Fish Oils May Be Lifesavers
By Miranda Hitti
WebMD Medical News Reviewed By Michael Smith, MD
Wednesday, August 30, 2006
Aug. 30, 2006 - Fish oils in fatty fish like salmon might be even better than heart devices called defibrillators at preventing sudden death from heart problems.
"Choosing fish two or three times a week is a good idea," researcher Thomas Kottke, MD, MSPH, tells WebMD.
"Grilled, baked, or broiled -- not fried," he adds. "Fried fish appears to lose all of its benefits."
The study by Kottke and colleagues will appear in the American Journal of Preventive Medicine's October edition.
Kottke works in St. Paul, Minn., at Regions Hospital's Heart Center.
Sudden Death Risk
Kottke's team created a computer model to check sudden death risk in a fictional group of people aged 30-84 in Olmstead County, Minn.
The researchers tested several scenarios.
In one scenario, people ate adequate amounts of omega-3 fatty acids from fish or fish oil supplements (in reality, the typical Western diet is short on omega-3 fatty acids).
In another scenario, automated external defibrillators (AEDs) were available in people's homes and in all public areas.
AEDs are used to shock the heart back into action if it develops a fatal rhythm problem that can result in sudden death.
In a third scenario, people who needed implantable defibrillators because of heart failureheart failure got those devices. Heart failure greatly increases the chance of sudden death.
Fish Oils Trumped Defibrillators
All three scenarios lowered sudden death risk. But omega-3 fatty acids yielded the best results -- even in healthy people.
Sudden death risk dropped 6.4% with adequate omega-3 fatty acid intake, compared with 3.3% for implantable defibrillators, and less than 1% with easy access to AEDS, the study shows.
What's more, about three-quarters of the imaginary lives saved in the omega-3 group were healthy people, note Kottke and colleagues.
The researchers aren't saying defibrillators don't work. Those devices can save lives, Kottke's team writes.
In fact, sudden death risk was reduced most by combining all three scenarios - getting enough omega-3s, distributing AEDs, and giving appropriate patients implantable defibrillators.
But when it comes to omega-3 fatty acids, the old saying that an ounce of prevention is worth a pound of cure may sum up the study's findings.
Kottke's computer model was based on omega-3 fatty acids from fish.
But omega-3 fatty acids aren't just in fish. Other sources include walnuts, flaxseed, canola oil, broccoli, cantaloupe, kidney beans, spinach, grape leaves, Chinese cabbage, and cauliflower.
Still, "fish oil has a lot more omega-3s than flax, and that's the same with Āc walnuts," Kottke tells WebMD.
Fish oil supplements containing omega-3 fatty acids are another option.
If you eat fish two or three times weekly, do you still need supplements?
"Probably not," Kottke says. "It appears that that's adequate and that the benefit actually comes at fairly low levels of consumption."
Supplements aren't regulated as strictly as prescription drugs. So, if you opt for that source of omega-3, do your homework and choose a high-quality supplement from a reputable company.
If you do decide to take fish-oil pills, tell your doctor. That way, your doctor can keep track of all the medicines and supplements you're taking.
Not a Cure-All
Kottke stresses that his study didn't directly test omega-3 fatty acids in actual people to prevent sudden death. Such studies are being done in Italy and the U.K., he notes.
Eating fish or taking fish oil pills won't make up for smoking, inactivity, and other heart hazards, Kottke warns.
"We need to prioritize nutritionnutrition and physical activity right up there with brushing our teeth," he says.
His short list of lifestyle tips:
* Don't smoke.
* Eat a healthy diet rich in fruits and vegetables.
* Limit saturated fat.
* Get enough physical activity - for example, taking 10,000 steps per day (a pedometer can help you keep count).
* A limited amount of alcohol may also be healthy (maximum one drink a day for women, two drinks for men).
* Eat a small amount of nuts regularly.
Kottke says he sprinkles almonds, banana, and peaches on his breakfast cereal. His evening snack is a glass of wine and some almonds instead of cheese and crackers.
"Nuts are very good for you," Kottke says. But nuts are high in calories, so don't overdo it.
The bottom line: Your daily habits -- including what you put on your plate -- matters. "It makes a huge difference," Kottke says.
SOURCES: Kottke, T. American Journal of Preventive Medicine, October 2006; vol 31. Thomas Kottke, MD, MSPH, The Heart Center, Regions Hospital, St. Paul, Minn. Health Behavior News Service.
