Fax: (805) 715-8106
Estimating recurrences prevented from using trastuzumab in HER-2/neu–positive adjuvant breast cancer in the United States
Version of Record online: 23 AUG 2010
Copyright © 2010 American Cancer Society
Volume 116, Issue 24, pages 5575–5583, 15 December 2010
How to Cite
Danese, M. D., Lalla, D., Brammer, M., Doan, Q. and Knopf, K. (2010), Estimating recurrences prevented from using trastuzumab in HER-2/neu–positive adjuvant breast cancer in the United States. Cancer, 116: 5575–5583. doi: 10.1002/cncr.25347
- Issue online: 3 DEC 2010
- Version of Record online: 23 AUG 2010
- Manuscript Accepted: 1 MAR 2010
- Manuscript Revised: 16 FEB 2010
- Manuscript Received: 23 NOV 2009
- breast cancer;
Breast cancer recurrence is associated with significant morbidity, mortality, and cost. Patients with early stage HER2+ tumors are at increased risk of recurrence. The use of trastuzumab for these patients has been shown to reduce recurrences and improve overall survival.
A Monte Carlo simulation was conducted based on Surveillance, Epidemiology, and End Results incidence rates for 2005, United States Census data for 2005, and the results of key trials of the adjuvant use of trastuzumab. Patients included in this analysis had incident, HER2+, stage I to III breast cancer. The number of recurrences that could be prevented with trastuzumab, the cardiac adverse events that might occur, and the associated cost savings were estimated.
Approximately 31,200 women had HER2+ breast cancer in 2005, of whom 7298 would have had a recurrence over the subsequent 5 years despite standard of care adjuvant treatment. If trastuzumab were added to their regimen, 2791 women might have avoided recurrence, and 948 may have had an asymptomatic or symptomatic cardiac adverse event, for a ratio of expected recurrences to cardiac adverse events of 3.2 (95% confidence interval, 1.5-5.9). In economic terms, avoidance of future breast cancer recurrences was associated with lifetime reduction in future direct and indirect costs on the order of $240 million to $1.7 billion.
On the basis of the simulation results, targeting HER2+ tumors with trastuzumab in the adjuvant setting should prevent a significant number of women from recurrence events, with important outcomes for patients, physicians, payers, and society. Cancer 2010. © 2010 American Cancer Society.
Recurrences of breast cancer are associated with significant morbidity, mortality, and cost.1 Women with HER2+ tumors are at higher risk of recurrence than other women, suggesting that the clinical and economic burden is even higher in this subgroup of cancer patients.2-4 Trastuzumab (Herceptin, Genentech, Inc., South San Francisco, Calif) has been shown in multiple studies to reduce recurrences in women with stage I to III HER+ breast cancer who have been appropriately diagnosed and treated.5-7 In addition to demonstrated clinical efficacy, several cost-effectiveness analyses of trastuzumab in the adjuvant setting have found its use to represent an economically attractive value proposition, according to commonly accepted standards.8-12 Taken together, these reports address the important issues of safety, efficacy, and value, however, none of them evaluates the impact on the US population of introducing this new therapeutic option in terms of either the recurrences prevented or the potential associated cost savings.
The purpose of this research was to estimate the annual number of patients in whom recurrence events (local recurrences, distant recurrences, contralateral tumors, and deaths) might be prevented. Typically, the concept of extrapolating results to a population has been focused on understanding risk factors for disease and the effect that eliminating the risk might have on the population. This concept of measuring attributable risk or etiologic fraction has been well documented, albeit the subject of some debate because of the variety of measures and interpretations.13
In this research, we are evaluating the benefit of a beneficial exposure (treatment), rather than the risk, and we are doing so using Monte Carlo simulation techniques rather than the results of any single study.14 Furthermore, we address not only the primary benefit of treatment, but also the primary risk, cardiac adverse events. Finally, we explore the economic impact of prevented recurrence events to society.
MATERIALS AND METHODS
HER2+ Breast Cancer Incidence
The Monte Carlo model uses data from the Surveillance, Epidemiology, and End Results (SEER) program as the basis for the simulation, with crude rates and confidence intervals obtained using SEERStat software version 188.8.131.52 Women aged 30 years and older with stage I to III breast cancer were included from the 17 available registries for 2005. The results were stratified by age (30-49, 50-69, and 70+), estrogen receptor (ER) status, progesterone receptor (PR) status, and nodal involvement (0 nodes vs > 0 nodes).
