To calculate a risk-benefit index of preventive tamoxifen, we used different data sources. For the preventive effects of tamoxifen, we could not know the exact chemopreventive effects of tamoxifen for Korean women because there were no prospective randomized controlled studies about tamoxifen chemoprevention in Korea. Therefore, we assumed that the effect of tamoxifen on Korean women was the same as on U.S. women. We used the relative risks of breast cancer, hip fracture, endometrial cancer and ischemic stroke from the NSABP-P1 study [1]. We summarized assumed relative risks with 95% confidence intervals in Table 1.

The incidence rates of hip fracture for Korean women were obtained from the study [4] that used the Korean National Health Insurance database. We re-calculated the incidence rate of female hip fractures according to 10-year age intervals [4]. There was no data under the age of 50, but a hip fracture at a young age is a rare event. So we assumed that the incidence is zero for those under age 50.

The incidence rates of stroke were obtained from another study [5] that also used the Korean National Health Insurance database. That study showed the 10-year probability of stroke per hundred individuals calculated for five-year age intervals. We re-calculated the incidence rate of female stroke according to the 10 year age intervals. Among those with cerebrovascular stroke, tamoxifen increases the incidence of ischemic stroke, but not hemorrhagic stroke. In the same study [5], the proportion of patients with an ischemic stroke was 58.9% of all females who had strokes. So we multiplied 0.589 by the incidence of strokes.

For the incidence rates of endometrial cancer, we used age-specific endometrial cancer incidence rates from the Korea Central Cancer Registry data [6].

The overall mortality rates and cause-specific mortality rates for patients with endometrial cancer and ischemic stroke were obtained from statistics on cause-specific death from the Korea National Statistical Office [7]. We searched the web-based Korea National Statistical Office's database by the International Classification of Disease (ICD)-10 diagnostic codes I63 for ischemic stroke and C54, C55 for endometrial cancer. The mortality rates for hip fracture were obtained from the hip fracture study [4]. The incidence rates and the mortality rates of hip fracture, endometrial cancer and ischemic stroke are summarized in Table 2.

To calculate the expected number of invasive breast cancer prevented over a 5-year period, we multiplied the estimated 5-year risk of invasive breast cancer by the relative risk reduced by tamoxifen. For example, if the 5-year risk of invasive breast cancer is 1%, it means that 100 breast cancers will occur among 10,000 people over a 5-year period. If they take a tamoxifen for prevention, 49 breast cancers will be prevented because the assumed relative risk of breast cancer is 0.51. The 5-year risk of invasive breast cancer was calculated by the Korean breast cancer risk assessment tool [3], a web-based breast cancer risk assessment tool for Korean women. In the tool, the risk factors for breast cancer (age, family history, body mass index, age of first delivery, breast feeding history and breast biopsy history) are entered, and breast cancer risk in the following 5 years and 10 years, and up to ages 64 and 74 years are calculated.

To calculate the expected number of hip fractures, endometrial cancers and ischemic strokes, we used Equation 1 which is the same method as that of Gail et al. [2]. In equation 1, the hazard rate means the incidence rate multiplied by the relative risk in Table 1. In tamoxifen untreated people, the relative risk is one, so the hazard rate is the same as the incidence rate. Mortality rates from other causes were calculated by subtraction of cause-specific mortality rates from overall mortality rates. In Table 3, we summarized the calculated number of hip fractures, endometrial cancers and ischemic strokes in the tamoxifen untreated and treated groups. The differences between tamoxifen treated and untreated cases also summarized. We compared these results with the results of Gail et al. [2] for U.S. women.

(Equation 1)

Expected number of non-breast event over 5 years

= hazard rate / (hazard rate + mortality rate from other causes) × [1 - exp{-5 × (hazard rate + mortality rate from other causes)}]

To estimate the risks and benefits of tamoxifen chemoprevention, we followed the methods of Gail et al. [2]. Life-threatening events related to tamoxifen are invasive breast cancer, hip fracture, endometrial cancer and stroke. The benefits of tamoxifen are prevention of both invasive breast cancer and hip fracture. The risks of tamoxifen are endometrial cancer and stroke (Figure 1). To summarize the risks and benefits of tamoxifen in a single number, we defined the risk-benefit index as follows:

(Equation 2)

Risk-benefit index

= expected number of invasive breast cancer prevented + expected number of hip fracture prevented - expected number of endometrial cancer - expected number of stroke

As the study of Gail et al. [2] suggested, we assumed that the magnitude of each event of breast cancer, hip fracture, endometrial cancer and stroke was the same. For example, one case of endometrial cancer that occurs is equivalent to one case of breast cancer prevented. If the risk-benefit index is negative, it shows that tamoxifen has more risk than benefit. For women with a previous history of hysterectomy, there is no risk of endometrial cancer. So we calculated this separately for women with a hysterectomy without adding the risk of endometrial cancer. We calculated the risk-benefit index for average 5-year risk of breast cancer according to the study of Kim et al. [3].