Preventing Sudden Death with n-3 (Omega-3) Fatty Acids and Defibrillators
American Journal of Preventive Medicine
Volume 31, Issue 4 , October 2006, Page 316
Thomas E. Kottke MD, MSPHa, , , Lambert A. Wu MDb, Lee N. Brekke PhDc, Mark J. Brekke MAc and Roger D. White MDd, e
aThe Heart Center, Regions Hospital, St. Paul, Minnesota
bCotton-O'Neil Heart Center, Topeka, Kansas
cBrekke Associates, Inc., Minneapolis, Minnesota
dDivision of Cardiovascular Diseases and Department of Anesthesiology, Mayo Clinic College of Medicine, Rochester, Minnesota, USA
eCity of Rochester Early Defibrillation Program, Rochester, Minnesota
Because interventions that prevent and treat events due to cardiovascular disease are applied to different, but overlapping, segments of the population, it can be difficult to estimate their effectiveness if formal calculations are not available.
Markov chain analysis, including sensitivity analysis, was used with a hypothetical population resembling that of Olmsted County, MN, aged 30 to 84 in the year 2000 to compare the estimated impact of three interventions to prevent sudden death: (1) raising blood levels of n-3 (omega-3) fatty acids, (2) distributing automated external defibrillators (AEDs), and (3) implanting cardioverter defibrillators (ICDs) in appropriate candidates. The analysis was performed in 2004, 2005, and 2006.
Raising median n-3 fatty acid levels would be expected to lower total mortality by 6.4% (range from sensitivity analysis=1.6% to 10.3%). Distributing AEDs would be expected to lower total mortality by 0.8% (0.2% to 1.3%), and implanting ICDs would be expected to lower total mortality by 3.3% (0.6% to 8.7%). Three fourths of the reduction in total mortality due to n-3 fatty acid augmentation would accrue from raising n-3 fatty acid levels in the healthy population.
Based on central values of candidacy and efficacy, raising n-3 fatty acid levels would have about eight times the impact of distributing AEDs and two times the impact of implanting ICDs. Raising n-3 fatty acid levels would also reduce rates of sudden death among the subpopulation that does not qualify for ICDs.
Sudden death due to cardiac disease imposes a significant burden on the population of the United States. Half a million people die suddenly each year,1 and, for half of these people, sudden death is the first indication that they have coronary heart disease.2 and 3 In some populations, sudden death is the heralding event in 15% of all cases of coronary heart disease.2
Prevention and treatment of sudden death currently focuses on three interventions: implantable cardioverter defibrillators (ICDs), automated external defibrillators (AEDs), and the consumption of n-3 fatty acids (n-3s). Each is supported by significant study evidence. There is randomized clinical trial evidence that ICDs reduce the risk of sudden death by preventing or treating ventricular tachycardia and ventricular fibrillation in individuals who have been previously resuscitated from ventricular fibrillation or who have depressed left ventricular systolic function.4, 5, 6, 7, 8, 9, 10, 11, 12 and 13 There is strong evidence that AEDs also reduce the risk of sudden death by interrupting the process as it is occurring. This is true both when an AED is used by a first responder and when used by a trained bystander in a public area.14, 15, 16, 17, 18, 19, 20, 21, 22 and 23 A meta-analysis commissioned by the Agency for Healthcare Research and Quality concluded that the evidence from primary and secondary prevention studies supports the hypothesis that consumption of n-3s, fish, and fish oil reduces all-cause mortality and various cardiovascular disease (CVD) outcomes such as sudden death, death from cardiac disease, and nonfatal myocardial infarction (MI).24 In observations and trials conducted in the United States and Europe, the predominant effect of high n-3 consumption is a reduction in risk of sudden death.
Each of these three interventions treats a different, but overlapping, segment of the population, and each has a different level of efficacy. Consequently, estimating the effectiveness of each is difficult without formal analysis. The analysis presented here compares the potential population impact of raising blood levels of n-3s for both primary and secondary prevention, distributing AEDs and training first responders and bystanders to use them, and implanting ICDs in appropriate candidates.