We assumed that patients with missing ER, PR, or nodal status would not be HER2 tested and are excluded from our rate estimates (7%-9% of patients, depending on age). Excluding these patients also minimizes the number of patients who either were diagnosed at autopsy or died very quickly after diagnosis and could not be tested. Furthermore, the model excludes both recurrent patients and incident metastatic patients.
Because HER2 status is not available in SEER, the number of HER2+ patients was estimated as a proportion of the total population within each of the 4 mutually exclusive strata defined by ER and PR status. This was done by applying estimates of the association between HER2 status and both ER and PR status based on a review of the literature. We applied the results from a single, large, population-based study using immunohistochemistry (IHC) for our base case.16 We then applied rates of fluorescence in situ hybridization positivity for the proportion of patients who were 2+ using IHC to reflect current recommendations for HER2 testing.17, 18 Because this is a critical aspect of the simulation, we compared the estimates of HER2+ by each of the ER and PR strata from our base case to those from other studies for validation purposes and found little variation (sensitivity analyses not reported).19-23
Population estimates of HER2+ breast cancer
The resulting incidence rates were applied to the 2005 US female population counts to estimate the number of HER2+ breast cancer cases.24 The estimated HER2+ patient counts were stratified by nodal status, and age according to their distributions in SEER (ie, they did not vary by HER2 status). A recent study suggested that most, but not all, patients (98%) who should be tested for HER2 status are receiving testing, and this is also incorporated into the model.25
Efficacy of Treatment, Recurrence, and Cardiac Adverse Events
Efficacy of treatment
The efficacy rate for trastuzumab was based on 2 studies that compared doxorubicin and cyclophosphamide followed by a taxane versus the same regimen plus 52 weeks of trastuzumab.5, 6 These results are consistent with other reports on the efficacy of trastuzumab in the adjuvant setting.7, 26 The relative risk of recurrence with trastuzumab was considered to be constant across subgroups, and to persist for 5 years.27 After 5 years, no recurrences were assumed to be prevented.
Recurrence events were defined as local recurrences, distant recurrences, contralateral tumors, and deaths. Overall recurrence rates from the placebo group were stratified by nodal status and pooled across studies. The rate for the trastuzumab group was estimated by applying the pooled hazard ratio to the placebo recurrence rates. The rate of recurrence was assumed to be constant over the 5-year risk period.13 Survival after recurrence was not modeled.
Two studies (National Surgical Adjuvant Breast and Bowel Project B-31 and North Central Cancer Treatment Group N9831) estimated virtually identical proportions of patients experiencing a cardiac event (primarily defined as asymptomatic ejection fraction decline below a threshold or frank congestive heart failure including symptoms of dyspnea with normal activity) based on patients with a maximum of 3 years of follow-up.28-30 Because the B-31 data were recently updated through 5 years and are very similar to the 3-year results, we elected to use the B-31 5-year cardiac data, because it matched the time frame we used to estimate recurrences.30, 31 Whereas the intent-to-treat trial analysis included patients who never received Herceptin because of anthracycline-related cardiac complications, the National Surgical Adjuvant Breast and Bowel Project B-31 analysis of cardiac adverse events excluded them. Therefore, these patients (approximately 7.5%) were also removed from the group considered to be at risk of trastuzumab-associated cardiac adverse events for consistency, and they did not benefit from trastuzumab therapy.28
Costs of Recurrence Events
We searched the literature to estimate the costs attributable to breast cancer recurrence including direct (medical and nonmedical) and indirect (loss income, productivity, and caregiving) costs from a societal perspective. The reference year for the analysis is 2008. Direct medical costs for recurrence events (except death) were adapted from a paper by Stokes et al1; we extracted the 12-month incremental cost of recurrence for women with each event compared with women with no recurrence. These costs were inflated to 2008 dollars using the medical portion of the consumer price index.31 In the study by Stokes et al, most of the recurrence events and costs, particularly for distant recurrence, were incurred in the first year. Therefore, we used the 12-month incremental cost as a proxy for the cost of the recurrence events. In this evaluation, costs of follow-up, including the cost of adverse events, were excluded.
We took 2 approaches to valuing the indirect costs of breast cancer recurrence: 1 excluding the cost of lost years of life, and the other including it. When excluding the cost of early death, we identified published estimates of patient time cost,32 early retirement,33 and lost time from work34 (we could not identify caregiver burden costs associated with recurrence). We assumed that patient time costs were similar for initial and recurrent cancer. Where necessary, we valued hours using national hourly labor rates.35 Finally, we incorporated published variability estimates in all costs to allow for variations in our inputs that reflect both uncertainty and variability.
