PATIENTS AND METHODS

The study of the biologic activity of tamoxifen at lower doses included a total of 127 healthy hysterectomized women enrolled in a single center who were randomized through two separate experiments 3 years apart. A detailed description of the study has recently been published.¹³ Briefly, the first experiment included 61 women who were randomized to tamoxifen 20 mg/d or placebo, whereas the second experiment involved 66 women who were randomized to tamoxifen 10 mg/d or 10 mg every other day. Study participants were told to take their tablets at dinner. Morning fasting blood samples from both experiments were drawn at baseline, and after 1 and 2 months of treatment. Part of the aliquots were frozen and stored at −70°C. Except for blood cell count, which was measured on fresh samples, and lipoprotein(a), which was measured on frozen samples at the end of each experiment, the remaining biomarkers from all of the 127 participants were simultaneously measured on the frozen samples in a single session, thus enabling us to analyze the results of the two experiments as a single trial. These biomarkers include the following: total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, triglycerides, fibrinogen, antithrombin III, osteocalcin, and insulin-like growth factor I (IGF-I). Because serum cholesterol is a reliable tamoxifen-regulated biomarker as well as an established surrogate end point for coronary heart disease,¹⁴ the primary end point of the study was the change in total cholesterol from baseline to 2 months of treatment. Moreover, because IGF-I can stimulate breast proliferation in primates¹⁵ and because higher circulating levels of IGF-I have been associated with an increased risk of premenopausal breast cancer in a prospective study,¹⁶ the change of IGF-I was considered as a suitable surrogate biomarker for breast cancer and, therefore, as another main study end point.

Blood concentrations of tamoxifen and its main metabolites, N-desmethyltamoxifen (metabolite X) and desdimethyltamoxifen (metabolite Z), were measured in blinded fashion as to treatment allocation with high-performance liquid chromatography (HPLC), using the methods previously described.^17,18 Briefly, tamoxifen and its metabolites were determined in an acetonitrile extract from serum or plasma and were separated by reverse-phase, low-dispersion liquid chromatography. The drug and its metabolites were detected by conversion to fluorophors, which was achieved by subjecting the effluent of the column to ultraviolet light while passing a transparent coil. Because of small plasma sample volumes, the injection volume had to be reduced from 250 to 200 μL. However, this did not affect the precision of the assay, because the coefficients of variation for tamoxifen and metabolite X at a concentration of 10 ng/mL were as low as 4.7% and 6.6%, respectively.

Of the 127 subjects initially enrolled, aliquots for drug concentration measurements were available for 105 subjects. Because depletion of frozen aliquots occurred in a random way among the participants, the loss of samples is not expected to affect the comparison of the biomarker changes among different tamoxifen concentrations. The main patient characteristics are listed in Table 1. All patients had undergone hysterectomy (44 underwent bilateral oophorectomy and 25 underwent monolateral oophorectomy). Women were considered postmenopausal if they were older than 50 years of age when one or both ovaries had been preserved. The proportion of patients who had undergone bilateral oophorectomy was 38% in the placebo group, 38% in the group of patients receiving 10 mg of tamoxifen every other day, 39% in the 10 mg/d group, and 46% in the 20 mg/d group. The mean (± SD) duration of hormone replacement therapy (HRT) was 2 ± 0 months, 31 ± 23 months, and 26 ± 33 months in the placebo, tamoxifen 10 mg every other day, and tamoxifen 10 mg/d groups, respectively. No patient started HRT after tamoxifen and no patient had the dose of HRT changed during the study. The route of HRT administration was transdermal in all patients in the lower tamoxifen doses and oral in the two patients in the placebo group. No significant change in baseline characteristics was observed compared with the larger cohort.¹³ Among the 27 aliquots that were available in the placebo arm, we selected, on the basis of the randomization order, nine cases (one of every three) as an internal control of HPLC assay. Thus a total of 87 aliquots were assayed by HPLC. Of these, 73 were composed of serum aliquots and 14 of plasma aliquots (the latter were distributed in the treatment groups as follows: two in the placebo group, one in the 10 mg every other day group, three in the 10 mg/d group, and eight in the 20 mg/d group). In six patients, both serum and plasma aliquots were available. Because the HPLC method for tamoxifen measurement was developed on serum samples,¹⁷ an intrapatient comparison of serum and plasma measurements was performed in these six patients and in nine additional breast cancer patients. Results showed an average 25% underestimation in the plasma aliquots (range, 18% to 31%). Consequently, two separate statistical analyses were performed: one on serum samples and the other including plasma aliquots corrected for a 25% increasing factor. As the results of the two analyses were not different (ie, no significant concentration-response relationship was evident in the serum samples), the larger is presented. Furthermore, because all nine cases in the placebo arm showed no amount of tamoxifen and metabolites in the blood, all of the 32 placebo cases of the original study were given a value equal to zero and were included in the analysis for statistical purposes. Finally, as the correlation coefficients between the concentrations of tamoxifen and metabolite X and metabolite Z were high (0.89 and 0.71, respectively), the main analysis on the biomarker changes was carried out using tamoxifen concentration as the predictive variable.

