A Short History of Bernard Fisher’s Contributions to Randomized Clinical Trials

Abstract

Dr. Bernard Fisher (1918–2019) was an early proponent of evidence-based medicine using the mechanism of prospective, multicenter, randomized clinical trials to test biological and clinical hypotheses. In this article, I trace how his early scientific work in striving to understand the nature of cancer metastasis through animal experiments led to a new, testable, clinical hypothesis: that surgery to remove only the tumor and a small amount of tissue around it was as effective as the more disfiguring operations that were then the standard treatment. Fisher’s work with the National Surgical Adjuvant Breast and Bowel Project (NSABP) using large, randomized clinical trials to demonstrate the veracity of this hypothesis led to a new paradigm in which the emphasis was placed on how systemic therapies used at an early stage of disease could effectively eradicate breast cancer for many patients. This new therapeutic approach led to the successful development of new treatments, many of which are widely used today. Ultimately, the new paradigm led to successfully preventing breast cancer in women who were at high risk for the disease but who had not yet been diagnosed with the disease. Throughout his entire career, Fisher championed the use of large prospective, randomized clinical trials despite criticism from many in the medical community who strongly criticized his use of randomization as a mechanism for testing clinical hypotheses. The approach he and the NSABP employed is still considered to be the highest standard of evidence in conducting clinical studies.

“In God We Trust; All Others Must Provide Evidence”

– Bernard Fisher (1918–2019)

Introduction

On October 16, 2019, one of the truly remarkable physician scientists of the last hundred years, Dr. Bernard Fisher, passed away at age 101 in Pittsburgh. The worlds of cancer biology, cancer treatment and public health are indebted to his innovations, seminal contributions and tireless leadership over a period of more than 60 years.^1–6 The world of clinical trials is also indebted to his early advocacy of randomized clinical trials that tested seemingly simple clinical hypotheses that addressed very subtle and deep biological questions. In this article, I discuss how Fisher took information he learned from very innovative laboratory experiments that he led from the 1950s through the 1970s and transmogrified them into simple but elegant breast cancer clinical trials (Table 1).^7–9 As a result of his work both in the laboratory and in subsequent clinical trials, profound changes were made from the crude, often unsuccessful breast cancer treatments of his early years to the more sophisticated and effective therapies still widely used today. What particularly distinguishes Fisher’s work is that he was among the first clinician scientists to translate his own work from “the laboratory to the bedside” using a standard of evidence that is still employed, the randomized, multicenter clinical trial. He conducted most of this work under the cloud of the prevailing milieu in the surgical and general medical communities, which frowned upon doing less extensive surgeries to treat breast cancer and particularly frowned on the idea of randomly assigning patients to vastly different surgical, radiotherapeutic, and/or systemic therapies.^10–13

Table 1.

