Toxicant deliveries (by machine smoking) are compiled and associated cancer risks are calculated for 13 carcinogens from 26 brands of conventional cigarettes categorized as “regular” (R), “light” (Lt), or “ultralight” (ULt), and for a reference cigarette. Eight “potentially reduced exposure product” (PREP) cigarettes are also considered. Because agency-to-agency differences exist in the cancer slope factor (CSF) values adopted for some carcinogens, two CSF sets were used in the calculations: set I [U.S. Environmental Protection Agency (EPA)–accepted values plus California EPA–accepted values as needed to fill data gaps] and set II (vice versa). The potential effects of human smoking patterns on cigarette deliveries are considered. Acetaldehyde, 1,3-butadiene, and acrylonitrile are associated with the largest calculated cancer risks for all 26 brands of conventional cigarettes. The calculated risks are proportional to the smoking dose z (pack-years). Using CSF set I and z = 1 pack-year (7,300 cigarettes), the calculated brand-average incremental lifetime cancer risk
The enormous health toll taken by tobacco is well documented (e.g., refs. 1, 2). For the known toxicants in tobacco smoke, predictions of the corresponding health effects, as well as the study of corresponding biomarkers, require reliable information about per-cigarette toxicant “yields” (aka “deliveries”). These yields can be expected to differ to some extent among the variety of brands that are available (3). Differences will also occur within a brand over time as caused by changes in cigarette design, additive formulations, crop characteristics, and tobacco blend. Among the important blend variables are amount and type of reconstituted tobacco sheet, essentially a tobacco paper made from stems, leaf fines, and additives. Besides brand family (e.g., Marlboro, Camel, Newport, etc.), cigarette products are distinguished by being (a) conventional in design, or a “potentially reduced exposure product” (PREP); (b) 85 or 100 mm; and (c) mentholated or nonmentholated.
Studies that report the per-cigarette smoke yields (aka deliveries) of a number of carcinogens with known cancer potencies are available (e.g., ref. 4). It is of interest to assemble relevant yield data, then use the best available methods to determine whether the risks of human cancer that can be calculated can account for the epidemiologically observed risks. Also of interest is whether there are meaningful differences in the predicted risks from (a) “regular” (R; aka full flavor) cigarettes; (b) “light” (Lt) and “ultralight” (ULt) cigarettes; and (c) new PREP cigarettes that are being designed by the industry (5). The assembled yield data will also be useful in future biomarker research.
A range of PREP marketing claims have appeared: (a) Advance: “All of the taste…Less of the toxins.” (6); (b) Eclipse: “May present less risk of cancer associated with smoking” and “70% lower smoking-related mutagenicity…” (7); and (c) Omni: “The three groups of carcinogens that have been significantly reduced are polycyclic aromatic hydrocarbons (PAHs), tobacco specific nitrosamines (TSNAs) and catechols. PAHs, TSNAs, and catechols are among the most potent and dangerous substances in tobacco smoke in relation to lung cancer incidence.” (8). It has been argued that marketing of PREPs is likely to have negative effects on public health, including promoting smoking among existing and new smokers and impeding cessation among existing smokers (9). If the perceptions of reduced risk are then also ultimately determined to be unfounded for any given PREP brand, the consequences associated with the availability of that brand would be doubly negative. Alternatively, if some PREP brands do offer some reduction of risk for a given number of cigarettes, it will be important to compare the magnitude of that benefit against the costs of any increase in smoking and/or decrease in cessation.
The public health concerns about PREPs are due in significant measure to the history of Lt and ULt cigarettes. Indeed, whereas such cigarettes have been marketed using wording intended to imply that they are safer and more protective of health than R cigarettes (e.g., see ref. 10), Burns et al. (11) report that epidemiologic data do not show that Lt and ULt cigarettes offer lower levels of risk as compared with R cigarettes. A recent U.S. court ruling states that “low tar/light cigarettes offer no clear health benefit over regular cigarettes” (12). The absence of any measurable benefit from smoking such cigarettes is due at least in part to the fact that consumers tend to smoke these cigarettes more intensely than R cigarettes to obtain certain nicotine deliveries (13, 14).
