BMJ  2004;328:72 (10 January), doi:10.1136/bmj.37936.585382.44

Paper

Cigarette tar yields in relation to mortality from lung cancer in the cancer prevention study II prospective cohort, 1982-8

¹ Department of Economics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA, ² Epidemiology and Surveillance Research, American Cancer Society, 1599 Clifton Road NE, Atlanta, GA 30329, USA

Abstract

Objective To assess the risk of lung cancer in smokers of medium tar filter cigarettes compared with smokers of low tar and very low tar cigarettes.

Design Analysis of the association between the tar rating of the brand of cigarette smoked in 1982 and mortality from lung cancer over the next six years. Multivariate proportional hazards analyses used to assess hazard ratios, with adjustment for age at enrolment, race, educational level, marital status, blue collar employment, occupational exposure to asbestos, intake of vegetables, citrus fruits, and vitamins, and, in analyses of current and former smokers, for age when they started to smoke and number of cigarettes smoked per day.

Setting Cancer prevention study II (CPS-II).

Participants 364 239 men and 576 535 women, aged 30 years, who had either never smoked, were former smokers, or were currently smoking a specific brand of cigarette when they were enrolled in the cancer prevention study.

Main outcome measure Death from primary cancer of the lung among participants who had never smoked, former smokers, smokers of very low tar ( 7 mg tar/cigarette) filter, low tar (8-14 mg) filter, high tar ( 22 mg) non-filter brands and medium tar conventional filter brands (15-21 mg).

Results Irrespective of the tar level of their current brand, all current smokers had a far greater risk of lung cancer than people who had stopped smoking or had never smoked. Compared with smokers of medium tar (15-21 mg) filter cigarettes, risk was higher among men and women who smoked high tar ( 22 mg) non-filter brands (hazard ratio 1.44, 95% confidence interval 1.20 to 1.73, and 1.64, 1.26 to 2.15, respectively). There was no difference in risk among men who smoked brands rated as very low tar (1.17, 0.95 to 1.45) or low tar (1.02, 0.90 to 1.16) compared with those who smoked medium tar brands. The same was seen for women (0.98, 0.80 to 1.21, and 0.95, 0.82 to 1.11, respectively).

Conclusion The increase in lung cancer risk is similar in people who smoke medium tar cigarettes (15-21 mg), low tar cigarettes (8-14 mg), or very low tar cigarettes ( 7 mg). Men and women who smoke non-filtered cigarettes with tar ratings 22 mg have an even higher risk of lung cancer.

Introduction

During the past 50 years, changes in the design and manufacture of cigarettes have markedly reduced their machine measured "tar" yields.^{1 2} The introduction of cellulose acetate filters in the 1950s, and subsequently more porous cigarette papers, reduced the average tar rating per cigarette in the United States from about 37 mg in 1950 to 22 mg in 1967.¹ The introduction of air ventilation holes in the filter tip in the late 1960s and expanded tobacco in the 1970s permitted manufacturers to market low tar (generally in the range of 8-14 mg per cigarette) and very low tar cigarettes ( 7 mg per cigarette). Concomitantly, the US average tar level per cigarette, as rated by the US Federal Trade Commission, declined to about 13 mg by 1990.¹ Similar trends in standardised tar yields have been reported in the United Kingdom^{3 4} and other countries.

While many case-control and cohort studies have examined risk of lung cancer in relation to type of cigarette smoked, nearly all have compared the risks of smoking high tar non-filter brands with smoking medium tar filter brands, or with the corresponding ranges of tar yield (see bmj.com for a detailed list of references). The three case-control studies that have included participants who smoked low tar brands^5-7 yielded negative or equivocal results, but the observation periods for these studies ended in 1980-1, when the combined market share of low tar and very low tar cigarettes in the United States had exceeded 10% for only five or six years.⁸ In most epidemiological studies, the observation period ended before 1986, when the market share in the United States had exceeded 10% for only a decade.⁸ Thus no large, long term prospective study has specifically compared the risk of lung cancer in smokers of medium tar filter cigarettes with that in smokers of low tar and very low tar filter cigarettes.

We analysed the relation between the tar rating of the brand of cigarette smoked in 1982 and mortality from lung cancer over six years among men and women in the cancer prevention study II (CPS-II), a nationwide prospective cohort of over one million US adults aged 30 years or older.