NSABP phase III breast cancer protocols initiated between 1971 and 1995

	Comparison therapies	Patient population	Date initiated	Number accrued
B-04	RM vs TM vs TM+XRT	Node positive / node negative	7/22/1971	1765
B-05	Placebo vs L-PAM	Node positive	9/22/1972	418 (380)*
B-06	TM vs Lump vs Lump+XRT	Node positive/node negative	4/8/1976	2163
B-07	L-PAM vs L-PAM + 5-FU	Node positive	2/3/1975	741
B-08	L-PAM + 5-FU vs L-PAM + MTX	Node positive	4/12/1976	737
B-09	L-PAM + 5-FU vs L-PAM + 5-FU + TAM	Node positive	1/1/1977	2697 (1891)*
B-10	L-PAM + 5FU vs C parvum Hydrocortisone	Node positive	5/1/1977	265
B-11	L-PAM + 5FU vs L-PAM + 5FU + A	Node positive, ER negative	6/1/1981	707
B-12	L-PAM + 5FU + A vs L-PAM + 5FU + A + TAM	Node positive, ER positive	6/1/1981	1106
B-13	M → F vs LV	Node negative, ER negative	8/1/1981	1116 (760)
B-14	Placebo vs TAM	Node negative, ER positive	1/4/1983	4127 (2892)*
B-14R	Placebo vs TAM	Node negative, ER positive from B-14 who were cancer free after 5 years of TAM therapy	4/7/1987	1172+
B-15	AC vs CMF × 3 vs CMF × 6	Node positive for all patients ≤ 49 years of age; node positive, PR negative for patients ≥ 50 years of age	10/1/1984	2338
B-16	TAM vs AC+TAM vs PFT (Last arm changed to PAFT after 6/10/1985)	Node positive for all patients 60–70 years of age; node positive, PR negative for patients 50–60 years of age	10/1/1984	1296
B-17	Lump + No further therapy vs Lump + XRT	DCIS	10/1/1985	1087 (818)*
B-18	Surgery → A 60 mg/m2×4 cycles + C 600 mg/m2 ×4 cycles vs A 60 mg/m2×4 cycles + C 600 mg/m2 ×4 cycles → Surgery	Node negative/node positive	10/17/1988	1523
B-19	M → F + LV vs CMF	Node negative, ER negative	10/17/1988	1095
B-20	TAM vs M → F + TAM vs CMF+TAM	Node negative, ER positive	6/1/1989	2363
B-21	XRT + Placebo vs XRT + TAM vs TAM alone	Node negative, tumors ≤ 1cm	6/1/1989	1009
B-22	A 60 mg/m2×4 cycles + C 600 mg/m2 ×4 cycles vs A 60 mg/m2×4 cycles + C 1200 mg/m2×2 cycles vs A 60 mg/m2×4 cycles + C 1200 mg/m2 ×4 cycles	Node positive	7/5/1989	2305
B-23	AC+Plac vs AC+TAM vs CMF+Plac vs CMF+TAM	Node negative, ER negative	5/12/1991	2008
B-24	Placebo vs TAM	DCIS	5/9/1991	1804
B-25	A 60 mg/m2×4 cycles + C 1200 mg/m2 ×4 cycles vs A 60 mg/m2×4 cycles + C 2400 mg/m2×2 cycles vs C 2400 mg/m2 ×4 cycles	Node positive	4/1/92	2548
B-27	PreOP AC vs Preop AC + PreOP Taxotere vs PreOP AC + PostOP Taxotere	Node negative/node positive	12/20/1995	2411
B-28	AC vs AC + Taxol	Node positive	8/1/1995	3060
P-1	Placebo vs TAM	Women without breast cancer who were at high risk to have breast cancer	6/1/1992	13388

^*( ) indicates the number of randomized patients in protocols that employed both randomization and registration of patients

⁺§ The patients in B-14R were accrued from B-14 patients who had been assigned to receive TAM and who were alive and event-free. These patients were rerandomized to TAM vs placebo. Thus, this was a test of 5 years of TAM vs longer.

Abbreviations

XRT: radiation therapy; RM: radical mastectomy; TM: total mastectomy (modified radical mastectomy); Lump: lumpectomy [sometimes referred to as segmental mastectomy]; L-PAM: L-phenylalanine mustard; 5-FU: 5-fluorouracil; MTX: Methotrexate; LV: leucovorin; A: Adriamycin; C: cyclophosphamide; AC, unless otherwise specified, refers to A 60 mg/m² × 4 cycles + C 600 mg/m² × 4 cycles; DCIS: ductal carcinoma in situ; TAM: tamoxifen; ER: Estrogen receptor; PR: progesterone receptor

Beginnings of the Fisher hypothesis: Laboratory studies and early clinical trials