Discussion and Conclusions
The results of this analysis using central parameter values suggest that increasing blood n-3 levels would reduce total mortality by 6.4%. The n-3 effect would be eight times as large as one produced by implementing an AED strategy that placed devices in all public areas and all homes and would be twice as large as implementing a strategy that implanted ICDs in all individuals with a left ventricular EF of 30% or less. About three fourths of the reduction in deaths from the n-3 strategy would accrue from increasing n-3 levels in apparently healthy individuals, a group not normally the target of interventions to reduce the burden of sudden death. Sensitivity analysis indicates that under some assumptions the impact of the n-3 strategy and the ICD strategy may be equivalent, but the n-3 strategy is expected to have a larger impact than the AED strategy under all assumptions. Most of the situations where ICDs are more effective than n-3 supplementation involve the assumption that n-3s have little or no efficacy for primary prevention.
Despite the fact that AEDs do save lives, they are unlikely to ever have a substantial impact on rates of sudden death. This fact is due primarily to the epidemiology of sudden death. As shown in Table 1, there is a chain of events that must occur if an individual is to survive an out-of-hospital cardiac arrest. The arrest must be witnessed by a bystander, the rhythm must be ventricular fibrillation or ventricular tachycardia when the AED is applied, the individual must survive to hospital admission, and once in the hospital, the individual must survive to discharge. The leaders of the Public-Access Defibrillation trial17 suggest that implementing a nationwide public-access defibrillation policy would save 2000 to 4000 lives per year. Even the upper estimate is less than 1% of sudden deaths and only 0.16% of all deaths in the United States.
The potential impact of ICDs is mitigated by the fact that half of the individuals who suffer an out-of-hospital cardiac arrest have no history of cardiac disease and thus would not be candidates for an ICD before they sustained their cardiac arrest. Additionally, the majority of individuals who die suddenly after MI have an EF of more than 30%. They do not currently qualify for prophylactic ICD implantation. Implanting biventricular pacemakers in everyone who qualified for an ICD would marginally increase the effect of the intervention.12
It is the relative size of the subpopulation being treated, not the level of risk, that tends to determine the potential population impact of interventions to prevent sudden death. For example, the death rates for the subpopulation of individuals who have survived an out-of-hospital cardiac arrest or have an EF of 30% or less after an MI are ten times the death rates of the apparently healthy subpopulation. However, the number of expected deaths among the apparently healthy subpopulation is 13 times the number of deaths in the subpopulations with systolic congestive heart failure, a previous out-of-hospital cardiac arrest, or an EF of 30% or less after an MI. Hence, the number of deaths prevented by n-3s in the healthy subpopulation could be nearly three times the number of deaths prevented by ICDs and AEDs in the subpopulations with systolic congestive heart failure, a history of out-of-hospital cardiac arrest, or an EF of 30% or less after an MI.
Critical knowledge gaps exist
One of these is the uncertainty that increasing n-3 blood levels changes outcomes in apparently healthy individuals. Although the American Heart Association34 and the American College of Cardiology35 have endorsed increased n-3 intake by individuals who are apparently healthy, and recent quantitative analysis has concluded that increasing fish consumption would have a large net positive impact on public health,36 a working group of the National Institutes of Health Office of Dietary Supplements and the National Heart, Lung, and Blood Institute concluded that a definitive clinical trial is needed to test the effects of omega-3 fatty acid intake on CVD mortality.34 The need for such a trial is underscored by a recent British meta-analysis that concluded that n-3s have little impact on either total mortality or risk of death from heart disease.33 Ongoing trials in Italy37 and England38 have the potential to satisfy these justified concerns.
Two other gaps in knowledge are the extent to which the data sources bias the estimates and the extent to which the estimates apply to minority populations and other subpopulations in the United States. Sudden death rates of physicians, the experience of largely white Olmsted County, MN residents, and the experience of individuals who have volunteered for clinical trials of ICDs may not apply to either the entire population or particular subpopulations. The data for AEDs, because they are collected from multiple communities, are likely to be the most representative. Even so, it is impossible to provide any sort of estimates for subpopulations because data simply do not exist.
While a formal cost-effectiveness analysis is beyond the scope of this paper, an examination of program implementation costs is telling. A hypothetical American population of 100,000 adults aged 30 to 84 would be part of a total population of about 172,000, would have 86,000 in the labor force, and would occupy 67,000 households.39 A comprehensive AED program would require AEDs for first responders, in work sites, and in homes. Based on the goals for equipping first responders in Rochester MN (95,000 total population in an area of 45 square miles), 65 AEDs would be required for first responders. Supplying one work site AED for every 50 workers would require 1720 AEDs. One AED in every home would require 67,000 AEDs. At $3000 per device,40 initial costs for devices would be $195,000 to equip first responders, $5.2 million to equip work sites, and $201 million to equip households. If the life of an AED were 10 years, the annual replacement cost would be about $21 million per year. Training and AED maintenance costs would be additional expenses.
