|Home | About | Journals | Submit | Contact Us | Français|
Women with BRCA1 or BRCA2 (BRCA1/2) mutations must choose between prophylactic surgeries and screening to manage their high risks of breast and ovarian cancer, comparing options in terms of cancer incidence, survival, and quality of life. A clinical decision tool could guide these complex choices.
We built a Monte Carlo model for BRCA1/2 mutation carriers, simulating breast screening with annual mammography plus magnetic resonance imaging (MRI) from ages 25 to 69 years and prophylactic mastectomy (PM) and/or prophylactic oophorectomy (PO) at various ages. Modeled outcomes were cancer incidence, tumor features that shape treatment recommendations, overall survival, and cause-specific mortality. We adapted the model into an online tool to support shared decision making.
We compared strategies on cancer incidence and survival to age 70 years; for example, PO plus PM at age 25 years optimizes both outcomes (incidence, 4% to 11%; survival, 80% to 83%), whereas PO at age 40 years plus MRI screening offers less effective prevention, yet similar survival (incidence, 36% to 57%; survival, 74% to 80%). To characterize patients' treatment and survivorship experiences, we reported the tumor features and treatments associated with risk-reducing interventions; for example, in most BRCA2 mutation carriers (81%), MRI screening diagnoses stage I, hormone receptor-positive breast cancers, which may not require chemotherapy.
Cancer risk-reducing options for BRCA1/2 mutation carriers vary in their impact on cancer incidence, recommended treatments, quality of life, and survival. To guide decisions informed by multiple health outcomes, we provide an online tool for joint use by patients with their physicians (http://brcatool.stanford.edu).
Women with BRCA1 and BRCA2 (BRCA1/2) mutations face substantially elevated lifetime risks of developing breast and ovarian cancer.1,2 The last decade of research with this high-risk population has identified preventive strategies that save lives, notably early bilateral salpingo-oophorectomy,3–6 and screening regimens incorporating magnetic resonance imaging (MRI) that diagnose most breast cancers early.7–12 Although randomized trials have not been performed, studies have reported on breast and ovarian cancer incidence, tumor biologic and clinical features, and quality of life with prophylactic and surveillance interventions for BRCA1/2 mutation carriers.1,2,13–23
Despite substantial progress in managing the cancer risks owing to a BRCA1/2 mutation, patients and their physicians struggle with choices about interventions, such as whether to replace breast screening with bilateral prophylactic mastectomy (PM) and when to pursue PM and/or prophylactic bilateral salpingo-oophorectomy (PO). Evidence-based practice guidelines recommend PO by age 40 years, but advise physicians and patients to discuss the options of PM versus MRI-based breast screening.24 Guiding decisions about these interventions is difficult, because no trials have compared them directly. We and others have used decision analysis to compare screening and prophylactic surgery in terms of survival and cost-effectiveness.25–33 However, prior studies have not fully characterized the patient experience with different interventions—for example, the likelihood that a woman who chooses breast screening will develop a cancer requiring adjuvant chemotherapy—although cancer treatments significantly impact quality of life and survivorship34–36 and may inform choices between risk-reduction strategies. Moreover, there is no practical way to compare multiple clinically relevant options, such as immediate PM and PO versus screening plus immediate PO and delayed PM, for an individual patient in real time.
We adapted a previously developed Monte Carlo simulation model to compare breast and ovarian cancer incidence, tumor prognostic features, recommended treatments, overall survival, and cause-specific mortality for BRCA1/2 mutation carriers. We translated this model into an online clinical decision support tool, enabling personalized cancer risk management for women with BRCA1/2 mutations.
We developed a computer simulation model that integrates published data (Table 1) to estimate breast and ovarian cancer incidence and tumor prognostic features, probability of survival to ages 70 and 80 years, and causes of death for women with a BRCA1 or BRCA2 mutation, starting from age 25 years.29 Risk-reducing interventions were modeled alone and in combination, at ages specified by practice guidelines24,51: breast screening consisting of mammography plus MRI started at age 25 years and continued annually to age 69 years, and PM and PO were modeled at ages 25, 40, and 50 years.