The influence of several variables (age, age at menarche, age at first birth, menopausal status, age at menopause, family history of breast cancer, body mass index, waist-to-hip girth ratio, current duration of HRT, and smoking habit) on tamoxifen and metabolite levels was analyzed by a one-way analysis-of-variance model adjusted for treatment group. Log transformation of tamoxifen and metabolite values was performed to achieve normality. The principal analysis was focused on the effects of tamoxifen concentrations on the percent change in the biomarkers after 2 months of treatment. Outliers were noted among the biomarker changes, namely, total cholesterol (one patient), lipoprotein(a) (three patients), and antithrombin III (one patient). These outliers were excluded from the analyses, because the baseline or end-of-study values were abnormally extreme. Tamoxifen concentrations were treated either as a continuous or a categorical variable, the latter using quartiles as cutoff points. Because the intrapatient coefficient of variation in serial tamoxifen measurements during steady-state is approximately 10% to 15%,¹⁹ the use of the categorical variable provides a more robust analysis. When tamoxifen concentration was treated as a categorical variable, four contrasts were used to compare the five groups: placebo versus tamoxifen, and, among the four categories of detectable tamoxifen levels, linear, quadratic, and cubic trends. All calculations were performed using PROC GLM in SAS (Version 6.11, SAS Institute, Inc, Cary, NC).

RESULTS

The mean (± SD) blood concentrations of tamoxifen, metabolite X, and metabolite Z in the four treatment arms after 2 months of treatment are presented in Table 2. The blood concentrations of tamoxifen and metabolites attained with 10 mg/d and 10 mg every other day were 60% and 80% lower, respectively, compared with the conventional dose of 20 mg/d. Except for a borderline effect of HRT duration, which was associated with lower tamoxifen concentrations (but based on only 15 women using HRT), no other variable had an influence on drug or metabolite levels (data not shown). The percent changes in the biomarkers according to blood tamoxifen levels are listed in Table 3. Compared with placebo, tamoxifen induced a significant change in all biomarkers except high-density lipoprotein cholesterol, triglycerides, WBC count, and fibrinogen. Among tamoxifen categories, no significant concentration-response relationship was observed for all of the biomarkers except platelet count (Table 2). For instance, no regression of tamoxifen levels upon the change in total cholesterol and IGF-I was observed (Fig 1). Notably, however, levels of triglycerides showed a borderline significant trend to a greater increase with increasing tamoxifen concentrations (P for linear trend = .06; Table 2 and Fig 2).

View larger version:

• In this window
• In a new window

• PowerPoint Slide for Teaching

Fig 1.

Effect of tamoxifen blood concentrations on the percent change in (A) total cholesterol and (B) IGF-I after 2 months of treatment. Abbreviation: eod, every other day.

View larger version:

• In this window
• In a new window

• PowerPoint Slide for Teaching

Fig 2.

Effect of tamoxifen blood concentrations on the percent change in triglycerides levels after 2 months of treatment. Abbreviation: eod, every other day.

