to list chronic drug abuse as the cause of death in this specific set of cases, just as may be done with chronic ethanol abuse. Like chronic ethanol abuse, the proper manner designation for these deaths from sequelae of chronic drug abuse is ''natural.''

[Reference]

References

1. Gruszecki AC, McGwin G, Robinson CA, Davis GG. Unexplained sudden death and the likelihood of drug abuse. J Forensic Sci. 2005;50:419-422.

2. Kaku DA, Lowenstein DH. Emergence of recreational drug abuse as a major risk factor for stroke in young adults. Ann Intern Med. 1990;113:821-827.

3. Davis GG, McGwin G Jr. Comparison of heart mass in seizure patients dying of sudden unexplained death in epilepsy to sudden death due to some other cause. Am J Forensic Med Pathol. 2004;25(1):23-28.

4. Gruszecki AC, Booth J, Davis GG. The predictive value of history and scene investigation for toxicology results in a medical examiner population. Am J Forensic Med Pathol. 2007;28(2):103-106.

5. DiMaio VJ, DiMaio D. Forensic Pathology. 2nd ed. Boca Raton, Fla: CRC Press; 2001.

6. Havranck EP, Nademanee K, Grayburn PA, Eichorn EJ. Endothelium-dependent vasorelaxation is impaired in Cocaine arteriopathy. J Am Coll Cardiol. 1996; 28:1168-1174.

7. Jones LF, Tackett RL. Chronic Cocaine treatment enhances the responsiveness of the left anterior descending coronary artery and the femoral artery to vasoactive substances. J Pharmacol Exp Ther. 1990;255:1366-1370.

8. Karch SB, Green G, Young S. Myocardial hypertrophy and coronary artery disease in male Cocaine users. J Forensic Sci. 1995;40:591-595.

9. Mehta SK, Super DM, Salvator A, et al. Heart rate variability in cocaineexposed newborn infants. Am Heart J. 2001;142:828-832.

10. Schlesselman JJ. Case-Control Studies: Design, Conduct, Analysis. New York, NY: Oxford University Press; 1982:213-216.