We initially built and validated a Monte Carlo model to analyze the effects of screening and treatment on the outcomes of patients with breast cancer, working within the Cancer Intervention and Surveillance Modeling Network.52,53 We then modified this model to simulate breast and ovarian cancer incidence, tumor characteristics, and prognosis under treatments recommended by practice guidelines (specific to tumor stage, size, and hormone receptors),1,13,18,20,46,54–56 and the performance of screening mammography and MRI,9,10,44,45 for BRCA1/2 mutation carriers.29,31 In sensitivity analyses, we varied parameters about which significant uncertainty exists, within CIs specified by published literature or more broadly (Table 1).
The model simulates life histories of a 1980 birth cohort of 1,000,000 female BRCA1/2 mutation carriers from age 25 years until age 100 years or death. We extrapolated BRCA1/2-associated cancer risks from meta-analyses.1,2 Because approximately 30% of BRCA1/2 mutation carriers undergo PO at a mean age of 45 years,42,57,58 and because premenopausal PO reduces breast cancer incidence by approximately 50%,4,5,40–42 we assumed that the incidence results from meta-analyses were affected by an unreported PO use of approximately 30%. To estimate breast cancer incidence in the absence of PO, we back-calculated the effect of a 50% reduction in subsequent breast cancer risk for 30% of the cohort as a result of PO performed by age 45 years.
We assumed a tumor grade distribution for BRCA1/2-associated breast tumors consistent with published reports; estrogen receptor (ER) expression was modeled as a function of grade and mutation.15,23 Tumor size and stage at clinical diagnosis were estimated as a function of grade, informed by the Surveillance Epidemiology and End Results (SEER) registry from the years before population-wide mammographic screening (1975 to 1981).52,53 We assumed that BRCA1/2-associated breast and ovarian cancers are treated with standard therapies based on tumor size, stage, grade, ER expression, and other prognostic features and that treatment efficacy and prognosis equal those in the general population.20,24,46,54,56,59,60 We assumed a median size detection threshold of 0.5 cm for MRI and that MRI has 85% sensitivity.9,10,12,44 For mammography, we assumed a distribution of detection thresholds, with a median of 1 cm, and that a proportion of tumors are undetectable until they grow to 5 cm; we derived this proportion separately according to age and BRCA mutation, calibrating to published mammographic detection rates in this population (Table 1).9,10,44,45
We assumed that PM reduces overall breast cancer incidence by 90%18,19 and PO reduces annual ovarian cancer incidence by 80%.5,6,42 We modeled the impact of PO as the annual probability of breast cancer detection at age i (Pi,PO) as Pi,PO = Pαi,NoPO, where i ≥ age at PO and α is the hazard ratio (HR). For ages less than 40 years, HR was 0.36, for ages 40 to 49 years, HR was 0.5, and for ages ≥ 50 years, HR was 1.0 (Table 1).6,40 We assumed that the breast cancer risk reduction from PM at any age, and from PO before age 50 years, persists indefinitely.41 We assumed that PO has no effect on breast cancer stage or ER expression.
We did not explicitly model use of menopausal hormone therapy, oral contraceptive pills, or prophylactic tamoxifen or raloxifene, given uncertainty about their effects on cancer risks for BRCA1/2 mutation carriers,61–69 but we did explore their impact through sensitivity analyses of the effect of PO on breast cancer incidence. After PO performed at age less than 50 years, we conservatively assumed that cardiovascular disease doubles47 and that osteoporotic hip fracture and dementia increase by 50%.48–50 We computed other-cause mortality using the Berkeley Mortality database,70 adjusting death rates from breast and ovarian cancer, cardiovascular disease, dementia, and hip fracture according to assumed relative risks (Table 1).71,72 Previously, we found that reducing the incidence of cardiovascular disease, osteoporosis, and dementia (as would be anticipated with menopausal hormone therapy) had only a small (2% to 3%) impact on other-cause mortality,29 so those sensitivity analyses were not repeated here.
We collaborated with software developers (L.C., P.R., M.S.) to build a model interface suitable for use as an online decision support tool. After obtaining human subjects approval from the Stanford University institutional review board and informed consent, we initiated pilot-testing of the tool with BRCA1/2 mutation carriers accrued through the Stanford Clinical Cancer Genetics Program and the Facing Our Risk of Cancer Empowered advocacy group73 and with clinicians from Stanford University Hospital and surrounding community practices.