The true beginnings of the “Fisher hypothesis” started in the 1950s. In the early years of his research, the young Fisher had many interests, among which were regeneration of liver tissue,^14–15 hypothermia¹⁶ and wound healing.¹⁷ Once his primary focus turned to the study of cancer metastasis, Fisher and his brother Edwin summarized the findings of dozens of predominantly animal experiments in this arena,^18–32 providing evidence that (1) circulating tumor cells were present at an early stage of disease;^18–20 (2) tumors can remain in a dormant stage for a long time and with an impetus can suddenly metastasize;^21–23 (3) cells can transmigrate the lymph nodes;^22,24–26 (4) the relationship between the tumor and its host (e.g., the immune system) is very complex and unpredictable;^{21, 27–31} (5) lymph nodes are of biological importance;^31,32 and (6) because of the complexities of the tumor-host relationship, surgical or other trauma can sometimes promote metastasis of cancer at any stage of the disease.^17,20,28 Many of these animal experiments involved cutting-edge microvascular surgical techniques on mice or rats under microscopic magnification.² At the time they were being conducted, many resulted in findings that were contrary to the then-current paradigm of cancer metastasis in human beings. Eventually, evidence from these experiments, along with preliminary evidence in human cancer studies, provided the basis for the Fisher hypothesis of cancer metastasis.

In 1956, while he was an active general surgeon also working on many animal experiments, Fisher got a call from his former mentor during his fellowship at the University of Pennsylvania, Dr. I.S. Ravdin, who was then the chairman of the Clinical Studies Panel of the Cancer Chemotherapy National Service Center at the NIH, and formerly had been a general in the army who had operated on President Eisenhower.³³ Along with Fisher and 22 other surgeons, Ravdin organized a group in 1957 called the Surgical Adjuvant Chemotherapy Breast Project and put its first patient on a protocol on April 4, 1958.⁸ The group’s name later changed to the National Surgical Adjuvant Breast Project and in 1977, with its initiation of colorectal trials, became known as the National Surgical Adjuvant Breast and Bowel Project (NSABP). In the early years of the group, Fisher became a leading contributor to implementing and reporting “first stage” clinical trials in women with breast cancer.^34–36 NSABP breast cancer trials before 1970 tested emerging chemotherapies that were typically given with radiation therapy after the patients underwent radical mastectomy. In 1967, Fisher was elected chair of the NSABP, and he subsequently moved the group’s headquarters to Pittsburgh in 1970.⁸ He was to remain chair of the group until 1994 and later became the director of science.

The first detailed communication about the Fisher hypothesis occurred in 1970.¹⁰ Around this time, a struggle in the surgical community about the worth of extensively disfiguring breast surgeries in the treatment of breast cancer was ongoing.^11,12 Previously, breast cancer was considered by most patients to be a death warrant and the disease was treated with extensive surgery (referred to as “en bloc radical resections”), from a layman’s perspective, casting a wide net [i.e., removing more tissue] to remove all of the cancer. Radiation therapy was often used along with these extensive surgeries. The prevailing treatment strategy seemed to imply that breast cancer emanated outward from the initial lesion in the breast and spread to other organs of the body in an orderly fashion, and therefore removal of extremely large amounts of tissue surrounding the lesion would be the optimal approach to treating the disease. Fisher’s animal experiments and observations of his patients’ outcomes indicated otherwise. But at the time, he didn’t possess enough evidence to appropriately challenge the prevailing paradigm. What was lacking, in Fisher’s view, was a scientific basis for the treatment of breast cancer which, until that time, had been largely based on observations made by Halsted in the 1890s.^37–39 Fisher summarized the state of affairs in the surgical treatment of breast cancer as follows: “It is almost tragic that such an important issue has, to the present, been decided by individual surgeons primarily as a result of emotionalism, information obtained from poorly carried out retrospective analyses of unsystematic case records and by comparisons of data made worthless because they were obtained from divergent series of patients.”⁹ He surmised that more extensive surgeries were not more effective in the treatment of breast cancer with respect to distant disease and survival and proposed that the best way to test his emerging theory was via randomized, multi-center clinical trials.