A recent cost-effectiveness analysis of ICDs has estimated that the lifetime cost of an ICD is between $68,000 and $101,000.41 Using an intermediate value ($85,000) and multiplying it by the number of new candidates per year (n =243) generates a cost of $21 million per year in the above hypothetical population.
As supplements n-3s can be purchased for as little as $0.16/1000 mg.42 If everyone in a community similar to the one in this simulation raised their n-3 levels through the use of a 1000-mg n-3 daily dietary supplement, the community cost would be $5.8 million per year. If however, a large proportion of the community ate fish high in n-3 content rather than other meat, much of the cost would be off-set.
Each one of the interventions considered here makes a particular contribution in easing the burden of sudden death. While AEDs will make relatively little impact on overall death rates, their effect on a few individuals who suffer an out-of-hospital cardiac arrest under relatively unique conditions is dramatic. The analysis shows that equipping first responders with AEDs is also relatively inexpensive. While the ICD strategy is more expensive and often less effective than the n-3 strategy, it is appropriate to provide devices to the class of patients who are recognizable as being at high risk of sudden death and for whom efficacy of ICDs has been demonstrated.
If increasing n-3 intake is indeed the efficacious primary prevention strategy that it appears to be in cohort studies, this population-wide strategy would have the greatest impact of the three strategies considered in this analysis. Increasing n-3 consumption would be less costly and more effective than an ICD strategy, it would reduce the risk of individuals who had not been identified as candidates for ICDs, and it would not require the unique situation that is required for an AED to have an impact. However, these outcomes have yet to be definitively demonstrated.
The distribution of the hypothetical population and the ranges of the estimates used at the start of each simulation appear in Table 2 along with the no-intervention scenario results. The largest subpopulation is the healthy one (n =94,734 of 100,000). The second largest subpopulation comprises individuals with previous MI and left ventricular EF >30% (n =3803). The subpopulations with congestive heart failure due to systolic and diastolic dysfunction each contain 659 individuals. A relatively small subpopulation (n =137) has a history of MI with severe left ventricular dysfunction. Nine individuals survived an out-of-hospital cardiac arrest.
Death rates in the subpopulations differ by as much as a factor of 20. Compared to the death rate for healthy individuals, the death rate for individuals with an MI and EF >30% is about four times larger, the rate for individuals with congestive heart failure or reduced left ventricular function after an MI is about ten times larger, and the death rate for individuals who have suffered an out-of-hospital cardiac arrest is 20 times larger.
Of the 912 (range=815 to 1013) deaths that occur during a simulated year, 700 (603 to 797) come from the healthy subpopulation, and 41 (20 to 82) each come from the subpopulations with congestive heart failure due to systolic and diastolic dysfunction. One (0.6 to 2.4) death occurs in the subpopulation that has survived an out-of-hospital cardiac arrest. The 117 (58 to 234) deaths in the subpopulation with EF >30% after an MI is nearly ten times the deaths in population with EF ĀÖ30% after an MI (n =12).
N-3 Fatty Acid Scenario
Raising blood n-3 levels is predicted by the analysis to save 58.2 (14.9 to 101.4) lives per year (Table 3), a reduction in total mortality of 6.4% (1.6% to 10.3%) (Table 4). About 75% of this impact would come from preventing deaths in the initially healthy subpopulation. Nearly all of the remaining impact would come from deaths prevented in the subpopulation with EF >30% after MI.
The analysis predicts that AEDs would be used about 36 times per year. An AED would be used 15.1 times for cardiac arrests occurring in the healthy subpopulation, 13.3 times for individuals with congestive heart failure, and 7.9 times for individuals with previous MI.
Because only a limited proportion of individuals on which AEDs are used ultimately survive, the AED scenario predicts that only 7.3 (1.8 to 12.1) deaths are prevented (Table 3), a reduction of 0.8% (0.2% to 1.3%) (Table 4). More than half (n =4.5) of the impact would come from preventing deaths in the initially healthy subpopulation. The next largest number of deaths prevented (n =1.2) would arise from the subpopulation with a left ventricular EF >30% after an MI.