In prior work, we reported survival with combinations of PM, PO, and screening29; we now expand our comparison of intervention strategies along the dual axes of incidence and survival. Figure 1 plots the probability of developing breast or ovarian cancer (combined) versus the probability of survival to age 70 years for BRCA1/2 mutation carriers. For BRCA1 mutation carriers, survival is maximized (80%) and cancer incidence minimized (11%) by the combination of PM and PO (PM + PO) at age 25 years. By comparison, PM + PO at age 40 years yields 3% lower survival probability (77%) with a 21% increase in incidence (32%), whereas breast screening plus PO at age 40 years, without PM, yields 6% lower survival probability (74%) with a 46% increase in incidence (57%). For BRCA2 mutation carriers, PM + PO at age 25 years maximizes survival (83%) and minimizes incidence (4%); by comparison, PM + PO at age 40 years reduces survival probability by 1% (82%), with an 11% increase in incidence (15%), whereas breast screening plus PO at age 40, without PM, yields 3% lower survival probability (80%) with a 22% increase in incidence (36%). For BRCA1/2 mutation carriers, PM, PO, and PM + PO reduce breast cancer incidence, with a magnitude inversely related to age at surgery. The same is true for PO and ovarian cancer incidence, whereas screening slightly increases breast cancer incidence (Table 2). Results were most sensitive to assumptions about the efficacy of PO for breast cancer risk reduction and BRCA1/2 mutation penetrance (Appendix Table A1, online only).
To characterize patients' treatment and survivorship experiences, we compared the outcomes of risk-reducing strategies in terms of breast tumor features at diagnosis: stage, size, ER expression, and recommended systemic treatments according to guidelines of the National Comprehensive Cancer Network.54 Without risk-reducing interventions, BRCA1 mutation carriers are most likely to develop an ER-negative tumor involving axillary lymph nodes (34%), corresponding to American Joint Committee on Cancer stage II to III74; next most likely diagnoses are a lymph node-negative, ER-negative tumor larger than 2 cm in size (mostly stage II; 16%) or a lymph node-positive, ER-positive tumor (stage II to III; 16%). These three most common scenarios require adjuvant chemotherapy; ER-positive tumors also require adjuvant hormonal therapy.54 Adding MRI-based screening shifts the tumor stage and ER distribution: screened BRCA1 mutation carriers have the greatest probability of a stage I, ER-negative tumor (42%, usually requiring chemotherapy), followed by a stage I ER-positive tumor (25%, requiring hormonal therapy and possibly chemotherapy) and a stage II to III, lymph node-positive, ER-negative tumor (14%, requiring chemotherapy; Table 3).54
Without intervention, BRCA2 mutation carriers are most likely to develop a stage II to III, lymph node-positive, ER-positive tumor (36%, requiring chemotherapy and hormonal therapy), followed by a stage II, lymph node-negative, ER-positive tumor larger than 2 cm (20%, requiring hormonal therapy and probably chemotherapy) or a stage I, lymph node-negative, ER-positive tumor (18%, requiring hormonal therapy and possibly chemotherapy; Table 3).54 With MRI-based screening, BRCA2 mutation carriers are most likely to have a stage I, ER-positive cancer (55%, requiring hormonal therapy and possibly chemotherapy), followed by a stage II to III lymph node-positive, ER-positive tumor (14%, requiring chemotherapy and hormonal therapy), and a stage I, ER-negative tumor (13%, usually requiring chemotherapy).54 Results were most affected by our assumptions about the sensitivity of screening MRI (Appendix Table A1). Figure 2 presents risks of developing breast cancer to age 70 years, stratified by likely systemic treatments,54 for BRCA1/2 mutation carriers under various strategies.
To inform an age- and mutation-specific decision tool, we performed 2,130 model runs, considering cancer-free women with BRCA1 and BRCA2 mutations, in the age intervals of 25 to 29, 30 to 34, 35 to 39, 40 to 44, 45 to 49, 50 to 54, 55 to 59, 60 to 64, and 65 to 69 years. Because each run simulates the individual histories of 1,000,000 women, our estimates incorporate outcomes of more than 1,000,000,000 women, stratified by age at BRCA1/2 mutation testing and risk-reducing intervention. Results of all runs are presented in the online decision tool.
