The laboratory aspects of Project MIX studies appear to have been conducted using well-accepted laboratory practices, so the raw data are probably unbiased. The essential conclusion that the PM investigators reached, on the basis of these data, that “The studies with ingredients added to cigarettes did not demonstrate any meaningful effect of the ingredients on the toxicity of cigarettes” 
, however, is a reflection of the way that the data were normalized in the studies of toxins in the smoke 
and low power of the animal toxicology studies 
. Despite these problems, Project MIX results have been widely promoted to the scientific community 
, the public at large, and the Institute of Medicine 
, as well as cited by other tobacco industry scientists 
The conclusion that the additives did not increase the yield of toxins in the smoke is a direct result of the fact that the cigarettes containing the additives produced 15%–28% more TPM than the control tobacco-only cigarettes. The fact that the additives lead to more TPM is, itself, an important indicator of increased toxicity because the TPM in cigarette smoke leads to substantial increases in risk of cardiovascular disease, with a steep and highly nonlinear dose-response at low levels of exposure 
. TPM exposure also disrupts physiological angiogenesis and contributes to ectopic pregnancy, spontaneous abortion, preterm delivery, sudden infant death syndrome, and slower wound healing 
In a letter to the editor of Food and Chemical Toxicology
, Vleeming et al. 
also commented on the fact that normalizing toxin production in smoke gave a misleading picture of the effects of the additives on smoke toxicity. They suggested that toxin levels should be normalized by the amount of tobacco in each cigarette. Since PM reduced the amount of tobacco when including the additives to maintain the mass of the cigarette constant, presenting the results normalized by tobacco weight led to even greater increases in the estimated effects of the toxins than we present (in our –). The PM researchers justified their use of TPM normalization on the grounds that “We choose TPM as our basis of comparison to be consistent with the animal and in vitro studies presented in this series of publications” and because “Since consumers choose to smoke cigarettes according to the tar delivery and taste, we chose to normalize the data to the TPM yield to reveal the effect of ingredients which contribute to the taste” 
. The first of these arguments is surprising since, as noted above, the original design of Project MIX did not anticipate reporting and analyzing results normalized by TPM 
and only introduced this normalization after PM had results showing that the additives led to higher levels of TPM (even though there was correspondingly less tobacco in the cigarettes). Also, as noted above, it is well-established that smokers smoke to control the delivery of nicotine, not tar (TPM) 
and normalizing toxin deliveries by nicotine delivery yields higher toxicity estimates than normalizing by TPM (–; ).
The fact that the in vivo toxicology studies 
were conducted at matched levels of TPM for the smoke from all cigarettes (to hold TPM constant) also meant, as PM scientist Reininghaus noted internally in 1998 
, that the rats breathing the smoke from the cigarettes with the additives were exposed to lower levels of toxins in the smoke than if the exposures had been matched on another smoke variable, such as nicotine delivery. To the extent that the mass of tobacco in the cigarette determines nicotine delivery, Vleeming et al.'s 
approach suggests that the underestimates of toxin exposures the rats received compared to human smokers could be substantial.
Another problem with the in vivo toxicology studies is that PM exposed the animals to fixed levels of TPM and the ratio of TPM to gas phase toxins changes with the different ingredient groups. If the production of gas phase toxins increases less than the increase in TPM, then exposing the rats to a fixed amount of TPM will reduce exposure to the gaseous constituents. For physiological endpoints affected by these gases the toxicity might appear to decrease on the basis of per TPM versus per cigarette exposures. Indeed, when PM instructed INBIFO to analyze the smoke chemistry data on a per TPM basis 
they noted that such an adjustment was already “compensated for by dilution in the subchronic study” 
The way the animal toxicology studies were designed with the relatively short (90 d) exposure period and follow-up after the end of the exposure (42 d) also raises concerns. PM's comparison of secondhand smoke with diesel exhaust (also conducted at INBIFO) had 99 rats of each gender in each exposure group, exposed the rats for 24 mo (730 d) with a 6-mo (183 d) follow-up 
. Longer exposure and follow-up would have increased the power of Project MIX to detect toxic effects of the additives. Even with these downward biases due to experimental design, however, our results suggest that an adequately powered design would have revealed a large number of toxic effects on the rats.
PM conducted extensive research on individual additives; the industry documents contain a list of 170 projects 
conducted between 1973 and 1995 that includes many single additive projects. These single additive projects include INBIFO Project Juice, in which citric acid (one of the 333 ingredients in the present study) was added in low (2.6%), medium (3.9%), and high (6.3%) quantities to cigarette filler 
. The amounts of citric acid reported for Project MIX ingredient group 1 were low level
44 ppm, and high level
122 ppm. It is not clear how these amounts of citric acid compare across studies. Studies were also conducted with lactic acid (Project Milk 
), propylene glycol 
, and cocoa 
. These studies were also conducted using low ratios between groups, along the order of only a 2- to 3-fold difference.