Epidemiologic studies that retrospectively examine disease incidence data for a given population can provide the final, quantitative measures of the risks of a particular per-person toxicant dose against which all other such estimates must be compared and reconciled (15). For smoking, examples of such epidemiologic studies include those of Doll and Peto (16), Villenueve and Mao (17), Holowaty et al. (18), and Bach et al. (19). However, because many smoking-related diseases take years to develop, discerning the effects of cigarette types/designs on disease incidence by means of such studies will neither be rapid nor easy. It is therefore useful to begin to address smoking risk by application of risk assessment methods that seek to predict risk based on the time-averaged doses of the carcinogens for a given toxicant portfolio. By this means, one can seek to move toward a more predictive, toxicant-by-toxicant understanding of the links between cigarette smoking and health outcomes.
Limitations and Utility of the Risk Assessment Approach as Applied to Smoking
For a given carcinogen i, risk assessment methods seek to compute the incremental (i.e., above-baseline) lifetime cancer risk ILCRi to an individual that is associated with an assumed exposure to the carcinogen. ILCRi is a fractional quantity (e.g., 0.0002). The standard approach is to compute each ILCRi based on the chronic daily intake CDIi (mg/kg-d) of i as averaged over some assumed lifetime (e.g., 70 years; ref. 20):
where CSFi [(mg/kg-d)−1] is the cancer slope factor (CSF; aka “cancer potency”) for i. To the extent that risk (i.e., chance) has no units, ILCRi may be thought of as being dimensionless. (Note that the units of CDIi and CSFi are designed to “cancel.”)
Any given sample of tobacco smoke is a very complex mixture. When there is simultaneous exposure to a specific known set of carcinogens (and especially when the carcinogens share the same mode of action), prediction of the total risk has often proceeded by a simple additivity model according to (21, 22)
Unfortunately, there is limited knowledge about how multiple carcinogens actually affect organisms (e.g., the extent to which the carcinogens act independently or interact synergistically; ref. 23). In addition, in the case of tobacco smoke, irritation from the smoke can lead to increased cell proliferation and thus increased likelihood of tumor development (24). Nevertheless, at present, there is no toxicant-specific risk assessment model for complex exposures that is clearly superior to Eq. B for this application.
The total number of pack-years smoked (represented here as z) is one common measure of the total smoking exposure. (1 pack-year = 365 packs = 7,300 cigarettes.) z is the measure directly related to computation of CDIi values. Assuming a specific lifetime (e.g., 70 years), a given carcinogen i, and constancy in the characteristics of the cigarettes smoked, then CDIi would be proportional to z: smoking 50 pack-years will give a CDIi that is fifty times that from 1 pack-year. By Eq. A, the cancer risk due to carcinogen i then also increases by fifty times, as does the total risk due to all carcinogens (Eq. B).
Application of Eqs. A and B to smoking assumes that variations in exposure intensity (packs/d) and duration (years of smoking) do not separately affect the risk: only their product as it appears in z is relevant. For an individual, however, it is well known that the risk of smoking-related cancer (e.g., lung cancer) is not a simple linear function of z (16, 19). Smoking-related lung cancer is thus affected not only by z but also separately by the smoking rate and the years of smoking, and indeed other variables such as age of smoking onset (16). The complexity in the dose-response relationship for smoking has motivated empirical, case-control–based searches for regression equations/models for predicting smoking risk (e.g., refs. 19, 25) and for new metrics for smoking exposure. Besides pack-years, metrics considered for lung cancer include (pack-years)1/2 and logcig-years (= [loge(cigarettes per day + 1)] × years; refs. 26, 27). The latter metric provides a way to separately consider the effects of smoking intensity and smoking duration (26, 27). However, even as there is a continuation of effort to improve the prediction of smoking risks by purely empirical means, interest remains exceedingly strong in considering the potential consequences of specific carcinogens in the smoke generated by both conventional cigarettes and PREPs (5). For that, at present, one must rely on Eqs. A and B, despite their simplistic reliance on z as the dose metric and the assumption of simple additivity. Eqs. A and B are the basis of an important check on the status of our understanding of the carcinogenicity of the smoke from conventional cigarettes, as well as the starting point for the examination of any implicit or explicit marketing claims of reduced harm from PREPs.
Prior applications of risk assessment principles to smoking risk have been reported (28, 29). Vorhees and Dodson (28) used yield data for smoke constituents from multiple brands of conventional cigarettes to calculate cancer risks due to 30 smoke carcinogens but did not include comparisons among brands within a given cigarette type or among cigarette types. Fowles and Dybing (29) surveyed multiple sources of cigarette yield data, then computed cancer risk indices for 41 smoke carcinogens (as well as health effects for 17 noncarcinogens), but solely for an estimate of the average conventional cigarette.