We developed an online user interface with a vertical bar-graph display, adapting figures from our prior survival analysis29 and from our current work on cancer incidence, features, and treatments (Fig 2). We selected a bar-graph format by analogy to the successful Adjuvant! model for breast cancer treatment.75,76 User-selected variables are patient characteristics (age and mutation) and intervention strategies (annual screening mammogram, annual screening MRI, PM, and PO, alone or in combination at different ages). Each vertical bar represents a selected strategy showing the following outcomes: survival probability to age 70 years, causes of death, breast and ovarian cancer incidence, and breast tumor features influencing treatment recommendations (stage and hormone receptor expression). Of six vertical bars, the middle four can be customized to report specific strategies; for comparison, the left-most anchoring bar shows outcomes if no risk-reducing interventions are undertaken and the right-most anchoring bar shows the outcomes of an age-matched woman with no BRCA1/2 mutation. The four user-customized bars can be ranked by survival, and the display can be printed. Figure 3 presents the main screen of the tool; the user can choose to display survival and causes of death only (Fig 3A) or to include features and guideline-recommended treatments of breast cancers (Fig 3B). Additional screens include (1) an introduction, specifying the intended use conditions under physician supervision, the intended population, and the modeling assumptions; (2) a glossary; (3) contact information; and (4) publication links. The decision tool is available online.77
We built a simulation model to estimate and compare multiple health outcomes for BRCA1/2 mutation carriers under various cancer risk-reduction strategies and converted it into an online decision tool for use by physicians and patients. We found that early prophylactic mastectomy and salpingo-oophorectomy most effectively prevent cancer, but alternatives that reduce cancer incidence far less substantially can offer comparable survival. MRI-based breast screening yields this benefit through a diagnostic stage shift, increasing the proportion of stage I tumors from 18% to 22% up to 67% to 68%; treatment recommendations vary by hormone receptor expression, which is correlated with the type of BRCA mutation. A BRCA2 mutation carrier who elects MRI screening may well escape a recommendation for adjuvant chemotherapy, because 81% of breast cancers diagnosed in this setting are smaller than 2 cm, node-negative, and hormone receptor-positive. The online decision tool77 enables direct comparison of many possible strategies for an individual patient, combining various screening methods with prophylactic surgeries undertaken at different ages and weighing their impact on cancer incidence, treatment experiences, and survival.
Decisions about cancer risk reduction are complex and highly personal. For most women in the population, the greatest challenge is estimating risk accurately.78 Imprecision in communicating risk-benefit ratios may contribute to under-utilization of effective strategies such as chemoprevention.79 For BRCA1/2 mutation carriers, cancer risks are higher and better defined, driving greater uptake of prophylactic surgeries and intensive surveillance.57,58 Nonetheless, a significant number of women with BRCA1/2 mutations never develop cancer, given variations in penetrance which are incompletely understood80–82; removing organs at risk thus remains a gamble. Use of prophylactic surgeries varies by country, age, and prior cancer diagnosis, and aspects of personal experience such as family cancer history and parity play a role.56,57,83,84 Prior decision analyses have assigned a set value to life after prophylactic surgery or with screening and reported results in quality-adjusted life years.25,27,31,85,86 Given our experience of substantial variation in patient preferences, we elected against ranking health states under different risk-reducing interventions. Instead, our estimates across multiple outcomes aim to guide patients in optimizing their quality of life, depending on their individual values: for one woman this could entail retaining her breasts for decades, with eventual diagnosis of an early-stage breast cancer that might require chemotherapy; for another, this could entail maximal cancer prevention by removing her breasts and ovaries early in life.
With rapid growth in therapeutic and diagnostic technology, decision aids are increasingly used in oncology practice. They synthesize diverse data sources and integrate comparisons across disparate (and often conflicting) scales of benefit, such as efficacy, toxicity, and cost. Trials have demonstrated improvements in decisional conflict and satisfaction with the use of decision aids in breast, colorectal, and thoracic oncology.76,87–92 Decision aids weigh the absolute magnitude of an intervention's benefit against competing risks and may align choices more closely with expected therapeutic gains.93 The online format of our decision tool facilitates access and personalization; results are customized for a patient's age, allowing women to revisit their decisions over time should their health status, life circumstances, or priorities change. Additionally, the online tool can be readily adapted to accommodate new data from emerging studies. No decision aid can replace any aspect of the physician–patient relationship; our tool aims rather to channel the discussion toward choices that more fully realize a patient's personal preferences.