Methods

Details of the cancer prevention study, initiated by the American Cancer Society in 1982, have been published elsewhere.^9-12 From the cohort of 508 318 men and 676 270 women, we excluded those who reported a history of cancer other than non-melanoma skin cancer; men who ever smoked pipes or cigars or chewed tobacco; and men and women whose current smoking status could not be ascertained. The resulting cohort comprised 364 239 men and 576 535 women. The outcome measure was death from cancer of the trachea, bronchus, or lung as the underlying cause, coded from the death certificate. During the six year follow up, 2622 men and 1406 women died from these cancers.

On the basis of brand name reported by each current smoker at enrolment, as well as the size (regular, king size, 100 mm, 120 mm), and presence or absence of menthol and of a filter, we assigned a tar rating.^{9 13} We then grouped current brand tar ratings into very low tar ( 7 mg), low tar (8-14 mg), medium tar (15-21 mg), and high tar ( 22 mg). Unspecified current brands, as well as those current brands that could not otherwise be classified, were considered as a separate category. All brands in the very low and low tar ranges, as well as 99% of brands in the medium range, were filter cigarettes. Those in the high tar range were exclusively non-filter cigarettes.

We restricted our mortality follow up to six years (1982-8) to reduce possible misclassification of exposure due to quitting or brand switching during longer follow up.

We used Cox proportional hazards methods to estimate hazard ratios and 95% confidence intervals of mortality from lung cancer in people who had never smoked, former smokers, and current smokers of very low, low, and high tar brands, relative to smokers of brands with tar ratings of 15-21 mg (medium tar). We adjusted for multiple covariates that reflected possible differences in participants' characteristics, dietary practices, occupational exposures, or medical histories.

Results

Descriptive characteristics of the cohorts of 100 868 men and 124 270 women who were current smokers at enrolment and for the cohorts of 263 371 men and 452 265 women who had never smoked or who had quit smoking at enrolment are given on bmj.com. Smokers of brands with medium or high tar ratings were more likely to be African American, more likely to have attained no more than a high school education, more likely to have a recent blue collar job or a history of potential occupational asbestos exposure, and less likely to report use of vitamins A, C, and E than participants who smoked lower tar brands, who never smoked, or who quit smoking before age 35 years. Current smokers of very low tar cigarettes (especially men) tended to smoke more cigarettes a day. Moreover, among the 13% of current smokers who were re-enrolled in the CPS-II nutrition cohort,¹⁰ those men and women who had smoked very low tar and low tar cigarettes in 1982 were more likely to have quit smoking by 1992.

Men and women who smoked very low tar ( 7 mg) and low tar (8-14 mg) brands had risks of lung cancer indistinguishable from those who smoked medium tar (15-21 mg) brands (Wald test for homogeneity of strata¹⁴ P = 0.27 for men, P = 0.80 for women). The risk was higher in those who smoked non-filter cigarettes and substantially lower in men (fig 1) and women (fig 2) who quit smoking. People who quit smoking before age 35 years had risks of lung cancer approaching those of people who had never smoked. Further adjustment for age when people started to smoke and number of cigarettes smoked a day showed nearly identical patterns in current and former smokers to those shown in figures 1 and 2 (see bmj.com).

View larger version (21K): [in this window] [in a new window]   Fig 1 Hazard ratios for lung cancer in men, 1982-8, by smoking status and tar yield of brand smoked, relative to current smokers of brands with tar ratings 15-21 mg

 

View larger version (22K): [in this window] [in a new window]   Fig 2 Hazard ratios for lung cancer in women, 1982-8, by smoking status and tar yield of brand smoked, relative to current smokers of brands with tar ratings 15-21 mg

 

Among men who never smoked, 93 died from lung cancer. Among men who quit smoking, 23 who quit aged 35 years, 344 who quit aged 35-54 years, and 540 who quit aged 55 years died from lung cancer. Among women who never smoked, 211 died from lung cancer. Among women who quit smoking, 16 who quit aged 35 years, 122 who quit aged 35-54 years, and 131 who quit aged 55 years died from lung cancer.

