CORRESPONDENCE Prevalence of Human Papillomavirus in Young Women
ADRIAN M. RENTON* LUKE WHITAKER
Odds ratio =
10+ 0
HPV +
HPV-
70 19
32 71
70x71 - 8 . 2 19x32 70x90 _ , 3
Relative prevalence
Academic Department of Public Health St. Mary's Hospital Medical School London, England
References (/)
(2)
LEY C, BAUER HM, REINGOLD A, ET AL: De-
terminants of genital human papillomavirus infection in young women. J Natl Cancer Inst 83:997-1003,1991 SCHLESSELMAN JJ: Case Control Studies. Monographs in Epidemiology and Biostatistics. London: Oxford Univ Press, 1982
Response We thank Dr. Renton and Dr. Whitaker for pointing out our casual (and technically incorrect) use of common epidemiologic terminology. We should have referred specifically to the relative odds of infection, rather than implying relative prevalence, in the sentence cited from the text.
= —T—
Thus, on the crude analysis, the women with 10 or more sexual partners are in fact only three times more likely to be infected with HPV than are the monogamous women. The discrepancy between these two measures of effect is explained by the relatively high prevalence of HPV infection. Odds ratios will provide good estimates of likelihood ratios or relative risks only where the prevalence of the disease is small. Their value as estimates of relative risks in case-control
M. MlCHELE MANOSt Department of Infectious Diseases Cetus Corporation Emeryville, Calif. ARTHUR REINGOLD
Epidemiology Program University of California, Berkeley MARK SCHIFFMAN
Environmental Epidemiology Branch Division of Cancer Etiology National Cancer Institute Bethesda, Md.
Retrospective Review of Neoadjuvant Chemotherapy for Osteogenic Sarcoma The retrospective analysis of chemotherapy for osteogenic sarcoma by Smith et al. (J) concluded that doxorubicin dose intensity is an important determinant of outcome. The authors used histologic response to preoperative (neoadjuvant) chemotherapy as a surrogate for disease-free survival. There are several concerns about this strategy. All of the studies cited by the authors showed a correlation between histologic response following initial surgery and subsequent disease-free survival, but that correlation varied. We have recently completed a retrospective review of our experience with neoadjuvant chemotherapy for osteogenic sarcoma (2). With longer duration of neoadjuvant chemotherapy, higher rates of histologic response were seen, but disease-free survival was unchanged. Good histologic response achieved after a short course of chemotherapy predicted a high probability of disease-free survival; good response achieved after prolonged chemotherapy predicted a lower probability (although still better than following a poor histologic response). For this reason, it is not safe to conclude that a high rate of good histologic response will automatically generate an equally high rate of disease-free survival. In addition, Smith et al. used the percentage of patients who achieved 90% tumor necrosis following neoadjuvant chemotherapy as the end point of interest. In the original description of histologic response to neoadjuvant chemotherapy, Rosen et al. (3) used a qualitative, not a quantitative, definition of response. Subsequent studies used a
^Correspondence to: M. Michele Manos, Ph.D., Department of Infectious Diseases, Cetus Corporation, 1400 Fifty-third St., Emeryville, CA 94608.
Correspondence to: Adrian M. Renton, M.S., Academic Department of Public Health, St. Mary's Hospital Medical School, Norfolk Place, London W2 1PG, England.
202
Journal of the National Cancer Institute
Downloaded from http://jnci.oxfordjournals.org/ at University of Otago on July 26, 2015
In their study of the prevalence of cervical human papillomavirus infection, Ley et al. (/) have presented the data as odds ratios and not as relative risks. They report an adjusted odds ratio of 11 for human papillomavirus (HPV) positivity for the group of women with more than 10 male sexual partners in their lifetimes relative to those with only one partner. On this basis, they conclude that "Women with 10 or more sexual partners were 11 times more likely to be infected with HPV than monogamous women. . . ." This conclusion is incorrect, as illustrated by the following calculation of the relative likelihoods and odds ratios from the data presented for the unadjusted odds ratio. lifetime number of male sexual partners
studies of rare diseases (2) has perhaps led to a situation in which the two are interpreted as generally equivalent. Because HPV infection was found to be common in the study by Ley et al., the odds ratios presented cannot be interpreted as equivalent to the relative risks of disease.