Here, according to the logic outlined in Table 1, we undertake calculations of cancer risks by brand and by cigarette type using averages for each of four main cigarette types of interest (R, Lt, ULt, and PREP). We also consider the results as compared with epidemiologically observed risks of lung cancer because that form of cancer is most clearly linked to smoking.
Materials and Methods
Multiple terms and symbols are used in the discussion of ILCRi values for individual carcinogens i as a function of the smoking dose z (e.g., 1 pack-year, 5 pack-years, etc.). For example, in consideration of Eq. B, when summing the risks from multiple carcinogens measured in tobacco smoke, notation is required that emphasizes that any such summation is a partial sum, denoted here as subΣ. (The enormous complexity of tobacco smoke means that it is likely that it will never be possible to know the levels and cancer potencies of all of the individual carcinogens in any sample of tobacco smoke.) Table 2 summarizes the terms that provide the focus of the final discussion; the definitions of other terms and abbreviations are given only where first introduced.
Effects of Machine versus Human Smoking on Cigarette Yield (Delivery) Values
Numerous smoker-to-smoker differences exist in the smoking process, and indeed from day to day for a given smoker. The differences are both idiosyncratic and stochastic in nature, and even a given smoker is subject to idiosyncratic differences because smoking habits tend to evolve with the passage of time. As a result, there is no single “human smoking condition,” and both “smoking topography” and degree of vent-hole blocking can vary widely. (“Smoking topography” = smoke flow rate versus time.) For particular cigarettes of interest, efforts to obtain cigarette yield data under human smoking conditions can therefore only capture yield information applicable to the subject smokers for the days (and hours of the day) of the testing.
Given the logistical difficulties, nonuniformities, and uncertainties associated with cigarette yield data obtained under human smoking conditions, considerably more cigarette yield data have been acquired using standardized machine smoking protocols than using human-derived smoking conditions. The yield values used here were obtained by machine smoking, mostly by the Massachusetts (MA) protocol. The possible magnitudes of typical differences caused by human smoking patterns versus standard machine smoking protocols may be considered as follows.
Yield values can be expected to depend primarily on the total smoke volume Vs,tot (mL, per cigarette), although variation in puff size over the course of the cigarette (e.g., constant puff size throughout versus larger initial puffs followed by smaller puffs) can also play a role. The Federal Trade Commission (FTC) machine smoking protocol (30) is characterized as (a) sequential 35-mL, 2-s puffs every 60 s; (b) no blocking of filter ventilation holes; and (c) smoking to a “butt length” equal to either the filter overwrap (“tipping”) paper + 3 mm, or the filter length + 8 mm, whichever is greater (although in any case not less than 23 mm). The number of puffs per cigarette varies by brand, moisture content, etc., but ∼8 puffs is typical, giving
In a review of human puffing patterns (31) with conventional cigarettes, the average puff volume is 43 mL and the average number of puffs is 11, suggesting Vs,tot ≈ 470 mL. In a study with 133 smokers, Djordjevic et al. (3) observed an average Vs,tot value of ∼550 mL for R cigarettes, and ∼640 mL for “low nicotine” cigarettes (Lt, etc.); Lee et al. (32) measured Vs,tot ≈ 630 mL for human subjects smoking their own brand. These human Vs,tot values are roughly 2 ×
The MA machine-smoking protocol was developed to address the view that cigarettes are typically smoked more intensely than described by the FTC protocol. The MA protocol (4) is (a) 50% blocking of filter ventilation holes; (b) sequential 45-mL, 2-s puffs every 30 s; and (c) the same butt length criterion as in the FTC protocol. As with the FTC protocol, the number of puffs per cigarette varies, but ∼9 puffs is typical, giving
Detailed studies have not been carried out on the effect of variable puff size on yield values for the range of toxicants of interest here. In the absence of such information, according to the above observations for the FTC and MA protocols [in particular, the results of Roemer et al. (33)], Vs,tot was judged to be a reasonable first-approximation scaling factor for yield values for different smoking topographies, including human topographies. Considering the results of Djordjevic et al. (3) and Lee et al. (32) discussed above, typical human Vs,tot values might be approximated as
Given that PREPs are designed with the general intent of reducing constituents that are characteristic of tobacco smoke, human smoking of such products may tend to be significantly more intense than with conventional cigarettes: observations of human smoking of the Eclipse PREP gave Vs,tot ≈ 1,350 mL (32). According to this view, yield values obtained for PREPs by machine smoking with Vs,tot = 300 to 400 mL may require upward adjustment by more than 2×, which would tend to increase the calculated toxicity of PREPs.