In this scenario, about 243 ICDs would be implanted each year in the hypothetical population. A total of 211 of the ICDs would be placed in individuals with CHF, 2 in individuals who have survived a cardiac arrest, and 30 in individuals who have suffered an MI.
Implantable cardioverter defibrillators would be expected to prevent 30.5 (5.7 to 84.1) deaths per year (Table 3), a reduction of 3.3% (0.6% to 8.7%) (Table 4). Nearly half (n =13.7) of these would arise from the initially healthy subpopulation who developed a condition that warranted implantation of an ICD (congestive heart failure due to left ventricular systolic dysfunction, out-of-hospital cardiac arrest, or MI with an EFĀÖ30%). Among those 13.7, ICD implantation would prevent 11.7 deaths among individuals who developed CHF, 2.0 deaths among individuals who had survived an acute MI, and 0.05 deaths among individuals who had survived an out-of-hospital cardiac arrest. Most (n =12.6) of the remaining deaths prevented would come from the subpopulation that already had congestive heart failure at the beginning of the simulated year but no history of MI. Only a small number of deaths (n =3.7) would be prevented in the subpopulation with an EF ĀÖ30% after a previous MI or the subpopulation that had survived a previous out-of-hospital cardiac arrest (n =0.5).
Combined Intervention Scenario
If all three interventions were to be implemented simultaneously, the analysis predicts that total mortality would be reduced by 9.8% (5.3% to 14.8%), corresponding to 89.4 (48.1 to 143.1) deaths (Table 3 and Table 4). About half of the deaths prevented would be due to increasing blood n-3 levels in the apparently healthy subpopulation. The subpopulation with congestive heart failure due to systolic dysfunction and the subpopulation with EF >30% after an MI would each experience a nearly equal number of deaths prevented, due to ICDs in the first case and due to n-3 supplements in the second case.
The range of mortality outcomes from the sensitivity analysis for each scenario is presented in Table 4 and noted above. The range of change in total mortality for each of the key scenario parameters is shown in Table 1. The parameter showing the largest range of possible outcomes-death reduction from 1.6% to 8.3%-is the relative risk of out-of-hospital cardiac arrest for healthy individuals if taking n-3 supplements. The next largest range, from 1.8% to 4.6% death reduction, is for the relative risk of death for individuals with CHF due to nonischemic cardiomyopathy if an ICD is implanted. All other key parameters have outcome ranges of <1.5%.
This analysis was performed in 2004, 2005, and 2006; the study is a Markov chain analysis.25 Total mortality is used as the main outcome measure for three reasons: many of the trials used in the analysis report survival as the outcome measure, total mortality avoids the ambiguity inherent in the identification of Āgsudden death,Āh and using total mortality eliminates the possibility that an intervention simply changes the cause of death rather than reduces the risk of death. In addition to deaths, the number of times that AEDs and ICDs were used was tallied in the analysis.
The conceptual model used in this analysis to estimate the impact of the interventions on CVD events and total mortality has been reported elsewhere.26 In this model, the probability that an event occurs, an intervention might be useful, or an outcome develops depends on risk factor levels, disease epidemiology, and the level of application of medical technologies in a particular population. While the full model contains eight modules, only four are used in this analysis: healthy individual, congestive heart failure, out-of-hospital cardiac arrest, and acute MI.
An overview of the reduced model is shown in Figure 1. To start the model process, characteristics of the hypothetical population being simulated are set including age, gender, and cardiovascular risk factor distributions. Individuals in the population are defined as initially being apparently healthy or having heart disease (HD). For this analysis, ĀghealthyĀh individuals have not had a cardiac arrest or an MI and do not have congestive heart failure; they may have angina pectoris or occult heart disease. Particular interventions (i.e., n-3 supplementation, ICDs, or AEDs) are active or not active depending on the scenario being simulated. The probability that an acute event occurs in a given year is calculated based on the assigned risk factors and interventions. An acute event can be an out-of-hospital cardiac arrest, the development of a nonischemic cardiomyopathy, an MI, death from heart disease while residing in a nursing home, or death from noncardiac causes. If an individual does not have an event, the individual is aged one year and the cycle is repeated. If an individual has an event, she or he may receive an appropriate intervention-AED for cardiac arrest, ICD when ejection fraction (EF) is ĀÖ30%, or the individual has survived an out-of-hospital cardiac arrest-if it is available in the current scenario. The probability of survival depends on the event experienced and the interventions applied. Individuals who survive are aged 1 year and the cycle is repeated. Individuals who have heart disease at the beginning of the year are routed through the continued heart disease intervention and pass through the acute event node in the same manner as the individual in the healthy population.
