Finally, U.S. Official Admits that Smoking Deaths are Declining

Dr. Phil Cole and I asserted in 2008 that smoking-related deaths in the U.S were on the decline. Finally, this month, an FDA official arrived at the same conclusion in a journal article.

In “Calculating the Big Kill,” which appeared in Regulation, the Cato Institute’s flagship publication (here), I made the case that “CDC [Centers for Disease Control and Prevention] estimates of smoking-related deaths do not add up.”  I noted that “U.S. smoking rates have been declining almost continuously since the mid-1960s, when 42 percent of American adults smoked.  The rate dropped to 28 percent in 1988 and has now declined to 21 percent.  Put simply, the proportion of Americans who smoke has been cut in half over the past 40 years.”

I asked, “Why have the CDC [big-kill] estimates not fallen?”

My question was based on a research study that Dr. Cole and I published a year earlier in Nicotine and Tobacco Research (abstract here), showing that “there were 402,000 deaths attributable to smoking in 1987 and 322,000 in 2002.”  We noted that despite the substantial decline in smoking prevalence, the estimate of smoking-attributable deaths from the Centers for Disease Control and Prevention had changed hardly at all. 

Five years later, my question is still relevant.  The answer is clear: The CDC chose not to acknowledge the fact that deaths had plummeted.

Brian Rostron, a scientist at the FDA, has now published a study, also in Nicotine and Tobacco Research (abstract here), demonstrating that smoking-attributable deaths declined from 398,000 in 2000 to 370,000 in 2007.

As I explained in my Cato article, most people mistakenly believe that CDC officials actually count the number of smokers who die each year.  The truth is that the CDC uses a model that estimates the number of current and former smokers based upon data in the National Health Interview Survey (NHIS); the CDC subjects those smoker estimates to relative risks developed by the American Cancer Society in the 1980s, and then produces an estimate of the number of deaths that would not have occurred if they had never smoked.  Like any model, the quality of the output is dependent on the quality of the input.

I had noted that the CDC was using 20-year-old data to produce its “big-kill” estimates; Dr. Rostron concurs, saying that “the CDC has not substantially revised the methodology or data that are used in this procedure since their introduction in the 1980s.” 

Rather than using decades-old risk estimates, Dr. Rostron updated the relative risks for smokers and former smokers by calculating them from NHIS subjects from 1997 to 2004 who were followed through 2006 with linkage to the National Death Index.  Unlike the outdated CDC estimates, he adjusted risks for age, race/ethnicity, education, alcohol consumption and body mass index, which are important confounding factors for some or all smoking-related diseases.

Here are Dr. Rostron’s big-kill estimates:

Smoking-Attributable Deaths Among Men and Women in the U.S., 2000, 2004 and 2007
Year	Men	Women	All
2000	211,000	187,000	398,000
2004	200,000	180,000	380,000
2007	195,000	175,000	370,000

Although declining, Dr. Rostron’s big-kill numbers may still be too high because of the way he estimated deaths among former smokers.  Using his model, former smokers accounted for about 177,000 out of the 370,000 deaths from smoking in 2007.  This seems to be excessively high, as the 2007 NHIS documents that 60% of the 39 million former smokers had quit over 15 years previously.  Any excess risk for a smoking-attributable death was minimal to nonexistent in this group. 

Dr. Rostron’s deaths among former smokers were likely concentrated in those who quit less than 5 years earlier, and who therefore have risks similar to those of current smokers.  But it is inconceivable that this group of 6.3 million former smokers produced almost as many deaths as the 27.5 million current smokers in 2007.  Dr. Rostron’s model should be revised to reflect the different risks among former smokers who have short or long quitting histories.

Regardless, the main message of this study is valid: Smoking-attributable deaths have been in decline for over a decade.  This is good news that the CDC and other federal agencies have ignored for far too many years.

Overweight and Low Mortality

Last week the Journal of the American Medical Association (JAMA) published a meta-analysis documenting that being “overweight was associated with significantly lower all-cause mortality.” (here).  The study was authored by Dr. Katherine Flegal of the U.S. Centers for Disease Control and Prevention (CDC) and colleagues at the National Cancer Institute and the University of Ottawa. 

 

This post will focus on the study’s scientific findings and the reaction to same, which is not dissimilar to the research and backlash surrounding tobacco harm reduction.

Dr. Flegal and colleagues conducted an extensive search for population studies reporting all-cause mortality according to categories of the body mass index (BMI), which is weight in kilograms divided by height in meters squared.  They used the following standard categories: 

Table 1. BMI Ranges for Weight Categories in the Flegal Meta-Analysis
BMI Range	Weight Category
Under 18.5	Underweight
18.5 to less than 25	Normal
25 to less than 30	Overweight
30 to less than 35	Grade 1 Obesity
35 to less than 40	Grade 2 Obesity
40+	Grade 3 Obesity

 

Dr. Flegal’s literature search initially identified 7,036 articles, of which 97 met the criteria for their study, which included adequate adjustment for age, gender and smoking.  The results were reported as hazard ratios (HRs), which are similar to relative risks.

Table 2. Hazard Ratios (HRs) for All-Cause Mortality According to Weight Categories in the Flegal Meta-Analysis
Weight Category	HR (95% Confidence Interval)
Underweight	Not Reported
Normal Weight	Referent Group
Overweight	0.94 (0.91 – 0.96)
Grade 1 Obesity	0.95 (0.88 – 1.01)
Grade 2-3 Obesity	1.29 (1.18 – 1.41)
All Obesity	1.18 (1.12 – 1.25)

 

 

The major finding is remarkable: Overweight people had a LOWER risk of mortality than people of normal weight.  In addition, Grade 1 obesity (up to BMI < 35) was also associated with lower mortality risk, although this was not statistically significant.  