We assumed that the effect of tamoxifen on Koreans was the same as on the U.S. women and used the relative risks of breast cancer, hip fracture, endometrial cancer and ischemic stroke from NSABP-P1. But it was not certain that the effect of tamoxifen on Korean women was the same as for U.S. women. So we conducted sensitivity analyses in which we varied our assumptions regarding the relative risks of breast cancer, hip fracture, endometrial cancer and ischemic stroke. We calculated the 5-year risk of breast cancer for specific age groups when the risk-benefit index was zero, which meant that the risk and the benefit for tamoxifen was the same. In one-way sensitivity analysis, each parameter of relative risk of breast cancer, hip fracture, endometrial cancer and ischemic stroke was varied separately from the lower limit to the upper limit of the 95% confidence interval. Then, we conducted sensitivity analysis including all relative risks in the same way.

The expected numbers of hip fractures, endometrial cancers and ischemic strokes were calculated using Equation 1. The expected numbers of hip fractures and ischemic strokes increased with age (Table 3). The expected number of endometrial cancers is highest in the sixth decade of life.

From the results of the study of Gail et al. [2], the expected numbers of hip fractures per 10,000 in tamoxifen untreated U.S. white women were 50 for women in their 50s, 116 for women in their 60s and 339 those in their 70s. For endometrial cancers, the expected numbers of the same U.S. white women were 10 for those in their 40s, 40 for those in their 50s, 70 for their 60s and 75 for their 70s. For ischemic strokes, the expected numbers of the same U.S. white women were 22 for ages 40-49, 54 for 50-59, 156 for 60-69, and 342 for age 70 and over. Expected numbers of hip fractures and endometrial cancers were lower in Korean women than in U.S. white women, but the expected number of ischemic strokes was much higher in Korean women.

The expected number of hip fractures was decreased by tamoxifen, but the expected numbers of endometrial cancers and ischemic strokes were increased by tamoxifen (Table 3). The number of ischemic strokes dramatically increased with age. This increase was exaggerated with tamoxifen treatment (Table 3). The magnitude of endometrial cancer was smaller than that of hip fractures or ischemic strokes induced by tamoxifen.

The risk-benefit index was calculated using Equation 2. In Tables 4 and 5, the shaded area shows a positive risk-benefit index that has more benefits than risks. For the specific 5-year risk of breast cancer, the net benefit of tamoxifen was reduced according to increases with age because of the high risk of stroke in older women. So older Korean women had more risk than benefit from tamoxifen chemoprevention.

Comparing non-hysterectomy (Table 4) and hysterectomy (Table 5) women, there was little difference in the risk-benefit index because the incidence of endometrial cancer was low. The effect of endometrial cancer is smaller than effects of hip fracture or ischemic stroke induced by tamoxifen.

According to the study of Kim et al. [3], the average 5-year risk of breast cancer in Korea was 0.18% for age 35, 0.38% for age 45, and 0.28% for age 55. At that average risk of breast cancer, the risk-benefit index was 0.4, -9.3, and -61.0 for ages 35, 45, and 55, respectively. Therefore, only women younger than age 40 had a positive risk-benefit index for average 5-year risk of breast cancer.

We calculated the 5-year risk of breast cancer at each specific age group when the risk-benefit index was zero, which meant that the risk and benefit from tamoxifen was the same. The numbers were 0.17% for women in their 30s, 0.57% in their 40s, 1.52% in their 50s, 2.81% in their 60s, and 3.14% for women in their 70s. Sensitivity analysis regarding the 95% confidence interval of breast cancer relative risk as parameters, the point was converged as the age decreased. Sensitivity analyses regarding relative risk of hip fracture, endometrial cancer and ischemic stroke showed similar results. Sensitivity analysis that included all relative risks in the same way, also showed similar results (Figure 2). It showed that the result of this study regarding women younger than age 40 is robust.