ADRIAN M. RENTON* LUKE WHITAKER
Odds ratio =
10+ 0
HPV +
HPV-
70 19
32 71
70x71 - 8 . 2 19x32 70x90 _ , 3
Relative prevalence
Academic Department of Public Health St. Mary's Hospital Medical School London, England
References (/)
(2)
LEY C, BAUER HM, REINGOLD A, ET AL: De-
terminants of genital human papillomavirus infection in young women. J Natl Cancer Inst 83:997-1003,1991 SCHLESSELMAN JJ: Case Control Studies. Monographs in Epidemiology and Biostatistics. London: Oxford Univ Press, 1982
Response We thank Dr. Renton and Dr. Whitaker for pointing out our casual (and technically incorrect) use of common epidemiologic terminology. We should have referred specifically to the relative odds of infection, rather than implying relative prevalence, in the sentence cited from the text.
= —T—
Thus, on the crude analysis, the women with 10 or more sexual partners are in fact only three times more likely to be infected with HPV than are the monogamous women. The discrepancy between these two measures of effect is explained by the relatively high prevalence of HPV infection. Odds ratios will provide good estimates of likelihood ratios or relative risks only where the prevalence of the disease is small. Their value as estimates of relative risks in case-control
M. MlCHELE MANOSt Department of Infectious Diseases Cetus Corporation Emeryville, Calif. ARTHUR REINGOLD
Epidemiology Program University of California, Berkeley MARK SCHIFFMAN
Environmental Epidemiology Branch Division of Cancer Etiology National Cancer Institute Bethesda, Md.
Retrospective Review of Neoadjuvant Chemotherapy for Osteogenic Sarcoma The retrospective analysis of chemotherapy for osteogenic sarcoma by Smith et al. (J) concluded that doxorubicin dose intensity is an important determinant of outcome. The authors used histologic response to preoperative (neoadjuvant) chemotherapy as a surrogate for disease-free survival. There are several concerns about this strategy. All of the studies cited by the authors showed a correlation between histologic response following initial surgery and subsequent disease-free survival, but that correlation varied. We have recently completed a retrospective review of our experience with neoadjuvant chemotherapy for osteogenic sarcoma (2). With longer duration of neoadjuvant chemotherapy, higher rates of histologic response were seen, but disease-free survival was unchanged. Good histologic response achieved after a short course of chemotherapy predicted a high probability of disease-free survival; good response achieved after prolonged chemotherapy predicted a lower probability (although still better than following a poor histologic response). For this reason, it is not safe to conclude that a high rate of good histologic response will automatically generate an equally high rate of disease-free survival. In addition, Smith et al. used the percentage of patients who achieved 90% tumor necrosis following neoadjuvant chemotherapy as the end point of interest. In the original description of histologic response to neoadjuvant chemotherapy, Rosen et al. (3) used a qualitative, not a quantitative, definition of response. Subsequent studies used a
^Correspondence to: M. Michele Manos, Ph.D., Department of Infectious Diseases, Cetus Corporation, 1400 Fifty-third St., Emeryville, CA 94608.
Correspondence to: Adrian M. Renton, M.S., Academic Department of Public Health, St. Mary's Hospital Medical School, Norfolk Place, London W2 1PG, England.
202
Journal of the National Cancer Institute
Downloaded from http://jnci.oxfordjournals.org/ at University of Otago on July 26, 2015
In their study of the prevalence of cervical human papillomavirus infection, Ley et al. (/) have presented the data as odds ratios and not as relative risks. They report an adjusted odds ratio of 11 for human papillomavirus (HPV) positivity for the group of women with more than 10 male sexual partners in their lifetimes relative to those with only one partner. On this basis, they conclude that "Women with 10 or more sexual partners were 11 times more likely to be infected with HPV than monogamous women. . . ." This conclusion is incorrect, as illustrated by the following calculation of the relative likelihoods and odds ratios from the data presented for the unadjusted odds ratio. lifetime number of male sexual partners
studies of rare diseases (2) has perhaps led to a situation in which the two are interpreted as generally equivalent. Because HPV infection was found to be common in the study by Ley et al., the odds ratios presented cannot be interpreted as equivalent to the relative risks of disease.