CDI and Smoking Dose z
where CDIi (mg/kg-d) and Ai (μg/cigarette) are defined in Table 2;
A given smoking dose z (pack-years) can be achieved in any number of ways. For z = 1 pack-year, one pack (20 cigarettes) can be smoked per day for 1 year, or 1 cigarette can be smoked per day for 20 years. The total quantity of cigarettes smoked is given by
Substitution into Eq. A gives as a function of z:
so that, when z = 1 pack-year
For carcinogen i, the ILCR for a smoking dose z = 1 pack-year is calculated according to
As a function of z
, we havewhere CSFi is the CSF (cancer potency) for carcinogen i with units commonly given as (mg/kg-d)−1, or, more precisely, risk/(mg/kg d); see Table 2.
By Eq. E,
Whereas a linear risk model as embodied in Eq. J is a standard approach in estimations of risk, it is important to reemphasize that the available epidemiologic evidence (e.g., refs. 16, 19, 26, 27) suggests that a linear, z-only model does not exactly apply to smoking related disease. This matter is addressed further below.
SubΣ Cancer Risks (ILCRsubΣ)
In a specific application of Eq. B, for human lung carcinogens, we define
where the right hand side is the sum over all human lung carcinogens for which cigarette yield and CSF data are available (in this work, formaldehyde, acrylonitrile, arsenic, and cadmium).
As noted above, the use of subΣ emphasizes that any attempt to sum ILCR values for tobacco smoke will yield a subtotal of the risk terms, not the full total. Nevertheless, the determined study of the composition and toxicology of tobacco smoke (including mixture, nonlinear, cancer-promoter, and irritation effects) may eventually allow predicted values of
In a second application of Eq. B, we define the subΣ for all carcinogens for which cigarette yield and toxicity data are available, regardless of cancer end point or animal model. For example, 4-(N′-nitrosomethylamino)-1-(3-pyridyl)-1-butanone (NNK) and benzo[a]pyrene are known rodent lung carcinogens. Therefore,
Cigarette Yield (Delivery) Data
Supplementary Table S1A to D contains the cigarette yield data considered here. The data pertain to multiple toxicants for each of 26 brands of conventional commercial cigarettes, one conventional reference cigarette (the 1R4F, a Lt cigarette), four PREP prototypes, and four PREP versions that have been commercially marketed. For the 26 R, Lt, and ULt brands and the 1R4F, all data were obtained from the 1999 Massachusetts Benchmark Study (4), which reported average yield values (five replicates) for 41 smoke constituents, plus tar, carbon monoxide, and nicotine. Only the carcinogen data are used here.
The Massachusetts Benchmark Study (4) used the MA machine-smoking protocol (see above). The cigarettes were categorized here according to tar yields that have been reported (30) using the FTC smoking protocol. In this work, three bins by tar yield are defined: (a) R, ≥15 mg; (b) Lt, 6 to <15 mg; and (c) ULt, 1 to <6 mg. (Whereas the descriptor “medium” has also been used with commercial conventional cigarettes, three bins were considered adequate.) The bin assignments made here do not in all cases equate with the designations used by the manufacturers. The data for the 1R4F were excluded when calculating means and SDs for ILCR values for the Lt category.
Some PREP cigarette yield data are available (6, 34-43), including data found in several once-secret tobacco industry documents (40-43). The PREP yield data used here were obtained as follows: TOB-HT prototype (34); “electrically heated cigarette” prototype (35); XDU 2-104 prototype (40); XDU 740 prototype (40); Premier (40); Advance “Light” Kings and “Light” 100s (6); and Eclipse (36). Only various subsets of the toxicants measured for the conventional cigarettes considered were measured for the PREPs in the studies considered. In addition, a number of the PREP studies contained only incomplete information on (a) the number of samples/replicates and (b) the smoking protocol used [e.g., Borgerding et al. (40) do not provide details on the smoking protocol used for the Premier and XDU prototypes]. No adjustments to the PREP yields were made here to compensate for such differences in smoking protocol: doing so would not have been straightforward given the incomplete method descriptions found in some of those studies, and, moreover, not likely to introduce much more than a factor of 2 change in the results (see above discussion of effects of smoking topographies).