The results run counter to the dominant public health message that overweight and obesity are killers, which is why the study was immediately denounced by many.  As Christopher Snowdon pointed out (here), extensive media coverage of the attacks gave credence to the notion that the analysis was flawed and unreliable.  Walter Willett, professor of public health at Harvard, declared: “This study is really a pile of rubbish and no one should waste their time reading it.” (here). 

This high-profile assault is similar to that which has been waged against tobacco harm reduction findings over a 20-year period.  In the 1990’s, Philip Cole and I published numerous scientific articles on the gaping difference in health risks associated with smokeless tobacco use and smoking.  We were sometimes ignored and often reviled; ad hominem attacks routinely obscured our important findings.

 

I have long had a professional interest in population studies of weight and health. In 2004, I published the first and only study showing that Swedish men who quit smoking by switching to snus avoided the weight gain usually seen with smoking cessation (abstract here).  My research experience provides some insights into the Flegal/CDC study.

Why are overweight people at less risk of dying than those of normal weight?  It’s possible that the normal-weight group included people who had lost weight because they were ill and were close to death, thereby raising the death rate in the reference group to which all other categories were compared.  In other words, an artificially higher death rate in the normal-weight group resulted in an artificially lower rate among overweights.  Although this is one of the more common criticisms discussed in the media, it is not likely to be a major factor, because the vast majority of the normal-weight population are healthy.

A more likely explanation for Dr. Flegal’s surprising result is that many in her “overweight” population belong in a properly defined “normal” weight category.

In my weight research I had to establish BMI categories.  A literature review revealed that major health organizations had changed the definition of overweight.  Up to the late 1990s, researchers had defined overweight as a BMI of 27+, which meant that normal weight went from 18.5 up to 27.  Dr. Flegal agreed with this definition, as she published a study using this classification in 1994 (here).

 

For reasons that I was never able to pin down, the definition of overweight changed in the late 1990s to a BMI of 25+.  Suddenly, people with a BMI from 25 to 27, who were previously normal weight, were now overweight.  The prevalence of overweight skyrocketed.

Since the newly-classified overweights continued to die at normal-weight rates, this reclassification might help explain Dr. Flegal’s finding of low mortality in the overweight category. 

 

One unresolved problem with Dr. Flegal’s analysis is that she defined three distinct obesity categories (1, 2 and 3 in Table 1), but then combined the mortality results for the highest two categories, and failed to report any results for the underweight category.  These are important omissions which Dr. Flegal needs to explain.

  

Despite these issues, CDC scientists should be commended for publishing this profoundly important finding: Compared with normal-weight individuals, overweight and slightly obese people have LOWER mortality.  It’s a shame that their colleagues at the CDC Office of Smoking and Health aren’t equally forthcoming about the minuscule risks for smokeless tobacco use.  For example, the first entry on the smokeless tobacco fact sheet from the CDC (available here) is - inexplicably – a webpage for gutka, a southeast Asian tobacco product that is rarely used in the U.S.  The rest of the CDC document conflates information regarding the health risks from smokeless tobacco use with those from smoking.

Like tobacco use, overweight and obesity are the objects of revulsion among some health advocates.  Scientific evidence, however, clearly shows that overweight or minimally obese people have lower mortality risks than those of normal weight, placing them at risk only for misguided, unscientific public health interventions like fat taxes (here) or soda restrictions (here).  Like tobacco users, the overweight and obese deserve health policies based on scientific facts, not fanaticism. 

High School Seniors’ Tobacco Use Pales in Comparison to Alcohol

Smoking by teens has declined to record-low levels, according to the latest University of Michigan annual Monitoring the Future Study, which examines youth tobacco, alcohol and drug use in 2012 (data tables here).  The figure at left, which shows the percentage of high school seniors using alcohol, marijuana, cigarettes and smokeless tobacco in the past 30 days over the period 1975-2012, is highly informative. 

Cigarette use continued a long-term decline, reaching 17%, the lowest prevalence since the survey’s introduction in 1975.  Cigarette use was lower than marijuana use (23%) for the fourth consecutive year.  Smokeless tobacco use among boys declined for the third consecutive year to 13.5%; use among girls was 1.6%. 

As I noted in 2009, (here), anti-tobacco forces have a brilliant strategy for dealing with tobacco use research: If it’s good, take credit; if it’s bad, blame the industry.  In a press release, the Campaign for Tobacco Free Kids did a little of both.  It implied that the gains were attributable to FDA regulation and tax increases, but it claimed that “…the tobacco industry still spends $8.5 billion a year — nearly $1 million every hour — to market its deadly and addictive products …that entice youth.”

If the tobacco industry is responsible for youth smoking, which industry is responsible for marijuana use?

For high school seniors, alcohol is by far the drug of choice.  While alcohol prevalence declined substantially from over 70% around 1980, it remains far too high.  In 2012, more than 40% of high school seniors had an alcoholic drink in the past 30 days; 28% of them had been drunk.  This is astounding, given that the legal age to consume alcohol is 21 years.  It is especially worrisome when considering how many teens and young adults die in auto accidents.  With the exception of the elderly, this age group (15-24 years) has by far the highest death rates from auto accidents.  It comprises 14% of the U.S. population but accounts for 27% of auto deaths (2,498 in 2009).

Any use of drugs among children is a problem that should concern every American. We need to better understand the social and behavioral reasons why children experiment with and adopt substance use, in order to develop prevention strategies.  But we need to prioritize limited resources based on the relative impact of each substance on children’s health. Cigarettes and smokeless tobacco should not be ignored, but the MTF survey documents that alcohol is the major threat.