When including death, we used the societal cost of lost productivity (including caregiver costs) from breast cancer death published by Bradley et al.36 In this study, the lost years of life, and their value in terms of productivity losses, decreases with age. To be conservative, we applied the lowest cost applicable to each of our 3 age groups (eg, the cost from Bradley et al for age 45-49 was used for our 30-49 age group). Furthermore, because not all cancer recurrences lead to death, and because life expectancy after breast cancer diagnosis can take 15 years to return to normal,37, 38 we only applied this cost to the proportion of recurrent patients who were expected to die within 5 years of diagnosis (79%).1 See Table 1 for key data inputs to the model.
|Model Input||Input (SD or CI)||Distribution||Reference|
|Incidence of stage I to III breast cancer in women by age, ER, PR, and nodal status (2005)||24 stratum-specific estimates||Normal||SEER15|
|Efficacy of trastuzumab (hazard ratio)||0.53 (0.40-0.72)||Log-normal||Joint analysis,5 BCIRG6|
|Proportion HER2+ by ER and PR||Normal||Francis 2007,16 Yaziji 2004,17 Owens 200418|
|ER+ and PR+||0.13 (0.12-0.14)|
|ER− and PR+||0.24 (0.21-0.27)|
|ER+ and PR−||0.25 (0.18-0.32)|
|ER− and PR−||0.35 (0.33-0.38)|
|Recurrence rates per patient-year at risk (no trastuzumab)||Normal||Francis 2007,16 Yaziji 2004,17 Owens 200418|
|Node negative||0.041 (0.030-0.056)|
|Node positive||0.083 (0.075-0.091)|
|Direct cost of recurrence (2008 $)||$30,400 ($27,300 to $33,400)||Normal||Adapted from Stokes 20081|
|Indirect cost of recurrence (2008 $ in millions)||Log-normal||Bradley 2008,36 Stokes 20081|
|Age, 30-49 years||$1.43 ($0.68 to $2.66)|
|Age, 50-69 years||$0.40 ($0.16 to $0.81)|
|Age, ≥70 years||$0.09 ($0.07 to $0.11)|
Several possible exclusions were considered as part of the scenario analyses. These included the presence of underlying cardiovascular disease based on published prevalence estimates, and the presence of lymph node-negative, small tumors (<1 cm), as suggested by National Comprehensive Cancer Network guidelines.39, 40
More specifically, we defined 3 scenarios selected to represent low and high estimates, with the base case selected to represent something in between. The base case excluded women who would not be tested for HER2, excluded half of women with small (<1 cm) lymph node-negative tumors defined by the National Comprehensive Cancer Network guidelines, and included all recurrence-related events defined in the trials (including second primary and contralateral tumors, and death).
The low estimate made the following changes from the base case: excluded women with severe underlying cardiovascular disease at baseline, excluded all women with small (<1 cm) lymph node-negative tumors defined by the National Comprehensive Cancer Network guidelines, and only included recurrence events (ie, no deaths or contralateral tumors). The high estimate made the following changes from the base case: assumed all patients would be tested for HER2, and included all lymph node-negative patients regardless of tumor size.
Finally, the duration of time over which recurrences were prevented was adjusted to evaluate the importance of this assumption on the final results.
Probabilistic model inputs were included as distributions (mean and variance) to reflect both variability and uncertainty. Normal distributions were used for continuous variables, with binomial distributions used for dichotomous events, and log-normal distributions used for cost data. Relative risks were sampled from a normal distribution on the log scale and transformed back.
When >1 data source was available, variance estimates were created using meta-analysis techniques and weighted according to the inverse of the variance of each estimate.41 Breast cancer incidence rate standard errors were based on estimates provided by SEERStat software. All other estimates were based on values in the literature, with the exception of the costs of productivity losses from Bradley et al,36 which were not reported. The standard deviations for these costs were estimated as proportions of the mean using a proxy. The proxy we selected was the ratio of the mean cost to its standard deviation in our simulation of the cost of recurrence, which excluded the cost of lost years of life (11% to 43% depending on age).
The model was run as a Monte Carlo simulation in Microsoft Excel using the @Risk 5.0 add-in (Palisade Corp., Ithaca, NY). For each iteration of the model, new values from each input distribution were randomly selected, and the results were recalculated automatically. Results taken across multiple iterations were used to estimate the results. Convergence was evaluated by comparing the stability of the key results across different numbers of iterations from 1000 to 10,000. On the basis of the stability of the trial runs, all simulations were run using 5000 iterations and reported as the mean and middle 95% of the empirical distribution, which is referred to as a confidence interval (CI).