In 1971, shortly after the manuscript articulating Fisher’s initial biological hypothesis was published, accrual to NSABP Protocol B-04 began (Table 1). To address the question of the worth of extensive local therapy, this trial tested whether patients who had a modified total mastectomy (which preserved the pectoralis major muscle and axillary regional lymph nodes) would have similar outcomes to those who underwent the more aggressive and disfiguring radical mastectomy.^40,41 The results of the study indicated that patients who had the less extensive local therapy did as well as those with more extensive local therapy. Several follow-up papers later on confirmed the results.^42,43 Another NSABP trial (Protocol B-05) was implemented in the early 1970’s testing the worth of a single chemotherapy, L-PAM, versus a placebo, both given after patients had undergone surgery (mastectomy) and radiotherapy. Early results indicated that the chemotherapy significantly improved survival in breast cancer patients, but analyses based on more follow-up indicated that the survival benefit was not as robust as was originally reported^44,45 Based on the mathematical models of Skipper asserting that multiple chemotherapies were more effective than a single chemotherapy,⁴⁶ two other chemotherapy trials were implemented by the NSABP in the 1970s. One compared L-PAM to L-PAM + 5-FU [B-07] and another compared L-PAM + 5-FU to L-PAM + 5-FU + methotrexate [B-08] (Table 1).⁴⁷ By 1976, after indications that less invasive local therapy was equivalent to more invasive local therapy, Fisher and the NSABP were ready to take the ultimate step, and the group initiated Protocol B-06 to test whether lumpectomies (excising the tumor and only a very small amount of tissue around it so that the tumor “margins” were free of cancer) with or without radiation resulted in equivalent outcomes (distant disease and overall survival) compared to a group of patients undergoing total mastectomies. By 1977, the NSABP had also initiated its first trial (Protocol B-09) testing the efficacy of tamoxifen in axillary node-positive breast cancer patients.

The Fisher hypothesis in full flower

By 1980, Fisher’s ideas about changes to the breast cancer metastasis paradigm and how to clinically test tenets of that hypothesis were in full flower. In his Karnofsky lecture published that year,⁴⁷ meant to reach a wide clinical and scientific audience, he laid out explicit details of his alternative hypothesis (Table 2) regarding cancer metastasis, comparing it to the prevailing Halstedian hypothesis, in a side-by-side presentation. As noted above, each component of Fisher’s alternative hypothesis had been vetted through the dozens of animal experiments he led, along with preliminary observations made in human breast cancer studies. What followed from this multi-pronged hypothesis was an assertion that could be tested by clinical trial methodology: breast cancer is a systemic disease in which malignant cells are disseminated in an unorderly fashion and likely disseminated throughout the body before diagnosis, and hence, variations in local therapy are unlikely to substantially change survival. This hypothesis was tested by Protocols B-04 and B-06 and similar trials around the world. The Karnofsky lecture also provided a forum for Fisher to express his ideas about how new cancer therapies might be employed. For example, he noted that “a probe capable of defining responders and nonresponders to the therapy used provides direction to the next stage of investigation [and] determination of the reason for the difference of response. With such information, patients can be selected for a particular therapy.” This idea was fruitful, as it eventually led to the advent of the neoadjuvant trial, Protocol B-18, in 1988, comparing adjuvant therapy given before surgery to the usual adjuvant schedule of systemic therapy (adjuvant therapy after surgery).⁴⁸

Table 2:

Explicit contrasts between the Fisher and the Halstedian Hypotheses

Halsted	Fisher (Alternative)
Tumors spread in an orderly defined manner based on mechanical considerations	There is no orderly pattern of tumor cell dissemination.
Tumor cells traverse lymphatics to lymph nodes by direct extension supported en bloc	Tumor cells traverse lymphatics by embolization challenging the merit of en bloc dissection.
The positive lymph node is an indicator of tumor spread and is the instigator of disease	The positive lymph node is an indicator of a host-tumor relationship which permits development of metastases
Regional lymph nodes (RLNs) are barriers to the passage of tumor cells.	RLNs are ineffective as barriers to tumor cell spread.
RLNs are of anatomical importance.	RLNs are of biological importance.
The blood stream is of little significance as a route of tumor dissemination.	The blood stream is of considerable importance in tumor dissemination
A tumor is autonomous of its host	Complex host-tumor interrelationships affect every facet of the disease
Operable breast cancer is a local-regional disease	Operable breast cancer is a systemic disease.
The extent and nuances of surgery (?) are the dominant factors influencing patient outcome	Variations in local-regional therapy are unlikely to substantially affect survival.

