In April of 1981, an article appeared in the Annals of Surgery¹ captioned, "A prognostic model for clinical stage I melanoma of the upper extremity. The importance of anatomic subsites in predicting recurrent disease." The authors were 19 in number, namely, CL Day Jr.,* AJ Sober,* AW Kopf,** RA Lew,* MC Mihm Jr,* P Hennessey,** FM Golomb,** MN Harris,** SL Gumport,** JW Raker,** RA Malt,* AB Cosimi,** WC Wood,* DF Roses,** F Gorstein,** A Postel,** WR Grier,** MN Mintzis,** and TB Fitzpatrick.* In that article, Day et al. told of having studied 13 variables, among them "subsites" anatomically that they claimed were important in predicting the likelihood of metastases (referred to confusingly by them as "recurrence"***) from a primary cutaneous melanoma situated on an upper extremity.

One hundred seven patients with clinical stage I melanoma were assessed by Day et al., the mean follow-up of those patients being 54 months. The authors divided the patients into two groups according to the location on the upper extremity of the primary melanoma, the location being designated by them either "hand-posterior upper arm (H-PUA)" or "forearm-anterior upper arm (F-AUA)." No good reason for that division was given, only a statement to the effect that " recurrence of melanoma was only reported in patients whose primary lesions were located either on the hand or posterior upper arm." This is what the coworkers found in regard to metastasis: "Five years disease free survival of 68% was reported for 44 patients with melanoma at these sites (H-PUA), with no recurrence in the 68 patients with melanoma of forearm and anterior upper arm (F-AUA)" which means that (70% of patients with primary melanoma on locations called H-PUA were likely to be free from metastases for five years." What, one is compelled to wonder worriedly, prompted the authors to identify those sites as having a worse prognosis for primary melanoma if 70% of patients with a primary melanoma positioned at that site had a good prognosis for five years at least.

The authors divided patients with melanoma at both sites, to wit, H-PUA and F-AUA, into two groups, high risk and low risk, based on apparent survival free of disease for five years. The low risk group was constituted of 84 patients whose primary melanoma measured >2.25 mm in thickness, three quarters of those melanomas being situated on the forearm and anterior arm (F-AUA) and one quarter of them on the hand or posterior upper arm (H-PUA), that latter fraction comprising one half of the total number of patients with melanoma on H-PUA. In the high risk group were 23 patients with melanoma on H-PUA, the average thickness of which was > 2.25 mm and rate of five-year survival free of disease was 37%. The major difference between the two groups was the thickness of the melanoma, those patients in whom thickness of the neoplasm was less than 2.25 mm (the low risk group), irrespective of precise anatomic site, had a much better rate of survival for five years (98.7%) than melanomas that were thicker (<2.25mm).

In sum, according to the data of Day et al. in 1981, (70% of the high risk group (H-PUA) showed no sign of what the coworkers referred to as "recurrence", that is, either persistence of melanoma at the primary site or a metastasis.

A rate of survival free of disease for five years of 98.7% should have called into question the enthusiasm shown by the collaborators for "importance of location in melanoma." Another problem limiting greatly the authenticity of the results of the investigation by Day et al. was the small number of patients in their series and the short period of follow up of them, those two considerations contributing to compromise seriously of capability to reach a conclusion that would be meaningful.

Last, the design of the work by Day et al. lacked precision. For but one example, yet a pivotal one, nonetheless, is the matter of where, exactly, is the posterior upper arm and what, definitively, are the boundaries of it? Even were it made clear just where the "posterior upper arm" begins and just where it ends, it hardly would be proper for a physician to inform a patient that prognosis is grim and an elective node dissection is requisite because a melanoma happened to reside within that ill-defined boundary—or just above or just below it.