The data inputs derived from the literature and used in the model are provided in Table 1. According to the base case estimates, 31,171 women had HER2+ breast cancer in 2005, of whom 29,159 might possibly have been treated given our assumptions. Of these women, 7298 would have had a recurrence over the following 5 years despite current treatment standards. Of these women who recurred without trastuzumab, 2791 would have avoided recurrence with the addition of trastuzumab (see Figs. 1 and 2). Trastuzumab use also would have been associated with an increase of 948 cardiac adverse events, the definition of which includes both symptomatic heart failure and asymptomatic reductions in left ventricular ejection fraction (see Fig. 3). On the basis of these figures, the ratio of expected recurrences to cardiac adverse events is 3.2 (95% CI, 1.5-5.9). This ratio was notably higher for the youngest cohort of women compared with the older 2 cohorts pooled together (7.2 vs 2.7). See Table 2 for ratios and 95% CI by age group.
|Quantity||Age, 30-49 Years, No. (95% CI)||Age, 50-69 Years, No. (95% CI)||Age, ≥70 Years, No. (95% CI)|
|Total recurrences prevented||692 (394-974)||1348 (763-1901)||751 (422-1074)|
|Cardiac adverse events||111 (24-202)||529 (267-809)||308 (153-470)|
|Recurrence prevented to cardiac event ratio||7.2 (2.6-22.2)||2.8 (1.3-5.5)||2.7 (1.2-5.3)|
|Economic, excluding lost productivity (death)|
|Recurrence cost per patient (excluding death; thousands)||$77 ($47-$152)||$107 ($59-$227)||$35 ($30-$46)|
|Total cost (excluding death; millions)||$53 ($24-$114)||$144 ($61-$330)||$27 ($14-$40)|
|Economic, including lost productivity (death)|
|Recurrence cost per patient (including death; thousands)||$1,458 ($706-$2,694)||$427 ($194-$844)||$119 ($102-$139)|
|Total cost (including death; millions)||$1,009 ($399-$2,008)||$575 ($217-$1,233)||$89 ($50-$132)|
The cost of recurrence was estimated both overall (Figs. 4 and 5) and separately for each age group (Table 2). The cost of recurrence was substantially lower when the cost of death was excluded. Furthermore, the costs varied across age groups. In the analyses excluding cost of death, because there were no early retirement estimates for women aged 30 to 49 years, the cost was higher for women aged 50 to 69 years. The overall cost was estimated to be $240 million. When death was included, the total recurrence cost was largest in the 30- to 49-year-old age group. In economic terms, avoidance of breast cancer recurrence was associated with savings on the order of $1.7 billion. More than half of this was from events prevented in the 30- to 49-year-old age group.
The base case was adjusted to reflect more conservative and more aggressive assumptions regarding the potential patients who would benefit from therapy. In the low case, the recurrences prevented were fewer (2267; 95% CI, 1288-3188). In the high case, the values were larger (3087; 95% CI, 1748-4356). Finally, when the time horizon was reduced to 4 years, the number of recurrences prevented was smaller (2340; 95% CI, 1331-3285). When the time horizon was extended to 6 years, the results were notably larger (3198; 95% CI, 1802-4502). The corresponding cost estimates for these different scenarios were proportional to the number of recurrences (data not shown).
In the United States, approximately 31,200 women will be newly diagnosed with stage I to III, HER2+ breast cancer each year, based on 2005 data. In the absence of trastuzumab, about 7300 of these women would have a recurrence event even with optimal chemo- and hormonal therapy. Of these, about 2800 women might have been prevented from a recurrence event if trastuzumab had been added to standard adjuvant treatment. This analysis uses 2005 as a proxy; most other years would be expected to be similar, because changes in incidence are relatively slow and would take some time to change these results substantively. Viewed a different way, over the next 25 years, HER2-targeted therapies would have the potential to prevent >55,000 cases of recurrent disease, assuming no substantive changes in the incidence of HER2+ disease. Given the potential for shortages in physician supply in oncology, fewer recurrences will help to move potentially scarce future resources to other patients.42, 43
The corollary to findings is that there are still many ways to improve outcomes, because there are approximately 4500 women who will have recurrences within 5 years despite optimal adjuvant therapy. This translates to approximately 90,000 women over the next 25 years for whom additional improvements in HER2 treatment are needed. Also, as with any therapy, there are risks to treatment. Cardiac adverse events represent the most significant complication associated with trastuzumab, although many of them are reversible to some degree. However, the 3:1 recurrence-to-cardiac event ratio suggests a high benefit-to-risk ratio given the potential improvement of the associated decline in cardiac functioning.