Project MIX also included in vitro genotoxicity and cytotoxicity studies 
using a standard test panel PM developed in 1996, including the Ames test 
. Not surprisingly, the results of these tests found that all of the cigarettes, whether or not they included additives, were genotoxic and mutagenic. These tests, however, are screening tests, not sensitive measures of dose-response 
. Therefore they are not appropriate for quantifying changes in toxicity associated with the additives. Absent the unlikely situation that the additives would eliminate tobacco smoke's genotoxicity and mutagenicity, the failure to find increased toxicity associated with the additives does not support the conclusion that “in vitro mutagenicity and cytotoxicity of the cigarette smoke were not increased by the addition of the ingredients” 
Scientists at British American Tobacco (BAT) conducted a similar project that in 2004 was also published in Food and Chemical Toxicology 
. The BAT studies assessed 482 additives in various mixtures “at or above their typical maximum levels used on cigarettes sold by BAT” through an analysis of 45 selected smoke constituents in mainstream smoke of test cigarettes; in vitro mutagenicity and cytotoxicity tests (Ames test, mammalian cell micronucleus test, and neutral red uptake test) and 90-d inhalation studies in male and female Sprague-Dawley rats (ten rats per group). Using comparable rates of exposure of rats to smoke (approximately 20% less than the PM study rats), the BAT scientists report similar deleterious histopathological findings as Project MIX: increased density of goblet cells in respiratory epithelium, increased hyperplasia/degeneration of the nasal cavity, and increased epiglottal metaplasia. Similar to the PM scientists, the BAT scientists downplayed the findings by concluding that they were inconsistent, indicating that the response of rats exposed to smoke was indistinguishable between test and control cigarettes.
The kind of manipulation of the presentation of scientific results demonstrated by the publication of the Project MIX results 
, is nothing new for the tobacco industry; industry researchers have a long history of doing so around a variety of issues related to secondhand smoke 
. While the procedures to collect the data themselves appear sound, the way that the data were analyzed and interpreted is not. An important implication of the analysis we present is that the scientific community and regulatory authorities cannot take the conclusions in tobacco industry (or industry-funded) research or research published in industry-dominated journals such as Food and Chemical Toxicology
at face value. Vigorous implementation of FCTC article 5.3, which seeks to protect the policy making process against tobacco industry interference and manipulation 
, underscores the need for such skepticism in considering research such as Project MIX (and the corresponding papers from BAT) at face value in the rulemaking process. It will be important for the US FDA, WHO, and regulatory agencies in other countries who are working to implement FCTC articles 9–11 to insist on receiving all drafts of the study protocol (taking particular care to not allow the tobacco companies to use lawyer involvement in the process as a way to avoid disclosure) together with the raw data to reduce the likelihood of the problems identified in this paper.
If one accepts PM's assertion that Project MIX evaluated additives in groups that “resemble typical commercial blended cigarettes” 
, the data PM collected could be used for policy making regarding the use of these 333 additives. That the Project MIX scientists examined cigarette additives in combination (rather than singly) allows for the possibility that the additives act either synergistically or antagonistically. Allowing for this synergism is important. A comparison of the estimated lung cancer effect of tobacco smoke produced by summing the individual effects of constituents of tobacco smoke produced an estimate of cancer risk much smaller than the observed epidemiological risk 
. Probably this is because all the carcinogens and cardiac toxins in tobacco smoke have not been identified and because there are likely interactions between the different constituents in the complex mixture that cigarette smoke represents.
It is also important to emphasize that the definition of “harm” implicit in PM's Project MIX and BAT's similar experiments 
—direct short-term toxic effects of additives—is just one dimension of the effects that additives have on disease. Additives also have an effect on product palatability and pharmacologic properties (Box 1
) in ways that make cigarettes less harsh and more pleasant to the user 
, leading to higher initiation and lower cessation and, so, more tobacco use and tobacco-induced disease 
. Indeed, it was primarily on the basis of these population-level impacts that the FDA Tobacco Products Scientific Advisory Committee concluded that, “Removal of menthol cigarettes from the marketplace would benefit public health in the United States” 
As noted in the Methods
, this study is based on the documents that PM made available as a result of litigation against the company. While we were able to find several drafts and detailed results of the Project MIX studies, that, for example, allowed us to determine that the protocol for analyzing the data was altered after the results indicating increased TPM were obtained, there are key missing pieces of information: why PM chose the additives that they did for study, as well as the levels and combinations and the reason for selecting the sample sizes in the toxicology studies. This situation represents a limitation of the available sources of information and highlights the importance of the FDA using its new authority to obtain all sources of data on additive testing and other features of product manufacturing and testing and to make them fully available for review.
The analysis in this paper shows that many of the toxins in cigarette smoke increase substantially when additives are put in cigarettes, including the level of TPM, and that, assuming that the toxicological results from Project MIX represent unbiased estimates of the true biological effects, these data show many adverse biological consequences (and that the failure to reach statistical significance was the result of underpowered studies rather than lack of an effect). In particular, regulatory authorities, including the FDA and similar agencies elsewhere who are implementing FCTC articles 9–11, could use the Project MIX data to eliminate the use of these 333 additives (including menthol, which is the major component of ingredient group 3) in cigarettes.
Any tobacco company would, of course, remain free to submit an application to the FDA, or other regulatory agency, to reintroduce use of an additive if they could provide convincing data from adequately powered studies that the additive truly did not have any adverse health consequences.