Adapted from: Cancer Res, 1980, 40: 3863–74, Fisher B. Laboratory and clinical research in breast cancer - a personal adventure: The David A. Karnofsky memorial lecture, 1980, with permission from AACR.

Not long after the Karnofsky lecture, Fisher and colleagues published the results of Protocol B-09 reporting a significant benefit in axillary node positive breast cancer patients of tamoxifen in terms of survival and disease-free survival.⁴⁹ The successful use of tamoxifen led to later trials in patients with axillary node-negative tumors, then in patients with ductal carcinoma in situ, and subsequently in women who were at high risk of developing breast cancer. In 1985, the highly anticipated results of the B-06 trial were published, and they indicated that lumpectomies with radiation and total mastectomies were equivalent with respect to survival and distant disease survival.⁵⁰ Updated results of B-06 were later published as more follow-up information became available.^51–54 The results of the B-04 and B-06 trials were later validated using the results of similar trials from all over the world.^55,56 Fisher himself later reflected on how his theory withstood the test of time and the results of many subsequent trials.^57,58

From 1981 – 1994, Fisher and the NSABP had a remarkable period of productivity during which 16 randomized phase III breast cancer trials that, in total, accrued 39,820 women were initiated by the group (Protocols B-11 through B-25 and Protocol P-1).^59–78 Two other phase III trials (Protocols B-27⁷⁹ and B-28⁸⁰) were conceived by 1994 and started in 1995 (Table 1), and several breast cancer phase II pilot trials and sub-studies were initiated. Seven phase III colorectal trials were begun under the direction of Dr. Norman Wolmark. In those years the NSABP was turning from the old questions of how much local therapy was appropriate to the new paradigm of what and how adjuvant or neoadjuvant therapies were appropriate for the systemic treatment of women with breast cancer or of those who were at high risk of developing the disease but who had not been diagnosed with it. Several NSABP trials (B-11,⁵⁹ B-13^60,62 B-14,^61,62 B-17,⁶⁷ B-19,⁶⁴ B-20,⁶⁵ B-24,⁷⁴ and P-1^76–78) tested therapies that significantly improved outcomes. Many of these therapies subsequently became the standard of care in the treatment of breast cancer in certain patient populations. Particularly satisfying to Fisher were the results of the P-1 trial showing that tamoxifen benefited women who were at high risk of having breast cancer.⁷⁸ The P-1 trial represented the end of a journey that began with very disfiguring surgical procedures to treat breast cancer that only addressed local considerations of the disease to an era where prevention of the disease was based on systemic considerations. Furthermore, results from the phase III neoadjuvant study, B-18,⁶⁷ established that adjuvant therapy given before surgery had outcomes equivalent to those resulting from the usual adjuvant therapy after surgery and also could be used as a probe for further systemic therapy as necessary.

Philosophy

From a scientific view, Fisher emphasized that any hypothesis should be testable. He invoked the philosopher Claude Bernard, who in 1865 asserted that a hypothesis was of value only if it could be appropriately tested scientifically and otherwise “is a purveyor of rubbish.”^81,82 Fisher’s view was that his hypothesis “…needed to be tested because only with scientific confirmation would it be permissible for a new generation of surgeons to abandon the surgical heritage that had been passed on to them by William Halsted.”⁸² Fisher also recognized that the process of changing an entrenched paradigm could not be answered in a single trial; such questions would have to be addressed in sequences of trials. What he and his many clinical colleagues developed using that approach was an intricate tapestry of trials in breast cancer (Table 3) that were instrumental in changing the paradigm of the prevention and treatment of breast cancer and that could be applied to other cancers as well. The idea was disarmingly simple: if an experimental therapy showed superiority in a particular trial, then that therapy would be used in future trials either as (1) a control therapy for testing a new therapy or (2) for testing the successful therapy in a new patient population. An example of the use of the strategy in item (1) was the companion surgical/radiation trials B-04 and B-06. Examples of using a successful therapy in different patient populations were the sequences of trials involving tamoxifen. In NSABP Protocol B-09 involving patients who had axillary node-positive breast cancer, it was shown that, in addition to total mastectomy and axillary dissection, tamoxifen and chemotherapy was superior to chemotherapy alone in patient outcomes. Subsequently, the worth of tamoxifen was (1) tested in other node- positive populations with different chemotherapies [B-15, B-16]; (2) tested in node-negative, estrogen receptor-positive patient populations [B-14,B-20]; (3) tested in node-negative, estrogen receptor-negative populations [B-19, B-23]; (4) tested in a population with ductal carcinoma in situ [B-24]; and (5) tested in populations of women who were at high risk for developing breast cancer [P-1].

