Smoking vs other risk factors as the cause of smoking-attributable deaths: confounding in the courtroom

JAMA. 2000 Aug 9;284(6):706-12. doi: 10.1001/jama.284.6.706.


Context: The surgeon general estimates that more than 400,000 deaths are attributable to smoking annually in the United States. The tobacco industry has criticized the surgeon general's estimates because they do not control for the lower educational and socioeconomic status of modern-day smokers.

Objective: To determine whether controlling for education, occupation, race, alcohol consumption, and various dietary factors, in addition to age and sex, substantially alters the relative and attributable risk estimates associated with tobacco smoking.

Design, setting, and participants: Nationwide American Cancer Society prospective cohort study of 974, 150 US adults aged 30 years or older, enrolled in 1982 and followed up through 1988. (The same study is used for the surgeon general and Centers for Disease Control and Prevention [CDC] estimates of smoking-attributable deaths in the United States.)

Main outcome measures: Death from each of the chronic diseases considered in the CDC's estimate of smoking-attributable mortality (cancers of the lung, oropharynx, larynx, esophagus, pancreas, kidney, bladder, and cervix; ischemic heart disease, arterial disease, and other heart conditions; stroke; chronic obstructive pulmonary disease; and other respiratory conditions). Estimates adjusted for multiple covariates were compared with those adjusted for age only among current and former vs never smokers.

Results: Adjusting for multiple covariates slightly decreased the relative and attributable risk estimates for current smoking in both men and women, but slightly increased the estimates for former smoking in women. Multivariate adjustment decreased the overall estimate of deaths attributable to smoking in the United States by approximately 1%, from 401,109 to 396,741 per year.

Conclusions: Our study suggests that federal estimates of deaths caused by smoking are not substantially altered by adjustment for behavioral and demographic factors associated with smoking beyond the current adjustment for age and sex. JAMA. 2000;284:706-712

MeSH terms

  • Adult
  • Behavior
  • Bias
  • Cause of Death
  • Female
  • Humans
  • Male
  • Middle Aged
  • Multivariate Analysis
  • Proportional Hazards Models
  • Prospective Studies
  • Risk Factors
  • Smoking / mortality*
  • Socioeconomic Factors
  • Survival Analysis
  • United States / epidemiology