DISCUSSION

Tamoxifen, now out of patent in Europe, is the gold standard endocrine treatment for breast cancer,¹ and it has recently been shown to decrease by 50% breast cancer incidence in at-risk women.² However, its minimal active dose is unknown. This is somewhat surprising, given the evidence from in vitro experiments of a plateau of antitumor activity after saturation of the ER^10,11 and the lack of a dose-response relationship in terms of efficacy in clinical trials of adjuvant treatment.¹ In contrast, the carcinogenic effect of tamoxifen follows a clear dose-response relationship in the rat liver⁵ and, presumably, also in the human endometrium,^6-9 where DNA adducts have been found in one study.²⁰ Together, these observations suggest that a dose reduction should be explored in an attempt to reduce toxicity while retaining activity. In this framework, we have recently shown that 10 mg/d and 10 mg every other day of tamoxifen are comparable to the conventional dose of 20 mg/d in modulating a broad spectrum of established or putative intermediate biomarkers of several diseases, including circulating IGF-I for breast cancer.¹³ To provide further insight into the search for the minimal effective dose of tamoxifen, we have measured the blood concentrations of drug and metabolites to correlate their levels with the biomarker changes.

Despite a strong correlation between the dose administered and blood levels, our results show no evidence for a significant concentration-response relationship on the vast majority of the end-point biomarkers. Importantly, drug and metabolite levels attained in our study with the conventional dose of 20 mg/d were consistent with those reported in the literature,^19,21 thus providing a reliable reference group. Therefore, a mean tamoxifen concentration of 25 ng/mL (ie, 20% of the level attained with 20 mg/d), was associated with comparable changes in most end-point biomarkers. Moreover, where a trend to a concentration-response relationship was noted (namely, on platelet count and triglycerides levels), the use of the lowest dose seems to be preferable. Indeed, preliminary results of our phase III trial show that severe hypertriglyceridemia may sometimes occur with 20 mg/d of tamoxifen,³ and the higher reduction of platelet count observed at lower tamoxifen concentrations might be associated with a diminished risk of deep venous thromboembolic events. It is important to emphasize the notion that our study was not powered to detect equivalence. Therefore, the lack of difference among tamoxifen categories does not necessarily mean that all biomarker changes will actually be equivalent.

Although modulation of most of the study biomarkers reflects activation of the ER pathway and may thus be relevant to the antitumor activity of tamoxifen,¹³ the present data are hypothesis-generating, because it remains to be determined that a similar biomarker modulation will translate into a comparable clinical efficacy and a reduced toxicity of the lowest dose. Thus our results indicate the biologic plausibility of a dose reduction and support the implementation of trials with clinical end points to assess the efficacy and the safety of tamoxifen at lowered doses in comparison with the dose currently in use. However, the lack of a dose-response relationship on most of the biomarkers suggests that a further dose reduction is worth exploring, particularly among postmenopausal women, for whom both drug concentrations^22,23 and severe adverse events² tend to be higher compared with premenopausal women. The concept of a dose reduction is further supported by the observation that tamoxifen has a very high tissue distribution, ranging from 5 to 60 times its blood concentrations.^18,24 Assuming that pharmacokinetics are linear, which seems to be true in the dose range used in the present study, the breast tissue level attainable with 10 mg every other day may still exceed by several times the growth inhibitory concentration of tamoxifen in breast cancer cell lines, which is approximately 35 ng/mL.^11,25,26 In addition, the concomitant activity of metabolite X, which has a significant growth inhibitory effect in breast cancer lines,²⁷ may further contribute to the total drug antitumor activity. Finally, recent in vivo studies in a spontaneous rat mammary tumor model indicate that a daily dose corresponding to approximately 1 mg/d of tamoxifen has a complete preventive efficacy on mammary tumor formation.²⁸ All of these considerations suggest that a further reduction in tamoxifen dose should be explored to establish the minimal active concentration.

In conclusion, the present study shows that a substantial dose reduction did not influence in a significant way the level of 11 of the 12 biomarkers studied, although there was a strong correlation between the dose administered and tamoxifen blood level. A minor decrease in platelet count was observed during tamoxifen administration, but this decrease was most pronounced at the lower tamoxifen blood levels. The mean concentrations of triglycerides showed a trend to an increase with increasing tamoxifen concentrations. The results of the present study suggest that the dose of tamoxifen may be reduced without a significant loss of beneficial effects on the lipid profile and other intermediate biomarkers, including plasma IGF-I. Additional studies are warranted to establish the minimal effective regimen of tamoxifen before a new generation of prevention trials is implemented.