Table 3.

Examples of how NSABP companion protocols were used to test therapies

B-04 ➔ B-06 ➔ B-17 ➔ B-32* (assessing the consequences of doing less invasive local therapy)

B-05 ➔ B-07 ➔ B-11 ➔ B-15 ➔ B-16 ➔ B-22 ➔ B-25 ➔ B-28 ➔ B-30* (testing optimal types, doses and timing of chemotherapies in the treatment of stage II breast cancer patients)

B-09 ➔B-12 (testing the efficacy of endocrine therapy [tamoxifen] in treating axillary node positive patients)

B-13 ➔ B-19 ➔ B-23 (testing the worth of tamoxifen and chemotherapies for treating node-negative, ER-negative patients)

B-14 ➔ B-20 (testing the efficacy of endocrine therapy [tamoxifen] and chemotherapy in axillary node negative, ER-positive patients; later used in the development of the Oncotype DX score)

B-17 ➔ B-21➔B-24 ➔ B-35* ➔ B-39* (testing surgical, radiation and endocrine therapies in DCIS patients or patients with very small tumors)

B-14 ➔ P-1 (testing the efficacy of endocrine therapy [tamoxifen] in axillary node negative, ER+ patients and as a preventive agent in women at high-risk of developing breast cancer)

P-1 ➔ P-2* (testing the efficacy and later, equivalence of endocrine therapies [tamoxifen, raloxifene*] in preventing BC in women at high risk for such disease)

B-18➔B-27➔B-40*(testing the worth of neoadjuvant chemotherapy vs adjuvant chemotherapy and if different therapies pre-operatively, post-operatively or both had efficacy in the treatment of stage I & II breast cancer)

^*Protocols conceived after the period of Fisher’s NSABP chairmanship but part of a continuing series of studies

As Fisher himself always emphasized, he did not take his journey alone. Through academic institutions, community networks,⁸³ NCI staff and scientists,⁵ and an extremely dedicated headquarters staff and statistical/data management/IT groups, he was able to lead an effort to accrue patient populations that were diverse demographically and who were treated within a broad spectrum of cancer care. This diversity, along with the randomization procedures, the large cohorts being accrued and the relatively simple comparisons being made, meant that statistical inference from the study results was very generalizable. Most of the results in these NSABP studies were later validated via a worldwide effort by the Early Breast Cancer Trialists’ Collaborative Group.

How did Fisher lead such a broad spectrum of talented people? By his intelligence, imagination, and charisma, and because nobody worked harder than he did. He also had a way of telling stories to motivate people. In one case, after the P-1 trial came just short of its accrual goal after one year of initiation, Fisher reminisced that he seemed to always come up a little short (according to him, finishing second in his high school class, for example) but always striving to do better. This message rang true to those he led.

Despite advocating for clinical trials, Fisher’s view about them mirrored that of Winston Churchill’s on democracy. Churchill said, “Indeed it has been said that democracy is the worst form of government except for all those other forms that have been tried from time to time.” Fisher modified this adage by asserting that large, prospective, randomized clinical trials are the worst form of testing biological and therapeutic hypotheses except for all other methods that currently exist. Nevertheless, he saw clinical trials as a necessary mechanism for conclusively testing biological and clinical hypotheses in animal and human populations.