Values of Body Weight, Averaging Time, and CSFi
Following U.S. EPA (20), for body weight and averaging time, the default values of 70 kg and 70 years × 365 days/y were adopted. ILCR values were calculated using two sets of the CSFi. For set I (see Table 3), values were obtained from the U.S. EPA Integrated Risk Information System data base (44) where available, and if gaps existed in the Integrated Risk Information System database, by using values from the California EPA Office of Environmental Health Hazard Assessment (45, 46). For set II (see Table 4), the converse process was used. The California EPA CSFi values for 1,3-butadiene, acrylonitrile, benzene, and cadmium are 5.5, 4.2, 3.7, and 2.4 times higher, respectively, than the corresponding U.S. EPA values.
For two carcinogens, quinoline and the tobacco-specific nitrosamine NNK, it was necessary to use CSF values based on p.o. administration (45, 47). CSFi values for 11 other carcinogens were obtained by conversion of tabulated values of the inhalation unit risk [IURi, (μg/m3)−1]. Following U.S. EPA (20), those conversions assumed BW = 70 kg and average daily inhalation rate (IR) = 20 m3/d:
For the carcinogens considered here, values of
Based on CSF set I, for 24 of the 26 commercial brands of conventional cigarettes and for the 1R4F, the rank order of the carcinogens for the three highest values is the same: acetaldehyde ≥ 1,3-butadiene ≥ NNK. The two exceptions are Marlboro King filtered Lt soft pack (acetaldehyde > acrylonitrile > 1,3-butadiene = NNK) and Now King filtered soft pack (acetaldehyde > 1,3-butadiene = acrylonitrile). Based on CSF set II, the rank order for the carcinogens with the three highest is the same for all 26 commercial brands of conventional cigarettes and for the 1R4F (i.e., 1,3-butadiene ≥ acrylonitrile ≥ acetaldehyde). For the PREPs considered, for both CSF sets I and II, the rank order for
For CSF set I, for all three types of conventional cigarettes,
Each brand-specific value of
Not all of the carcinogens measured for R, Lt, and ULt cigarettes in the Massachusetts Benchmark Study (4) were determined in the various PREP studies. The subΣ risk values presented in Fig. 2D therefore pertain to various more limited subΣ values than could be computed for the other cigarette types. Nevertheless, each of the “all carcinogen” subΣ risk values for the PREPs was found to be >10−6, whether by CSF set I or II. Moreover, just as compensation has been observed with Lt and ULt cigarettes, PREPs may tend to be smoked more intensely than conventional cigarettes. For example, as noted, Lee et al. (32) observed total smoke volumes (Vs,tot) of ∼1,350 mL for smokers with the Eclipse as compared with the ∼405 mL/cigarette expected from the MA protocol. A factor of 3 × increase in deliveries due to intense smoking for PREPs would bring
Measures of the Observed Average Lung Cancer Risk
Placing the per pack-year cancer risk values calculated here into proper context requires an understanding of the actual cancer risks indicated by available epidemiologic data. This can be undertaken most easily with lung cancer given that ∼90% of male lung cancer deaths and 75% to 80% of female lung cancer deaths each year are due to smoking (48, 51).
Lifetime probabilities of lung cancer for Canadian current smokers and “never smokers” were studied by Villeneuve and Mao (17) for 1987-1988. Their results suggest that the lifetime (0-85+ years) smoking-related risk of lung cancer for “current smokers” is ∼0.16 for men and ∼0.11 for women. Because the current male/female ratio for lung cancer incidence rate in Canada is ∼60:40 (52), these values imply an average risk of ∼0.14 for all current smokers in that work. Data for 1985-1991 indicate a population in Canada for ages 15+ years of ∼20,000,000 (53), a smoking prevalence of ∼30% for that group (54), and thus ∼6,000,000 smokers. Sales of manufactured cigarettes in Canada over the preceding decade averaged ∼55 billion cigarettes per year (55). Adding in another ∼16% due to a “roll your own” component (56), the result is ∼64 billion cigarettes per year, or ∼9,000,000 pack-years each year. This indicates an average cigarette consumption of ∼1.5 pack-year/y for all smokers (11,000 cigarettes/y); it is assumed that the smoking rate (among smokers) was roughly similar in preceding years. Assuming that a typical smoking duration is 30 to 50 years, the result is ∼45 to 75 pack-years. Pankow et al.4
J.F. Pankow, K.H. Watanabe, D.F. Austin. Lung cancer incidence rate as related to cigarette consumption rate for 61 nations: the revolver model of lung cancer risk. Submitted for publication, 2006.