Our work has some limitations. The results of any simulation model depend on its assumptions. We initially developed and validated our model with SEER registry data, modeling tumor growth and ER expression as a function of grade52,53,94; we subsequently applied this same strategy to BRCA1/2-associated breast cancers, using appropriate tumor grade and ER distributions.13,15,23,55,95,96 Although justified by reports that patients with breast cancer with and without BRCA1/2 mutations have similar outcomes,20,59,60,97 this fundamental approach is difficult to validate. The model lacks data on breast tumor progesterone receptor and human epidermal growth factor receptor 2 expression, given their absence in SEER; because human epidermal growth factor receptor 2 overexpression is rare in BRCA1/2-associated tumors,13,55,95,96 this limitation is unlikely to change our conclusions. An additional limitation is our use of average BRCA1/2 mutation penetrance estimates, derived from meta-analyses1,2; our model does not incorporate factors that may mediate individual risk variation, such as birth cohort, family history, lifestyle and environmental exposures, or single-nucleotide polymorphisms in other genes.81,82,98,99 We did not assign a separate prognostic category to carcinoma in situ or consider emerging treatments such as poly (ADP-ribose) polymerase inhibitors. We varied input parameters widely in sensitivity analyses and found that our assumptions about BRCA1/2-associated cancer risks, the sensitivity of MRI, and the effect of PO on breast cancer incidence were most influential. If MRI-based breast screening detects preinvasive cancers,9,45,100 which have optimal survival and no requirement for chemotherapy,74 or if BRCA1/2 mutation-targeted cancer therapies improve survival with few adverse effects,101–104 then breast screening may provide a better outcome than we estimate; conversely, if mutation penetrance is higher or cancer prognosis worse than we estimate, prophylactic surgeries would appear more favorable. The online decision tool focuses on cancer-free women; it does not report second primary cancer risks, mortality from a prior cancer, or benefit from a procedure performed in the past. We have not measured the tool's impact on decision outcomes in a clinical trial, but pilot-testing among 60 patients and providers yielded high rankings on clinical relevance and ease of use, with a full analysis underway. Future work is warranted to address these limitations of the model and decision tool.
We calculated cancer incidence, tumor prognostic features that influence treatment and quality of life, overall survival, and cause-specific mortality under many possible risk-reduction strategies for BRCA1/2 mutation carriers. We customized these results by age and BRCA mutation and adapted them into an online tool to support joint decision making by patients and physicians. By characterizing the multiple health outcomes associated with cancer risk-reduction options, our decision tool aims to clarify a patient's priorities and guide choices that preserve them.
We thank Bronislava M. Sigal, PhD, who made important contributions to earlier versions of the simulation model code and consulted on modifications made for this study, and the Facing Our Risk of Cancer Empowered patient advocacy group, who assisted in pilot-testing the online decision tool.
|Parameter||BC Incidence (%)||BC Stage (% local)||BC ER (% positive)||OC Incidence (%)|
|BRCA1 mutation carriers|
|BC by 70 years, BRCA1||0.65||0.47||0.85||0.65||0.47||0.85||0.65||0.47||0.85||0.65||0.47||0.85|
|No S* or PM or PO||68||53||89||42||42||42||34||34||34||38||38||38|
|S*, no PM or PO||70||55||90||79||75||80||34||34||34||38||38||38|
|S* and PO, no PM†||52||39||77||77||74||78||34||34||34||10||10||10|
|S* and PO and PM†||25||19||38||81||80||82||37||37||37||10||10||10|