We used multivariate sensitivity analyses in current smokers to examine whether varying the exclusion criteria or the boundaries of the very low tar and low tar categories materially altered the results (tables 1 and 2). In both men and women, the findings were essentially unchanged when people with emphysema and other diseases attributable to smoking were excluded, when the analyses were restricted to people who had smoked their current brand for a minimum of five or 10 years, or when the boundaries of the very low or low tar categories were altered slightly (see bmj.com).

View this table: [in this window] [in a new window]   Table 1 Multivariate analyses of mortality from lung cancer of men who were current smokers, according to tar level of cigarette smoked in 1982*

 

View this table: [in this window] [in a new window]   Table 2 Multivariate analyses of mortality from lung cancer of women who were current smokers, according to tar level of cigarette smoked in 1982*

 

Discussion

Although smokers of non-filter high tar cigarettes with tar ratings 22 mg had the highest risk of lung cancer, we detected no difference in risk among people who smoked medium tar cigarettes (15-21 mg), low tar cigarettes (8-14 mg), or very low tar cigarettes ( 7 mg). This pattern persisted after we adjusted for demographic characteristics, dietary habits, and occupational and medical histories. Moreover, our results were robust in sensitivity analyses that were restricted to people who had smoked their current brand for a minimum of 10 years or excluded smokers with emphysema and other smoking related diseases (see bmj.com). Similarly, the findings were essentially unchanged by minor variations in the boundaries of the low tar and very low tar categories or by omission of the number of cigarettes smoked a day as a covariate.

We observed the smoking habits of all participants in the cancer prevention study only at enrolment in 1982. However, among the 13% of participants who were re-enrolled in the CPS-II nutrition cohort,¹⁰ those who smoked very low tar and low tar cigarettes in 1982 were more likely to have quit smoking by 1992. If this finding of differential cessation applies to the entire cohort, then our estimates understate the actual risk of lung cancer associated with continued smoking of low tar and very low tar cigarettes.

Non-linear relation between tar levels and risk of lung cancer
Our findings challenge the assumption that the association between tar rating and lung cancer risk is necessarily linear. As the data points for current smokers in figures 1 and 2 show, extrapolations based on comparisons of only the highest and lowest tar groups, or parametric models in which tar yield is a continuous linear variable, can obscure a non-linear relation between tar levels and risk of disease. Indices of lifetime cumulative tar exposure^{15 16} are especially problematic because they confound tar yield with the number of cigarettes smoked a day and the number of years of smoking.

By the end of follow up in 1988, low tar and very low tar cigarettes had been on the US market for about two decades. Participants in the cancer prevention study who smoked very low tar cigarettes had used their current brand for an average of four to five years, while those who smoked low tar cigarettes had smoked their current brand for an average of seven to eight years. We did not attempt to analyse the tar levels of brands that participants recalled smoking in the past. Because the median age at entry was 53-54 years, most current smokers in the cohort could not have smoked low tar or very low tar brands exclusively over their lives. Accordingly, we could not evaluate the effect of the exclusive use of low and very low tar cigarettes from adolescence onward.

Findings consistent with other evidence
Our finding that there was no difference in the risk of lung cancer between people who smoked medium tar filter, low tar filter, and very low tar filter cigarettes is consistent with evidence of compensatory smoking. Addicted smokers who switch from a higher to lower tar cigarette can maintain their nicotine intake by blocking ventilation holes, increasing the puff volume or the time during which the smoke is retained in the lungs, and smoking more cigarettes.¹⁷ As a result, the actual dose of toxicants to the smoker may be much higher than is predicted by machine measured yields. Changes in inhalation patterns induced by lower tar cigarettes may increase the surface area of the lung exposed to carcinogens in smoke and thus result in greater deposition of submicron sized particles deeper into the airways.¹⁸ An increase in the depth of inhalation may have contributed to the marked increase among smokers in the incidence of adenocarcinoma of the lung (a cancer that arises in the more peripheral tissues of the lung) in the United States and other countries.¹⁹ In fact, adenocarcinoma of the lung was found to be more strongly associated with cigarette smoking in the second cancer prevention study (1982-8) than in the first (1960-72).¹⁹ Finally, changes in tobacco curing and blending have increased the delivery of carcinogenic, tobacco specific nitrosamines^{1 20} even as average tar yields have declined. Tar yield is a relatively weak predictor of a brand's delivery of carcinogenic, tobacco specific nitrosamines.²¹