Percent Match between Predicted and Observed Risks
If it is assumed that the carcinogen Ai values considered here for conventional cigarettes are representative of historical Ai values for cigarettes from the preceding decades, then one can examine the degree of match between the predicted and observed risks of lung cancer. By none of the methods used below to consider this match, however, do the predicted risks come close to matching the observed risks.
First, the match M (%) between the subΣ risk for human lung carcinogens from a 1 pack-year dose of conventional (conv) cigarettes and the actual observed population-average lung cancer risk
Based on CSF set I, for R, Lt, and ULt conventional cigarettes,
Another definition of the risk match for lung cancer is obtained as follows.
is defined as a z-dependent definition of the risk match. Because of the “concave-up” nature of the nonlinearity in the dependence of lung-cancer risk on z (16, 19), much of the risk for a population that is incorporated in
K.H. Watanabe, J.F. Pankow, S.D. Stellman, D.F. Austin, unpublished data, 2006.
Overall, even allowing for some differences between machine and representative human smoking conditions (see above), the match values obtained here indicate that the subΣ−lung estimates of cancer risk from conventional cigarettes that can currently be calculated are either (a) too low (e.g., as due to incorrect CSF values) or (b) quite incomplete due to incomplete consideration of the roles of important carcinogens, cancer promoters, and/or irritant chemicals that promote cell proliferation as described by Preston-Martin et al. (24). These results are consistent with the conclusion of Fowles and Dybing (29) that the total estimatable cancer risk from the average conventional cigarette is much less than the observed smoking-related mortality risk for all cancers.
Risk Reduction Offered by a PREP
For a given PREP, RPREP (%) is defined as the amount by which the calculatable per pack-year risk of lung cancer is reduced by switching from a particular type of conventional cigarette to that PREP:
The reduction in the risk can be no greater than the risk that can be accounted for with the type of conventional cigarette of interest, hence RPREP ≤ M, as noted in Eq. P. The largest RPREP (and M) values will result when
By analogy with Eq. O, we also define
Based on the results of Watanabe et al.,5 for z = 10, 15, and 20 pack-years, we obtain all R′PREP ≤ 4.5% (CSF set I) and ≤9% (CSF set II).
Overall, the RPREP and R′PREP results indicate that there is not sufficient justification for viewing any PREP versions considered here (including the currently available Eclipse) as providing predicted reductions in the risk of human lung cancer that are significant relative to the observed risk for cigarettes as they are smoked by populations and quantified by
The risk assessment framework provides a useful means to begin the work of considering the toxicant-specific aspects of the cancer risks of smoking cigarettes. This framework allows connections to be made between the levels of carcinogens in cigarette smoke and the observed health risks of smoking. However, for conventional cigarettes (i.e., R, Lt, and ULt cigarettes), as they are smoked by populations such as the United States and Canada, the lung carcinogen results obtained here indicate that, currently, it is only possible to account for ≤4% of the observed per pack-year risk for lung cancer. This is based on an estimate of the observed average per pack-year risk (
Needed improvements in the toxicant-specific risk assessment modeling of tobacco smoke include (a) more analytic information on the identities and levels of the myriad toxicants present in cigarette smoke by cigarette type and brand; (b) more complex risk models that allow consideration of the nonlinearity of the dose response, effects of carcinogen mixtures, and the roles of cancer promoters; and (c) high-quality toxicity data for additional toxicants.
The current inability to use toxicant-specific methods to account for the observed cancer risks of smoking carries an important implication for PREP cigarettes. Namely, all expressed and implied promises of “reduced harm” that now accompany the marketing of PREPs (including ostensible PREPs) must be viewed as speculative and unverified. Indeed, because dose considerations for known tobacco smoke lung carcinogens account for ≤4% of the lung cancer risk of conventional cigarettes as they are smoked by North American populations, then lowered levels of these toxicants in PREPs still leave PREPs in the possible position of being as harmful as conventional cigarettes. Expressed another way, even if a PREP design were to succeed in removing all currently measured known human lung carcinogens from cigarette smoke (and even perhaps all other currently measured carcinogens), there would be little reason to be confident that such removal would by itself lead to any observable reduction in smoking related lung cancer.
Grant support: Funds provided by Michael J. Dowd, Regina M. Dowd, Patrick J. Coughlin, the Cooley Family Fund for Critical Research of the Oregon Community Foundation, the Medical Research Foundation of Oregon, and the Oregon Opportunity Program of Oregon Health and Science University.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this article are available at Cancer Epidemiology Biomarkers and Prevention Online (http://cebp.aacrjournals.org/).