Detailed methods including the weighted, directed graph for each of the modules, the sources of the data, the central values, and ranges used in the analysis, and the process used to generate the results are available in an appendix posted on the American Journal of Preventive Medicine website (www.ajpm-online.net).
A program written in SAS, version 9.1 (SAS Institute Inc., Cary NC, 2004) was used to run a simulation for a no-intervention scenario, a scenario for each of the three interventions individually, and a scenario combining all three interventions. For each scenario, the nodes, links between nodes, and the decisions defining the scenario were entered into an Excel spreadsheet. The link probabilities and the ranges used in the sensitivity analysis were entered into the spreadsheet either as numeric values or as references to stored SAS macros. Each scenario required two steps: (1) populating the scenario, and (2) running the scenario. The 100,000-person hypothetical population used for this analysis resembles the population of Olmsted County, MN, aged 30 to 84 in the year 2000. Details of each step are described in the appendix.
Description of the scenarios
The parameters in the model that define each scenario are shown in Table 1. These include a central value and the range used for the sensitivity analysis for each parameter, the source(s) of the value, and the minimum and maximum change in deaths predicted from the sensitivity analysis.
Central value and range for model parameters that define n-3 fatty acid supplement, ICD, and AED scenarios, and associated minimum and maximum outcomes (as % change in deaths from no intervention) from sensitivity analysis
See published article for Table.
While the meta-analysis by Wang et al.24 concludes that increasing n-3 consumption reduces total mortality and sudden death in primary prevention, it does not combine data in a manner that can be used to estimate the potential impact of an intervention. Therefore, values for healthy individuals for the n-3 scenario were drawn from the Physicians' Health Study (PHS).27 Data from the PHS are both consistent with the other data analyzed by Wang et al.,24 and are the only data based on n-3 levels measured in the blood rather than estimated from dietary recall. The n-3 scenario assumes that the healthy population has sudden death rates that are the same as those of PHS subjects in the top quartile of the n-3 blood level distribution. For individuals with a previous MI, the scenario assumes that they are receiving supplemental n-3s and that their risk of sudden death is consistent with rates observed in the GISSI-11 study.28
The AED scenario assumes that AEDs are available in all public areas and homes. It also assumes that having a defibrillator in the home increases the probability of surviving a cardiac arrest at home. Corresponding probabilities were drawn from several community AED studies.14, 15, 18, 19, 20, 21, 22, 23, 29 and 30
The ICD scenario assumes that ICDs are implanted in all individuals with a left ventricular ejection fraction (EF) ĀÖ30% and in individuals who have survived an out-of-hospital cardiac arrest. Survival probabilities and relative risk of death were drawn from a variety of published observations and trials.5, 6, 8, 10, 11, 13, 31 and 32
The scenario combining all three interventions includes the assumptions and rates from all three scenarios. The no-intervention scenario assumes that n-3 levels in the blood are not changed through supplements or diet, ICDs are not implanted in any individuals, and AEDs are not available. Each scenario assumes 100% adherence (although unrealistic) to see what the maximum effect might be.
The no-intervention scenario was run first and used as the baseline against which the remaining four scenarios were compared. Central values for all model parameters were used in each scenario. The change in total deaths, expressed as the percent change from the no-intervention scenario, is the outcome measure used for comparison.
The potential variation in outcomes reported for each scenario is the range resulting from a one-way sensitivity analysis. This range can represent the effect of variations in event rates, the expected effectiveness of the interventions, and the potential uptake of each intervention. The sensitivity analysis was conducted by individually setting all model parameters, including those in Table 1, to their minimum or maximum values and running each scenario for each of these parameter changes. For parameter values taken from single studies, the range is the value's 95% confidence interval. Where multiple studies were used for a value, the range was derived from the extremes observed in the studies. Because some analyses suggest that n-3s do not change outcomes in the healthy population,33 and because adherence to n-3 supplements might be low in any event, the relative risk of sudden death for healthy individuals taking n-3 supplements was expanded to include no effect. To account for the possibility that AEDs are less effective in homes than in public due to a lower witnessing rate, the lower bound of the probability that an out-of-hospital cardiac arrest is witnessed was extended to include the assumption that none are witnessed in homes. The proportion of the initial population with heart disease was also varied from 0.5 to 2.0 times the proportion's central value.