What is already known on this topic Nearly all previous epidemiological studies of risk of lung cancer in relation to the type of cigarette smoked have compared smokers of high tar non-filter cigarettes ( 22 mg tar) with those of medium tar filter cigarettes (15-21 mg) No large, long term prospective epidemiological study has specifically compared the risk of lung cancer in smokers of medium tar filter brands with the risk in smokers of low tar (8-14 mg) and very low tar ( 7 mg) filter brands What this study adds The risk of lung cancer was no different in people who smoked medium tar cigarettes, low tar cigarettes, or very low tar cigarettes Men and women who smoked non-filtered cigarettes with tar ratings 22 mg had even higher risks of lung cancer All current smokers, regardless of the tar level of their current brand, had substantially greater risks of lung cancer than those people who had never smoked or who had quit smoking

While our finding that smokers of high tar non-filter cigarettes had higher risks of lung cancer may reflect unmeasured differences between smokers of non-filter and filter cigarettes,¹⁸ it is none the less consistent with many other case-control and cohort studies (see bmj.com for detailed references). Reducing the use of high tar non-filter cigarettes may thus provide limited public health benefits in those countries where such products are commonly used. Although non-filter cigarettes currently represent no more than 1% of cigarette sales in the United States and the United Kingdom,^{22 23} they comprised about 20% of cigarettes sold in China,²⁴ 15% in France, and 6-20% in Eastern Europe as late as 1996.²²

---

This is an abridged version; the full version is on bmj.com

Contributors: See bmj.com

Funding: No specific funding.

Competing interests: JEH has testified as an expert witnesses on behalf of plaintiffs in tobacco-related litigation. MJT has testified as an uncompensated expert on behalf of plaintiffs and the American Cancer Society in tobacco related cases.

Ethical approval: Approved by the institutional review board at Emory University.