Controversy and aftermath

The most devastating period in Fisher’s career came after the 1991 finding by NSABP auditors that a University of Montreal surgeon had falsified data from his institution, St. Luc Hospital. Shortly after these falsifications were confirmed, Fisher abruptly halted accrual from St. Luc and reported the incident to the NCI, which informed the Office of Scientific Integrity (which shortly thereafter folded into the Office of Research Integrity, or ORI).⁸⁴ A joint ORI and NSABP audit of all St. Luc data found falsified data on 99 of the 1511 patients submitted by St. Luc to NSABP.⁸⁵ The falsified data involved manipulation of entry data so that patients who would have been ineligible for the clinical trial subsequently became eligible. During the audit, the involved NSABP staff were instructed by the ORI not to discuss the matter except on a “need to know basis”.⁸⁴ In early 1992, using information from the audit up to that time and with permission from the ORI, reanalyses of NSABP studies were presented to the NSABP executive committee and shortly thereafter to representatives of the ORI, NCI and NIH. The final results of the audits were published in the ORI newsletter in March 1993 and the Federal Register in July 1993.⁸⁴ Once the audits were complete and the full extent of the falsifications was known, NSABP began to reanalyze all 22 NSABP studies with St. Luc information, including eight studies which had no falsifications. Like earlier analyses, the new analyses showed that study conclusions were unchanged, both when all St. Luc patients were removed, or when only those patients with falsified data were removed. These results were expected because the studies involved were large, multicenter randomized clinical trials, and the St Luc contributions to accrual were typically relatively modest. Furthermore, given that it had been established that the treatment, follow-up and endpoint information had not been falsified then one issue with the removal of St. Luc patients going forward was that if the intention to treat principle was to be adhered to on future analyses then even ineligible patients should be included. Also, it seemed that there was an ethical duty to include St. Luc patients who had, in good faith, allowed themselves to be in the studies, but had the unfortunate experience of having had an investigator falsify entry data on other participants.

While Fisher and colleagues were completing the last set of re-analyses and deliberating on how best to communicate the situation to the scientific community and the public, to emphasize the seriousness of the transgression while also indicating that study results were not affected, the press picked up on the story.⁸⁶ In addition to denouncing the fraud, Fisher himself was targeted personally by the press and others.⁸⁷ A congressional investigation had been initiated which led to pressure on the University of Pittsburgh and the NCI to remove Fisher from his chairmanship at NSABP. For a brief time, some of Fisher’s publications were flagged with statements of possible misconduct. In some cases, publications without St. Luc data were falsely flagged.⁸⁷ Additionally, B-06, which had six patients with falsified biopsy dates, was singled out for an NCI audit of NSABP institutions other than St. Luc. After the NCI audit was completed, the reanalysis of B-06 results was published in the New England Journal of Medicine⁵³ along with the results of the NCI audit.⁸⁸

Although all study results were vindicated, the personal damage to Fisher was ruinous. It took until 1997 for him to be cleared of misconduct charges by the ORI.⁸⁹ Also in 1997, his settled his then ongoing lawsuit against the University of Pittsburgh and the NIH.^90,91 Part of the settlement required that public apologies be made to Fisher by all entities in the suit. After these events occurred, he was named as the scientific director of NSABP.

Fisher was very productive in the next 10 years publishing long-term results of earlier NSABP studies^{43, 54} and first results of more recent studies including those of the P-1 prevention trial.^67–78 Towards the very end of his career, Fisher was still active—working on a regular basis until the age of 99. He was still assessing and revising his hypothesis,⁵⁸ providing overviews of the history of surgery,⁸³ and offering insightful updates of earlier surgical and breast cancer interests such as multicentricity.⁹² He continued to write and approached each manuscript as if it were most important paper that he had ever written.

Conclusions

Bernard Fisher changed the paradigm of how and why breast cancer metastasizes and how to treat and possibly prevent such metastases. Furthermore, his early advocacy and use of large, randomized multicentered clinical trials to scientifically test proposed biological and clinical hypotheses set a standard of evidence that is still in use today.⁹³