How much pleasure does smoking bring?

News of a controversial cost:benefit calculation contained in new federal tobacco regulations subjecting many tobacco products to cigarette-like regulations, and creating new graphic warning label (GWL) regulations for cigarettes. The question at hand is whether the cost:benefit analysis underpinning the regulations is correct.

Most attention has focused on estimates of the benefits (pleasure, etc.) that smokers derive from smoking that were used in the analysis, and which increase the cost of the regulation (lost benefit to smokers = bigger cost of regulation). Frank Chaloupka and several other leading researchers in the economics of smoking have written a useful critique of the economic analysis undertaken to evaluate this rule. They focus on the issue of “lost consumer surplus “–or the pleasure/other benefits that smokers derive from smoking as a cost of the GWL regulation. From page 3:

The most critical concern about FDA’s cost estimation is the agency’s reliance on lost consumer surplus as a cost of smokers’ quitting in response to the GWLs.  We describe in detail why the notion of consumer surplus, predicated on well‐informed rational behavior, does not apply in this instance in which the vast majority of smokers begin smoking, and become addicted, before the age of majority.

This is important because any rule that has an expected cost of over $100 Million (in 1995$) has to undergo a detailed cost:benefit analysis to demonstrate that the benefits of the regulation outweigh the costs. As the estimate of “lost consumer surplus” rises, the net benefit of the regulation decreases, making the case for its promulgation less clear. A book that I co-authored “The Price of Smoking” (5 part series on the book from June, 2011) with Duke colleagues figures heavily in the FDAs cost:benefit analysis, and our top line findings illustrate the stakes. We estimated the net present value of the societal cost of a pack of cigarettes in 2000$ to be $40/pack, allocated as follows:

  • $33 private cost: borne by the individual, primarily through a substantially shortened lifespan
  • $5.50 quasi-external cost: borne by the smokers’ family through increased health costs, slightly lower wages and other factors
  • $1.50 external cost: borne by society, and representing the net effect of things like taxes paid, Medicaid and Medicare payments, and Social Security received

Most of the cost of smoking is borne by smokers via shortened lifespan, so netting out the “lost consumer surplus” or pleasure from smoking greatly changes the calculus of assessing the costs and benefits of a regulation whose predicted impact is smoking cessation. Note that the disagreement about the magnitude of this “lost consumer surplus” is mostly one of theory application (I think). Two polar opposite interpretations of the economic and epidemiological literature are possible: smoking is simply another economic decision, and therefore the benefits of smoking must be similar to the costs expressed in terms of lost years of life. At the other end of the spectrum, the addictive nature of the product and juvenile initiation means smoking is irrational and therefore not given to a calculus based on the economic rationality of decisions. Surely the truth lays somewhere in between.

The most unusual aspect of the economic analysis undertaken by the FDA is to reduce, by around half, the benefits of smoking cessation to account for “lost consumer surplus” or foregone benefits of smoking, which greatly increases the cost of the regulation. While some allowance may seem reasonable, Chaloupka and colleagues argue persuasively (see p. 11-12) that we don’t have enough empirical evidence to determine the size of the impact, and most crucially that there is no reason to expect smokers to quit smoking in response to graphic warning labels if they had undertaken smoking in a fully rational manner. This is persuasive to me that reducing the benefits of cessation that are expected to result from the regulation by half is an over-estimate of this effect, resulting in an under-estimate of the net benefits of the regulation. It does not make clear what the correct estimate might be.

There are many other interesting issues raised by the FDA regulation and the Chaloupka et al response related to the economics of smoking that I will address in several posts over the next few weeks.

Cost effectiveness of smoking cessation

I was prepping over the weekend for a meeting Wednesday on the economics of tobacco control and ran across this very clear brief from the British Medical Journal that I thought was worth highlighting. Many of the health benefits of cessation accrue via life extensions, but they point out the morbidity differences by smoking status (measured by self rated health). Precise measurement of the morbidity benefits of cessation is likely important for continued cessation success (the chart below is cross sectional and doesn’t show changes). Communicating the benefits of cessation in different forms and formats maximizes the chance that change will be initiated.

More to the point, the article nicely lays out the cost effectiveness of simple smoking cessation interventions in terms of the cost of a life year saved as compared to common strategies to prevent heart attack.

The paper notes the following caution:

Care should be taken when extrapolating the results of these evaluations, as cost effectiveness estimates are likely to be time and country specific and highly dependent on the healthcare system in question. In a system of fee for service, as in the United States, monetary rewards may be necessary to encourage provision. On the other hand, if patients who stop smoking place a reduced burden on the primary care budget in future years, the incentives to provide such services may be inherent in the system.

While I don’t think I would describe the U.S. health care system (systems?!/non-system?!) as being simply fee for service, that helps to underline their point that precise estimates of the costs and benefits of smoking cessation are needed for each nation and likely sub-population to best target smoking cessation strategies. Smoking cessation is an old problem that remains a top public health priority. The CDC has declared tobacco to be a “winnable battle” and there is much work to do in this area.

Smoking Cessation-III: What is the Correct Counterfactual?