References

1. Hoffmann D, Djordjevic M, Brunnemann K. Changes in cigarette design and composition over time and how they influence the yields of smoke constituents. In: Shopland D, ed. The FTC cigarette test method for determining tar, nicotine, and carbon monoxide yields of US cigarettes: report of the NCI expert committee. Bethesda, MD: US National Institutes of Health, National Cancer Institute, 1996: 15-37. (NIH Publication No 96-4028; NCI smoking and tobacco control monograph No 7.)
2. Kozlowski LT, O'Connor RJ, Sweeney CT. Cigarette design. Risks associated with smoking cigarettes with low machine-measured yields of tar and nicotine. Bethesda, MD: US National Institutes of Health, National Cancer Institute, NCI Smoking and Tobacco Control, 2001: 13-37. (Monograph No 13.)
3. Wald N, Doll R, Copeland G. Trends in tar, nicotine, and carbon monoxide yields of UK cigarettes manufactured since 1934. BMJ 1981;282: 763-5.
4. Kiryluk S, Wald N. Trends in cigarette smoking habits in the United Kingdom, 1905-1985. In: Wald N, Froggatt P, eds. Nicotine, smoking and the low tar programme. Oxford: Oxford University Press, 1989.
5. Vutuc C, Kunze M. Lung cancer risk in women in relation to tar yields of cigarettes. Prev Med 1982;11: 713-6.[CrossRef][ISI][Medline]
6. Vutuc C, Kunze M. Tar yields of cigarettes and male lung cancer risk. J Natl Cancer Inst 1983;71: 435-7.
7. Wilcox HB, Schoenberg JB, Mason TJ, Bill JS, Stemhagen A. Smoking and lung cancer: risk as a function of cigarette tar content. Prev Med 1988;17: 263-72.[CrossRef][ISI][Medline]
8. US Federal Trade Commission. Domestic market share of cigarettes by tar yield. Federal Trade Commission cigarette report for 2000. Washington, DC: US Federal Trade Commission, 2002.
9. Stellman SD, Garfinkel L. Smoking habits and tar levels in a new American Cancer Society prospective study of 1.2 million men and women. J Natl Cancer Inst 1986;76: 1057-63.
10. Thun MJ, Calle EE, Rodriguez C, Wingo PA. Epidemiological research at the American Cancer Society. Cancer Epidemiol Biomarkers Prev 2000;9: 861-8.[Free Full Text]
11. Garfinkel L. Selection, follow-up, and analysis in the American Cancer Society prospective studies. Washington, DC: Government Printing Office, 1985: 49-52. (NIH Publication 85-2713, National Cancer Institute monograph 67.)
12. Calle EE, Terrell DD. Utility of the national death index for ascertainment of mortality among cancer prevention study II participants. Am J Epidemiol 1993;137: 235-41.[Abstract/Free Full Text]
13. US Federal Trade Commission. Report of `tar,' nicotine, and carbon monoxide of the smoke of 200 varieties of cigarettes. Washington, DC: US Federal Trade Commission, 1981.
14. Rothman KJ, Greenland S. Modern epidemiology. 2nd ed. Philadelphia: Lippincott Williams & Wilkins, 1998.
15. Zang EA, Wynder EL. Cumulative tar exposure. A new index for estimating lung cancer risk among cigarette smokers. Cancer 1992;70: 69-76.[CrossRef][ISI][Medline]
16. Kunze M, Vutuc C. Threshold of tar exposure: analysis of smoking history of male lung cancer cases and controls. In: Gori GB, Bock FG, eds. A safe cigarette? Banbury report 3. Cold Spring Harbor, NY: Cold Spring Harbor Laboratory, 1980: 29-36.
17. Benowitz N. Compensatory smoking of low-yield cigarettes. In: Shopland DR, Burns DM, Benowitz NI, Amacher RH, eds. Risks associated with smoking cigarettes with low machine-measured yields of tar and nicotine. Bethesda, MD: US National Institutes of Health, National Cancer Institute, 2001: 39-64. (NCI smoking and tobacco control monograph No 13.)
18. Burns DR, Major JM, Shanks TG, Thun MJ, Samet JM. Smoking lower yield cigarettes and disease risks. In: Shopland DR, Burns DM, Benowitz NI, Amacher RH, eds. Risks associated with smoking cigarettes with low machine-measured yields of tar and nicotine. Bethesda, MD: US National Institutes of Health, National Cancer Institute, 2001: 65-158. (NCI smoking and tobacco control monograph No 13.)
19. Thun MJ, Lally CA, Flannery JT, Calle EE, Flanders WD, Heath CW Jr. Cigarette smoking and changes in the histopathology of lung cancer. J Natl Cancer Inst 1997;89: 1580-6.[Abstract/Free Full Text]
20. Hoffmann D, Hoffmann I. Tobacco consumption and lung cancer. In: Hansen HH, ed. Lung cancer. Boston: Kluwer Academic Publications, 1994: 1-42.
21. Harris JE. Smoke yields of tobacco-specific nitrosamines in relation to FTC tar level and cigarette manufacturer: analysis of the Massachusetts benchmark study. Public Health Rep 2001;116: 336-43.[CrossRef][ISI][Medline]
22. Forey B, Hamling J, Lee P, Wald N, eds. International smoking statistics. 2nd ed. London: Oxford University Press, 2002.
23. US Department of Agriculture, Economic Research Service. Tobacco outlook, Washington, DC: USDA, April 2002. (TBS-252.)
24. Yang G, Fan L, Tan J, Qi G, Zhang Y, Samet JM, et al. Smoking in China: findings of the 1996 national prevalence survey. JAMA 1999;282: 1247-53.[Abstract/Free Full Text]

(Accepted 7 November 2003)

CiteULike    Complore    Connotea    Del.icio.us    Digg    Reddit    Technorati    What's this?

Relevant Article

Lower tar makes no difference
BMJ 2004 328: 0. [Full Text] [PDF]

This article has been cited by other articles:

• Cropsey, K., Eldridge, G., Weaver, M., Villalobos, G., Stitzer, M., Best, A. (2008). Smoking Cessation Intervention for Female Prisoners: Addressing an Urgent Public Health Need. Am. J. Public Health 98: 1894-1901 [Abstract] [Full text]  
• O'Connor, R J, Hammond, D, McNeill, A, King, B, Kozlowski, L T, Giovino, G A, Cummings, K M (2008). How do different cigarette design features influence the standard tar yields of popular cigarette brands sold in different countries?. Tobacco Control 17: i1-i5 [Abstract] [Full text]  
• O'Connor, R J, Hurley, P J (2008). Existing technologies to reduce specific toxicant emissions in cigarette smoke. Tobacco Control 17: i39-i48 [Abstract] [Full text]  
• Melikian, A. A., Djordjevic, M. V., Chen, S., Richie, J. Jr., Stellman, S. D. (2007). Effect of Delivered Dosage of Cigarette Smoke Toxins on the Levels of Urinary Biomarkers of Exposure. Cancer Epidemiol. Biomarkers Prev. 16: 1408-1415 [Abstract] [Full text]  
• Brennan, P., Crispo, A., Zaridze, D., Szeszenia-Dabrowska, N., Rudnai, P., Lissowska, J., Fabianova, E., Mates, D., Bencko, V., Foretova, L., Janout, V., Fletcher, T., Boffetta, P. (2006). High Cumulative Risk of Lung Cancer Death among Smokers and Nonsmokers in Central and Eastern Europe. Am J Epidemiol 164: 1233-1241 [Abstract] [Full text]  
• Tindle, H. A., Rigotti, N. A., Davis, R. B., Barbeau, E. M., Kawachi, I., Shiffman, S. (2006). Cessation Among Smokers of "Light" Cigarettes: Results From the 2000 National Health Interview Survey. Am. J. Public Health 96: 1498-1504 [Abstract] [Full text]  
• Sassen, A. W., Richter, E., Semmler, M. P., Harreus, U. A., Gamarra, F., Kleinsasser, N. H. (2005). Genotoxicity of Nicotine in Mini-Organ Cultures of Human Upper Aerodigestive Tract Epithelia. Toxicol Sci 88: 134-141 [Abstract] [Full text]  
• Godtfredsen, N. S., Prescott, E., Osler, M. (2005). Effect of Smoking Reduction on Lung Cancer Risk. JAMA 294: 1505-1510 [Abstract] [Full text]  
• Hecht, S. S. (2005). Carcinogenicity studies of inhaled cigarette smoke in laboratory animals: old and new. Carcinogenesis 26: 1488-1492 [Abstract] [Full text]  
• Kleinsasser, N. H., Sassen, A. W., Semmler, M. P., Harreus, U. A., Licht, A.-K., Richter, E. (2005). The Tobacco Alkaloid Nicotine Demonstrates Genotoxicity in Human Tonsillar Tissue and Lymphocytes. Toxicol Sci 86: 309-317 [Abstract] [Full text]  
• Lewis, S, Arnott, D, Godfrey, C, Britton, J (2005). Public health measures to reduce smoking prevalence in the UK: how many lives could be saved?. Tobacco Control 14: 251-254 [Abstract] [Full text]  
• Hammond, D., Fong, G. T., Cummings, K. M., Hyland, A. (2005). Smoking Topography, Brand Switching, and Nicotine Delivery: Results from an In vivo Study. Cancer Epidemiol. Biomarkers Prev. 14: 1370-1375 [Abstract] [Full text]  
• Hecht, S. S., Murphy, S. E., Carmella, S. G., Li, S., Jensen, J., Le, C., Joseph, A. M., Hatsukami, D. K. (2005). Similar Uptake of Lung Carcinogens by Smokers of Regular, Light, and Ultralight Cigarettes. Cancer Epidemiol. Biomarkers Prev. 14: 693-698 [Abstract] [Full text]  
• Stewart, B. W., Coates, A. S. (2005). Cancer Prevention: A Global Perspective. JCO 23: 392-403 [Abstract] [Full text]  
• Doll, R., Peto, R., Boreham, J., Sutherland, I. (2004). Mortality in relation to smoking: 50 years' observations on male British doctors. BMJ 328: 1519- [Abstract] [Full text]  
• Davies, H E (2004). Non-linear relationship between lung cancer risk and cigarette tar yield. Thorax 59: 354-354 [Full text]  
• (2004). From the library. Br. J. Ophthalmol. 88: 600-600 [Full text]  
• (2004). Hit Parade. BMJ 328: 774-774 [Full text]  
• (2004). Low Tar Doesn't Mean Low Risk. Journal Watch Cardiology 2004: 4-4 [Full text]  
• (2004). Low Tar Doesn't Mean Low Risk. JWatch General 2004: 3-3 [Full text]  

Rapid Responses:

Read all Rapid Responses

DNA Adducts in Smokers

John M McLaren Howard
bmj.com, 14 Jan 2004 [Full text]

The answer to a quandary?

Alfred P J Lake
bmj.com, 18 Jan 2004 [Full text]