Furthermore, there is some evidence that newer regimens may have lower rates of cardiac adverse events than seen in studies B-31 and N9831.6, 7, 27 In particular, in the Breast Cancer International Research Group study evaluated the efficacy and safety of 52 weeks of Herceptin administered as a component of both an anthracycline-based (doxorubicin/cyclophosphamide/taxane/trastuzumab) and a non– anthracycline-based (cyclophosphamide/taxane/trastuzumab) adjuvant regimen in comparison to doxorubicin/cyclophosphamide/taxane for women with early stage HER2+ lymph node-positive or lymph node-negative cancer. Because the Breast Cancer International Research Group study has shown that cyclophosphamide/taxane/trastuzumab and doxorubicin/cyclophosphamide/taxane/trastuzumab are not statistically significantly different in terms of overall survival at 5 years, these simulation results may reflect a conservative estimate of the cardiac adverse event risk from adjuvant therapy.44
We also used this model to explore economic consequences. Avoidance of breast cancer recurrence can lead to substantial economic savings. Depending on whether the productivity gains (ie, the economic value) from lost years of life are included, this estimate ranges from $240 million to $1.7 billion over the lifetimes of each 1-year cohort of patients. It should be cautioned that appropriately valuing the indirect cost of cancer is not a simple undertaking.45, 46 The cost of cancer is very broad and includes some difficult-to-measure aspects; therefore, the inference from our analyses should not be that these numbers represent the definitive estimate of the value of preventing recurrences. Rather, what is clear from exploring the cost data is that there are significant costs of breast cancer recurrence, most of which accrue from lost years of life. Accordingly, this is the actual value of treatments that can prevent recurrences, particularly those that are associated with high mortality risk, such as HER2+ breast cancer.2-4
Because our purpose was to estimate the number of recurrences and to explore their long-term economic implications, we did not address the cost of treatment. However, at least 3 cost-effectiveness analyses have done so.9, 11, 12 Assuming no missed doses or dose reductions, and including costs for time missed from work, Kurian et al12 estimated, per patient, that 1 year of trastuzumab and cardiac monitoring would cost $72,285 in 2008 dollars, and would yield approximately 1.4 quality-adjusted life-years (QALYs). This would translate to approximately $2.1 billion in total costs and 41,000 QALYs for the women treated in each year. Although these numbers are not directly relevant to our study, they should provide some context for the estimates presented.
As with any model, there are limitations based on the data and assumptions used to construct the base case analyses. In general, we tried to remain conservative (underestimate) where possible. The efficacy of trastuzumab was based on an intent-to-treat analysis. However, crossover from placebo to trastuzumab occurred in 12% of patients, which suggests that trastuzumab efficacy may be underestimated. Also, trastuzumab may reduce recurrences for >5 years, but this is not considered in the model except for the sensitivity analyses. Although the base case does not assume that every woman would be treated, and we conducted scenario analyses to explore alternative base case definitions, it may be that the base case does not capture the current state of treatment. Finally, effectiveness in the real world, may be very different from trial experience, both because healthier patients may have been selected for the trials, and because patients may not adhere as well outside of the trials.
In terms of the economic impact, which excluded the value of lost years of life, we did not include estimates of caregiver costs, and we did not value early retirement in our youngest age group. In the estimate that included lost years of life, we applied no costs to the subset of patients who recurred but did not die within 5 years. This is likely to be conservative, because even long-term survivors are likely to incur some costs, as evidenced by research showing that cancer survivorship is associated with risk of unemployment.47 Furthermore, because we assigned the lowest relevant value for productivity losses in each age group, these are likely to be underestimated.
Although we incorporated uncertainty in our inputs into the model, because our model represents 1 year, we did not address potential changes in the population dynamics. The relationship between HER2 status and ER and PR status may be different by age, or may change over time. For example, there is some indication that recent decreases in breast cancer incidence have been observed in proportionally greater numbers in strongly hormone receptor-positive patients and this trend may not be as strong in the HER2+ population.48, 49 Furthermore, although this model is based on current trial results, there is likely to be improvement in outcomes over time as the treatment of HER2+ tumors evolves. As the SEER program begins to capture HER2 status, and other changes take place over time, this model can be updated to provide more accurate information on the societal implications of treatment.