This is the fourth in a series of posts looking more closely at quantifying the impacts of smoking cessation; earlier posts:

There was a great deal of data and analytical work behind the development of the model that allowed us to simulate different smoking counterfactuals to provide estimates of the life extension gained from smoking cessation. This effort does not change the key question: what is the appropriate counterfactual?

Table 5 from the paper lays out the life extensions from the counterfactuals we estimated, all of which could be combined to tell different stories:

Using 35 year old male smokers to illustrate, we identified 6.9* life years gained if he quit smoking at 35 and did not relapse. The comparison we highlighted was:

  • 35 year old smoker continues smoking until his death, on average at age 69.3
  • Compared to the counterfactual expected age at death of 76.2 if he quit at age 35 (76.2 – 69.3 = 6.9 years)

But, is that the correct counterfactual?

  • We also estimated the counterfactual life expectancy of a 35 year old smoker whose smoking status (continuation, cessation, relapse) followed the population-observed averages, with death predicted at age 72.9.

If this is used as the counterfactual, then the life extension due to smoking cessation is 3.6 years (72.9-69.3) years instead of 6.9. Quite a difference. In the paper, we highlighted what is essentially the high water mark of life gains from cessation by comparing the life expectancy of a 35 year old smoker who quits to one who continues to smoke until death. Epidemiologically speaking though, some of those smokers who do not to quit at 35 later do so, so perhaps the counterfactual we highlighted was too optimistic, at least compared to the average behavior of a 35 year old smoker.

The exact same model produced both results, the only difference is the counterfactual chosen, and the story being told. Which is the correct one? It is rare that the most important questions are easy.

*This is life extension unadjusted for smoking misclassifciation.

Donald H. Taylor, Jr, Vic Hasselblad, Jane Henley, Michael J. Thun, Frank A. Sloan. Benefits of Smoking Cessation for Longevity. American Journal of Public Health 2002;92(6):990-996.

II-Smoking Cessation: Selection Effects

This is the third in a series of posts looking more closely at the methods used to quantify the benefits of smoking cessation using this paper as an example; earlier posts:

A RCT of smoking cessation couldn’t and wouldn’t be done, so some sort of observational data must be used to quantify the benefits of smoking cessation. The choice of the Cancer Prevention Study II (CPS-II) presented benefits and costs:

  • Biggest benefit: large number of person-years of follow up (7.2 million for females, 4.3 million for males)
  • Biggest cost: selection effects that its use introduced

By selection, I mean the participants in CPS-II differed from the overall population:

  • They were whiter (93% of CPS-II v. 80% 1990 Census) and they had higher levels of education (30% college degree in CPS-II v. 9% U.S. adults).

What does this mean for our estimates of the relative risk of death by smoking status? First, the direction of any bias due to selection is ambiguous, unlike measurement error in smoking status that provides a predictably conservative bias toward identifying no effect. Using racial differences as an example, two stories seem plausible. First, whites could receive more benefit from cessation because they have a longer background life expectancy than minorities. On the other hand, non whites could receive more benefit from cessation at later ages because of healthy survivor effects. I am unsure what direction the net effect that selection bias of this sort may go.

In the end, 11.5 million person years of follow up covers a multitude of sins, and we were able to control for race (white v. non white) and education in the estimation of the relative risk of mortality calculations that underlie our models. However, even though we had large enough cell counts to control for these variables, there is still worry that the underlying population is different from that used to obtain the relative risk estimates.

Other cohort studies available were even whiter and based on one geographic locale (Framingham), or were based on the experience of General Practitioners in the U.K. (British Doctor Study). It would be great to have a long term follow up database that was representative of the overall population in terms of race and education to update this study, particularly given the increase in the Hispanic population of the U.S. The Health and Retirement Study, which is approaching 20 years of follow up could be an option for updating, and the tradeoff would be smaller cell sizes vs. being more representative. Are there other databases that should be considered?



Smoking Cessation-I: As Compared to What?

I am going to do some follow up posts on the discussion last week of how to use smoking cessation research to answer practical questions about how much longer a person would live if they stopped smoking. The point of the paper on which the post was based was to denominate the benefits of smoking cessation in a format that was readily understandable with a goal of providing information that could help spur smokers to attempt cessation. We felt that life years gained from cessation was about as clear as it could get.

The first question that needs to be answered when seeking to quantify the benefits of smoking cessation is “as compared to what?” The counterfactual is the conceptual key to estimating and communicating the magnitude of any cessation benefits. To provide a quantitative estimate, we needed several things:

  • Long term follow up of a population at risk of smoking. We used the American Cancer Society’s (ACS) Cancer Prevention Study II (CPS-II), a prospective cohort study begun in 1982 that included 1.2 Million persons as of December 31, 1996. There are selection effects introduced by the use of this survey that will be discussed in later posts (more educated, more white, than U.S. population).
  • Data on smoking, collected via self report in the CPS-II.
  • Data on mortality for CPS-II participants, collected via ACS inquiries in 1984, 1986, 1988 and through checks against the National Death Index in December of 1989, 1991, 1994 and 1996 (Dec. 31, 1996 is censor date). Death certificates with multiple causes were collected for 98.6% of deaths.