|OC by 70 years, BRCA1||0.39||0.39||0.46||0.39||0.39||0.46||0.39||0.39||0.46||0.39||0.39||0.46|
|No S* or PM or PO||68||68||68||42||42||42||34||34||34||38||38||46|
|S*, no PM or PO||70||70||70||79||79||79||34||34||34||38||38||46|
|S* and PO, no PM†||52||52||52||77||77||77||34||34||34||10||10||12|
|S* and PO and PM†||25||25||25||81||81||81||37||37||37||10||10||12|
|No S* or PM or PO||68||68||68||42||42||42||34||34||34||38||38||38|
|S*, no PM or PO||70||69||71||79||59||82||34||34||34||38||38||38|
|S* and PO, no PM†||52||50||54||77||58||81||34||34||34||10||10||10|
|S* and PO and PM†||25||23||24||81||68||86||37||37||37||10||10||10|
|HR for BC after PO||0.5§||0.1||1||0.5§||0.1||1||0.5§||0.1||1||0.5§||0.1||1|
|No S* or PM or PO||68||68||68||42||42||42||34||34||34||38||38||38|
|S*, no PM or PO||70||70||70||79||79||79||34||34||34||38||38||38|
|S* and PO, no PM†||52||27||71||77||77||76||34||35||34||10||10||10|
|S* and PO and PM†||25||23||27||81||84||80||37||37||37||10||10||10|
|BRCA2 mutation carriers|
|BC by 70 years, BRCA2||0.45||0.4||0.85||0.45||0.4||0.85||0.45||0.4||0.85||0.45||0.4||0.85|
|No S* or PM or PO||50||45||86||47||47||47||78||78||78||11||11||11|
|S*, no PM or PO||53||48||88||80||77||83||78||78||78||11||11||11|
|S* and PO, no PM†||34||31||72||79||76||80||78||78||78||2||2||2|
|S* and PO and PM†||13||12||34||83||82||86||81||80||80||2||2||2|
|OC by 70 years, BRCA2||0.11||0.11||0.27||0.11||0.11||0.27||0.11||0.11||0.27||0.11||0.11||0.27|
|No S* or PM or PO||50||50||50||47||47||47||78||78||78||11||11||27|
|S*, no PM or PO||53||53||53||80||80||80||78||78||78||11||11||27|
|S* and PO, no PM†||34||34||34||79||79||79||78||78||78||2||2||6|
|S* and PO and PM†||13||13||13||83||83||83||81||81||81||2||2||6|
|No S* or PM or PO||50||50||50||47||47||47||78||78||78||11||11||11|
|S*, no PM or PO||53||50||54||80||63||82||78||78||78||11||11||11|
|S* and PO, no PM†||34||32||36||79||62||81||78||78||78||2||2||2|
|S* and PO and PM†||13||13||13||83||71||86||81||80||80||2||2||2|
|HR for BC after PO||0.5§||0.1||1||0.5§||0.1||1||0.5§||0.1||1||0.5§||0.1||1|
|No S* or PM or PO||50||50||50||47||47||47||78||78||78||11||11||11|
|S*, no PM or PO||53||53||53||80||80||80||78||78||78||11||11||11|
|S* and PO, no PM†||34||15||52||79||79||78||78||79||78||2||2||2|
|S* and PO and PM†||13||12||15||83||86||80||81||81||80||2||2||2|
NOTE. Parameters, varied from base case to lower and upper cases, are BC and OC risks by age 70 years owing to BRCA1/2 mutations (BC by 70 years, OC by 70 years), MRI sensitivity for BC detection, and HR for BC development after premenopausal PO; interventions evaluated are S, PM, and/or PO.
Abbreviations: BC, breast cancer; ER, estrogen receptor; HR, hazard ratio; MRI, magnetic resonance imaging; OC, ovarian cancer; PM, prophylactic mastectomy; PO, prophylactic oophorectomy; S, screening.
See accompanying editorial on page 471
Supported by the National Institutes of Health (Grants No. R01 CA829040, U01 CA088248, and R01 CA66785 to S.K.P), the Robert Wood Johnson Foundation (Physician Faculty Scholars Award, Grant No. 64317 to A.W.K.), and a Stanford Cancer Institute Developmental Research Award.
Presented in part at the Robert Wood Johnson Foundation Physician Faculty Scholars Meeting, December 8-10, 2010, Fort Lauderdale, FL.
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
The author(s) indicated no potential conflicts of interest.
Conception and design: Allison W. Kurian, Lauren Clarke, Sylvia K. Plevritis
Financial support: Allison W. Kurian, Sylvia K. Plevritis
Provision of study materials or patients: Allison W. Kurian, Elizabeth A. Schackmann, Sylvia K. Plevritis
Collection and assembly of data: All authors
Data analysis and interpretation: Allison W. Kurian, Diego F. Munoz, Elizabeth A. Schackmann, Meredith A. Mills, Sylvia K. Plevritis
Manuscript writing: All authors
Final approval of manuscript: All authors