In terms of interpreting the economic measures, it should be noted that the model does not address any costs of adjuvant treatment, including the costs of treating cardiac (or any other) adverse events. Because of this, these economic results should not be considered to be a net gain for society. Rather, the estimate represents the value of preventing recurrences as a consequence of treatment, regardless of the cost of treatment. From the societal perspective, these savings, realized in a subset of HER2+ patients, could be envisioned as helping to offset costs in the larger population, including screening, diagnosis, testing, treatment, and adverse events. It should be emphasized, however, that because of the nature of the costs considered in this model, the value of preventing recurrences does not directly result in tangible savings that can be immediately invested in the healthcare system.
Regardless of the limitations of this simulation model, the policy implications are clear. The burden of HER2+ disease can be significantly ameliorated with existing targeted therapies. Although this is important, there is also room for more improvement. Consequently, targeting HER2+ tumors in the adjuvant setting could prevent a significant number of women from recurrence events, with important clinical, humanistic, and economic consequences for patients, physicians, payers, and society.
CONFLICT OF INTEREST DISCLOSURES
This research was funded by Genentech. Drs. Brammer and Lalla are employees of Genentech.
- 6Phase III trial comparing AC-T with AC-TH and with TCH in the adjuvant treatment of HER2 positive early breast cancer patients: second interim efficacy analysis. Paper presented at: 29th Annual San Antonio Breast Cancer Symposium; December 14-17, 2006; San Antonio, Texas., , , et al.
- 13Modern Epidemiology, 3rd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 2008.
- 15Surveillance Research Program. SEER*Stat Software. Bethesda, MD: National Cancer Institute; 2008.
- 24Annual Estimates of the Population by Five-Year Age Groups and Sex for the United States: April 1, 2000 to July 1, 2006 (NC-EST2006-01). Washington, DC: US Census Bureau, Population Division; 2007.
- 25Human epidermal growth factor receptors 2 (HER2) testing and treatment patterns among newly diagnosed breast cancer patients. Paper presented at: American Society of Clinical Oncology Meeting; September 5-7, 2008; Washington, DC.,
- 27Updated results of the combined analysis of NCCTG N9831 and NSABP B-31: adjuvant chemotherapy with or without trastuzumab (H) in patients with HER2-positive breast cancer. Paper presented at: American Society of Clinical Oncology Meeting; June 1-5, 2007; Chicago, Illinois., , , et al.
- 28Five year update of cardiac dysfunction in NSABP B-31: a randomized trial of AC to paclitaxel versus AC to paclitaxel with trastuzumab in HER2-positive, lymph node-positive, operable breast cancer. Paper presented at: American Society of Clinical Oncology Meeting; June 1-5, 2007; Chicago, Illinois., , , et al.
- 29Assessment of cardiac dysfunction in a randomized trial comparing doxorubicin and cyclophosphamide followed by paclitaxel, with or without trastuzumab as adjuvant therapy in lymph node-positive, human epidermal growth factor receptor 2-overexpressing breast cancer: NSABP B-31. J Clin Oncol. 2005; 23: 7811-7819., , , et al.
- 31Consumer Price Index. Series ID: CUSR000SAM. Washington, DC: US Department of Labor, Bureau of Labor Statistics; 2008.
- 35May 2007 National Occupational Employment and Wage Estimates. Washington, DC: US Department of Labor, Bureau of Labor Statistics; 2008.
- 40National Comprehensive Cancer Network (NCCN). Breast cancer screening and diagnosis guidelines. NCCN Clinical Practice Guidelines in Oncology v.2.2008. Fort Washington, PA: NCCN; 2008. Available at: http:/www.nccn.org/professionals/physician_gls/PDF/breast-screening.pdf.
- 41CooperH, HedgesLV, ValentineJC, eds. The Handbook of Research Synthesis and Meta-Analysis. 2nd ed. New York, NY: Russell Sage Foundation; 2009.
- 44Phase III randomized trial comparing doxorubicin and cyclophosphamide followed by docetaxel (AC -> T) with doxorubicin and cyclophosphamide followed by docetaxel and trastuzumab (AC -> TH) with docetaxel, carboplatin and trastuzumab (TCH) in HER2 neu positive early breast cancer patients: BCIRG 006 study. Paper presented at: San Antonio Breast Cancer Symposium; December 9-13, 2009; San Antonio, Texas., , , et al.