There are several other key aspects of the study methodology:

  • We excluded CPS-II participants with incomplete data on smoking and/or history of pipe or cigar smoking (excl. 138,669 men; 72,459 women) and persons reporting being sick at study entry because they may have changed smoking status due to illness (34,824 men, 61,522 women).
  • After exclusions, the analysis sample included observations from date of study entry until December 31, 1996 resulting in 7.2 Million person years of female follow up and 4.3 million male person years.
  • We used all cause mortality and not simply death from smoking related diseases.
  • A sub-study conducted in 1992 assessed smoking status changes for a subset of the CPS-II cohort, allowing us to adjust for smoking status changes over time that would otherwise have to be assumed to be constant based on initial assessment when a subject entered the CPS-II cohort.
  • We identified the relative risk of death by age, race and smoking status from CPS-II and applied these to the 1990 Census population to estimate age, sex, race and smoking specific mortality rates. (**more detail below the figure)
  • Finally, we constructed a model that allowed us to simulate how cessation at particular ages would alter mortality and quantify the life extension of smoking cessation as compared to different counterfactuals.

The estimated benefits of smoking cessation for longevity are summarized in the figure; over 8 years for males stopping at age 35 as compared to a 35 year old smoker male who smoked until their death. For females, cessation at age 35 increased expected lifespan by nearly 8 years. Even cessation at age 65 resulted in meaningful life extensions, again as compared to a 65 year old smoker who continued until death. In future posts, I will talk about the selection biases introduced by using CPS-II to obtain the relative risk of death due to smoking, the model simulations used to quantify longevity extensions due to cessation, what all this means for interpreting/using our results and whether there are opportunities and a need to update this research.

**More detail on RR estimation. We estimated the relative risk of death by smoking status using the CPS-II data using Cox proportional hazards model controlling for age (in 1 year increments) and sex. Final models were estimated separately for men and women, and for those older and younger than age 70, controlling for age in 1 year increments within those groupings. Initially, smoking was assumed to remain constant during the follow up period (1982-1996, or death) as measured by the self report in CPS-II.

Assuming static smoking status could result in two types of bias: persons classified as current smokers at baseline who later quit, and former smokers at baseline who relapsed. Both sources of bias would serve to underestimate the positive impacts of smoking cessation on longevity. If current smokers at baseline quit, then the negative impact of smoking would be underestimated. If former smokers relapsed, the the benefits of cessation would be underestimated; both errors bias towards no difference in life span by smoking status. The 1992 follow up study for a subset of CPS-II showed that cessation was far more common (56.8% of male current smokers in 1982 did not smoke in 1992; 52.7% for females). Relapse rates were much lower, on the order of 3%, and never smokers initiating in middle-to-late age were miniscule.

Relative risk of mortality rates obtained from the CPS-II stratified by age, race, sex and smoking status were applied to the 1990 Census population, stratified by age, race and sex. We then divided by a factor representing the weighted average of the category-specific relative risks to obtain estimated mortality rates for each age, race sex and smoking status group. Life extensions were shown with and without an adjustment for smoking cessation (those in graph above are with this adjustment) but no adjustment was made for relapse. Again, this biases findings toward no effect of cessation.

Donald H. Taylor, Jr, Vic Hasselblad, Jane Henley, Michael J. Thun, Frank A. Sloan. Benefits of Smoking Cessation for Longevity. American Journal of Public Health 2002;92(6):990-996.


Dad, Can You Stop Smoking and Not Die?

My 11 year old son posed this question to me last week after a group visited his school and discussed the dangers of smoking. They were of course seeking to discourage smoking initiation. We discussed some of what he retained from this session, some correct, some wrong. I could tell the real question lurked. He then asked me this:

If Pop Pop had stopped smoking would I have met him?

Pop Pop was my father in law, who was a lifelong smoker and who died about 3 months  before my son was born. My wife talks about her father often with the kids (the older ones are 16 and 14), and she usually doesn’t miss the opportunity to reinforce the dangers of smoking cigarettes.

11 year old’s like direct answers, so I told him that if his Pop Pop had stopped smoking before he got sick (around a decade before his death) then he probably would have met him.

A paper I published around a decade ago in the American Journal of Public Health, Benefits of Smoking Cessation for Longevity provides the essence for my answer; Table 5 denominates the benefits of smoking cessation by a given age in terms of life years gained as compared to continuing to smoke:

Note, the expected survival of never smokers should be 78.2 (for men) in the table, as compared to that of 69.3 for those who smoked until death (a typo, Methuselah was not a subject). The benefits of cessation for longevity are notable, and we found benefits even for smokers who stopped at age 65 (contingent upon their not already being sick when they quit, which is actually the primary reason that older smokers quit). The figure below summarizes the life extension gained due to smoking cessation at a given age (35, 45, 55 and 65), contingent upon surviving to that age.

There are a lot of things going on in the model that produced these results; I will amplify them over the next  few days.

Donald H. Taylor, Jr, Vic Hasselblad, Jane Henley, Michael J. Thun, Frank A. Sloan. Benefits of Smoking Cessation for Longevity. American Journal of Public Health 2002;92(6):990-996.