Shortly after publication of the article by Day et al., a spate of other articles by the same authors, more or less, about the same subject appeared. In August, 1981, Day and 16 other authors (CL Day Jr.,* AJ Sober,* AW Kopf,° RA Lew,* MC Mihm Jr.,* FM Golomb,° A Postel,° P Hennessey,° MN Harris,° SL Gumport,° JW Raker,* RA Malt,* AB Cosimi,° WC Wood,° DF Roses,° F Gorstein,° TB Fitzpatrick*) proposed, in the American Journal of Surgery, "A prognostic model for clinical stage I melanoma of the trunk. Location near the midline is not an independent risk factor for recurrent disease."³ The study by them was based on 15 variables chosen because of their purported usefulness in predicting "recurrent disease" in 254 patients with clinical stage I melanoma of the trunk. Once again, the authors defined "recurrent disease" either as persistence locally at the primary site or as metastasis. This is what they had to say about features they regarded as being important for predicting prognosis accurately:

Moreover, they made these assertions:

Patients with lesions 1.70mm to 5.49 mm in thickness were separated into three groups thus: (1) High risk:<3.6mm in thickness located in upper trunk (irrespective of lymphocyte response or mitoses/mm2), (2) Intermediate risk: a) <3.6 mm in thickness on lower trunk. b) =3.6mm in thickness and <6 mitoses/mm irrespective of lymphocyte response and location, or c) = 3.6 mm in thickness located on upper trunk with nearly absent or minimal lymphocyte response, and (3) Low risk: =3.6mm in thickness with = 6 mitoses /mm located on the lower trunk. On the basis of these findings, the collaborators concluded that "The trunk lesions located near the midline did not have a worse prognosis than more lateral melanoma of similar thickness."

The most egregious mistake made by the coworkers was that as long as thickness of a melanoma was less than 1.70 mm or more than 5.5-mm, the variable of thickness was shown to be the major factor in prognosis, irrespective of the location of the neoplasm anatomically. For them, when the very same melanoma was between 1.70 mm and 5.5 mm in thickness, location, that is, on the upper trunk or the lower trunk, became the major factor in prognosis. In short, the authors came to the same wrong conclusion as they had a few months earlier that year.¹ A second error is that Day and co-workers failed, once again, to define what, precisely, were the boundaries of the upper trunk and the lower trunk. The criteria they established for midline and lateral trunk were these:

Nothing was said by them, however, about how those boundaries were arrived at, but they stressed, nevertheless, that "the trunk lesions located near the midline did not have a worse prognosis than more lateral melanoma of similar thickness." It remains unclear how it was determined that a patient with a melanoma on the upper trunk has a worse prognosis than one with a melanoma on the lower trunk when, in fact, the boundaries employed for identifying those supposed sites were never stated. It seems that midline and lateral trunk are part of what the authors call upper and lower trunk. A third serious limitation of the work is apparent in their Tables 1 and 2 . Scrutiny of those tables leads to the conclusion that there is no difference in survival based on where on a trunk a melanoma is located. According to the data of Day et al., the rate of survival changes only by virtue of the thickness of the melanoma, not the anatomic site of it. Last, the authors continue to use the term "recurrent" in a sloppy way, that is, for both persistence of a melanoma at the primary site and for metastasis of melanoma. It is obvious those two phenomena are very different from one another and, perforce, the implications of them for survival of a patient are very different.

Following publication of the article by Day et al.² in August 1981, another article was brought forth by nearly the same authors and in it BANS was proposed for the first time as an important guide to prognosis. The new work titled, "Prognostic factors for melanoma patients with lesions 0.76–1.69 mm in thickness. An appraisal of thin level IV lesion," appeared in the Annals of Surgery and was authored by 21 dermatologists, pathologists, and surgeons, senior among them again being Day (CL Day Jr.,* MC Mihm Jr.,* AJ Sober,* MN Harris,** AW Kopf,**TB Fitzpatrick,* RA Lew,* TJ Harrist,* FM Golomb,** A Postel,** P Hennessey,** SL Gumport,** JW Raker,* RA Malt,* AB Cosimi,** WC Wood,° DF Roses,° F Gorstein,** D Rigel,° RJ Friedman,** MM Mintzis°).⁴ All 21 collaborators were affiliated with either the department of dermatology or of surgery at the Massachusetts General Hospital, Harvard Medical School, or with one of those departments as well as pathology at New York University School of Medicine. Day, himself, was a resident in dermatology at Harvard. In this study, 14 variables were tested for their usefulness prognostically in 203 patients with clinical stage I melanoma whose primary malignant neoplasm was between 0.76 and 1.69 mm in thickness. This is what the authors averred by way of conclusion:

Consideration also was given by the researchers to testing the efficacy in these patients of elective regional node dissection. This third article was more comprehensive than the first two publications about the subject by Day et al.^1,2 and established beyond doubt in the minds of the members of the team the importance of location of melanoma in regard to predicting prognosis accurately, the crucial locations for that being captured by them in the mnemonic BANS, an acronym for what the authors called "high risk locations" of primary cutaneous melanoma.

These are some of the errors committed by the coworkers:

According to their own data concerning the 12 patients who died, seven already had had an elective lymph node dissection (one of those showing metastasis to a node) and five did not undergo such dissection. Based on this data, it is reasonable to conclude that lymph node dissection made no difference in the survival of any of these patients over the course of a period of five years. But lymph node dissection never had any benefit for patients and the proposal by Day et al. that it be performed predicated on a primary melanoma having been present on a BANS site was as ill-conceived as it was potentially harmful.

Following soon on publication of these three articles devoted to the matter of BANS came a series of articles that called into question the very legitimacy of BANS itself. The first of those articles was by Woods et al.⁷ (JE Woods, WF Taylor, DJ Pritchard, FH Sim, JC Ivins, EJ Bergstralh) who studied 427 patients with stage I melanoma seen at the Mayo Clinic between 1973 and 1981. This is what they had to say about BANS:

In August 1986, Rogers et al. (GS Rogers, AW Kopf, DS Rigel, ML Levenstein, RJ Friedman, MN Harris, FM Golomb, P Hennessey, SL Gumport, DF Roses), all of them from New York University, drew attention to the "Influence of anatomic location on prognosis of malignant melanoma: attempt to verify the BANS model."⁶ A few of the authors listed were among the ones who had set forth the concept of BANS originally.⁴ By the time of their publication in 1986, however, they had lost confidence in the concept and the model of BANS,⁴ that being a consequence of irrefutable criticism by other students of the subject. Rogers et al. assessed findings in 1,150 patients with primary cutaneous melanoma who were followed closely from 1972 to 1982, information about those patients having been entered, prospectively, in the data base of the Melanoma Cooperative Group at New York University Medical Center (the very same data that had been tapped for use in the very first article about BANS⁴). The authors, by their own admission, were motivated to repeat the study because of doubts raised about the concept of BANS in several articles, that providing them the impetus to analyze data gathered since publication of the original article about BANS in 1981 by the Melanoma Cooperative Group. What they found was stated by them thus:

In short, the contingent from New York University acknowledged forthrightly that the claim in the original article about BANS that patients with melanoma 0.76–1.69 mm thick located on BANS sites (84%BANS, 99% Non-BANS) had a 15% worse chance for survival than did patients whose melanoma was located on a non-BANS site was not in synchrony with the results published subsequently by Blois et al.,⁷ Cascinelli et al.,⁸ and Woods et al.⁵

The conflict in data prompted serious questions to be raised about the very legitimacy of the idea of BANS itself. Blois et al., for example, recorded only a 6% poorer survival for patients with primary melanoma that arose at BANS sites, which is not significant statistically; Cascinelli et al. found a 2.5% better 10 year rate of survival for patients with melanoma at BANS sites; Wood et al. noted no difference in survival between melanoma on BANS and non-BANS sites. Wood and coworkers made this statement about the authenticity of BANS:

They stated further that "The data suggest, however, that BANS areas may not be unique" and, that being the case, "We encourage others to test the BANS versus Non-BANS subsites on their data bases, in this way the validity or invalidity of the BANS sites can be definitely established." More criticism of BANS came from those with a particular interest in melanoma in different parts of the world. By 1986, BANS, which had been launched as a concept just four years earlier, in 1982, was abandoned, even by some who had been the most enthusiastic advocates of it.