March 10, 2003
Robert C. Young, M.D.
Fox Chase Cancer Center
7701 Burholme Avenue
Philadelphia, PA 19111
I want to ask for your advice on matters relating to the enclosed manuscript, submitted October 31, 2002 to the JNCI, and the enclosed manuscript reviews, received February 11, 2003. This is of relevance to you for two reasons. First, your position in the field of ovarian cancer research. Second, the fact that you are the major reason why I've devoted my entire career to cell culture drug resistance testing (CCDRT) since 1979.
To explain the last first: In the summer of 1978, I was a Clinical Associate in the Medicine Branch, when Salmon's paper was published in the NEJM. You personally assigned me to summarize and discuss the paper at a Clinical Associates' weekly journal club-like conference. In my typical contrarian fashion, I decided that everyone could just read the paper for themselves, but I wanted to research also the background, prior history of the field of "culture and sensitivity" testing in cancer. So I came to the conference prepared to summarize the state of the entire field, as it existed at the time, including the different types of approaches and clinical correlations, dating back to the original publication in a 1954 JNCI, by Black and Speer. The Salmon paper was simply the latest contribution to the field. In my preparation for the conference, I became very interested in the variety of approaches which had been attempted previously. A little ways into my presentation, you more or less cut me off, and led the discussion yourself. You were annoyed, though polite, in saying that "what we were supposed to have heard was...."
Anyway, later that year (Fall of 1978), I was working in Marc Lippman's laboratory, when my assigned project with Marc had to be put on hold, because his MCF-7 cells (he had a contracting laboratory growing and freezing these cells for him) were found to be contaminated with mycoplasma. Marc gave me the freedom to do what I wished during this "downtime," and I did a little project in collaboration with Dan Von Hoff to see if I could get neuroblastoma cells (which he received for "human tumor stem cell" assays from UW Milwaukee) to grow axons and dendrites. But I also got interested in the concept of looking at alternative assay endpoints for cell culture drug resistance testing, based on the knowledge that I had acquired preparing for my ill-fated journal club presentation. I had, in fact, previously done a little chemosensitivity testing project in 1970-71, when I was a graduate student at Michigan working in Ray Ruddon's lab. This was based on Trypan Blue staining of Burkitt Lymphoma cells which would not proliferate in the DMSO vehicle required for solubilizing a series of purine analogues synthesized by Alan Sartorelli at Yale, which I was testing in the Burkitt cells.
All of the above led, by the winter of 1979, in Marc's lab, to my "invention" of the key feature of what I later called the "DISC assay," namely taking a Trypan Blue-stained cell suspension and spinning it down on a cytospin centrifuge to get rid of the "unincorporated" Trypan Blue and depositing the blue-stained (dead) cells and clear, unstained (living) cells on a permanent microscope slide. Later on, in the Fall of 1979, when I had moved to the Long Beach Veterans Hospital, I substituted Fast Green for Trypan Blue and leaned that I could counterstain the living cells with such things as H&E or Wright-Giemsa, to allow determination of which "living" cells were tumor versus normal cells. But the cytospinning of the Trypan Blue stained cell suspension, which was the key part of the overall method, was developed at the Medicine Branch.
My first funded grant, in 1980, was to compare and contrast different endpoints for in vitro chemosensitivity testing of fresh human tumors. I looked at 5 different endpoints, including the Human Tumor Stem Cell assay and my newly-developed "DISC" assay. For various reasons, I settled on the latter as the preferred method, and I've been working with this ever since. The DISC assay has cell death as an endpoint, as opposed to the cell proliferation endpoint of Salmon's assay and its variants. In the early 1980s, virtually no one had ever heard of apoptosis in the context of the mechanism of action of many forms of chemotherapy. Cell death was considered to be a crude and irrelevant endpoint. But my papers of the era served as the model and prototype for later studies of the DISC assay and a related family of other cell death assays, as described in the enclosed invited review prepared for publication in ONCOLOGY (NY), which was subsequently rejected and which was published only on my website, www.weisenthal.org. Also on my website is the single review provided to me by ONCOLOGY, and copies of correspondence relating to this review manuscript.
In 1990-92, Robert Nagourney and I wrote the draft protocol for a planned GOG study to correlate assay results with clinical outcomes in ovarian cancer. We proposed a head to head comparison of the cell death endpoint (DISC assay) with the cell proliferation endpoint (thymidine-incorporation variant of Salmon's assay). I left Oncotech January 1, 1992 and personally recruited John Fruehauf as my replacement. As I was the only person at Oncotech with experience and expertise in using the DISC assay for solid tumors, I asked the GOG that I be permitted to do parallel DISC assay studies in my own laboratory (at my own expense), in the usual cases in which there would be sufficient tumor to send portions both to Oncotech and to my laboratory. I viewed this head to head comparison of the cell death to cell proliferation endpoints as being the most important component of the study. However, despite many conversations, letters, and phone calls, I was unable to convince the GOG to allow me to participate, despite the fact that Nagourney (who had also left Oncotech) and I were the ones who had originally proposed the study and written the draft protocol.
The assay correlation study (GOG 118) has, to this date, never been analyzed or published, and, if ever published, will not include my planned head to head comparison of the cell death versus cell proliferation endpoints.
Though not given the opportunity by the GOG to do my study through them, I tried my best to do the study on my own, using specimens submitted to my laboratory from around the country. This work is described in the enclosed (rejected) JNCI manuscript, with attached reviews. Also enclosed is a "rejection response" I sent to the JNCI's editor.
Now, the reason I am writing to you is this: If you take the time to read and consider the enclosed manuscripts, I think that you'll be convinced that there is by now a relatively large body of good science behind this work. In Emil Frei's Karnofsky address of, I think, 1982, he spoke of the dichotomy between "investigators" (who work comfortably with their peers and who are esteemed by their peers, for their methodical, single step by single step approach) and "discoverers" (who tend to go against the grain, annoy their peers, often fail, but who do occasionally produce multi-step advances). Frei lamented that "discoverers" have decreasing opportunities in clinical research, as the peer review component of the clinical research system has come to be increasingly dominated by "investigators."
I happen to think that the entire clinical trials system has, in advanced ovarian cancer, produced virtually nothing of substance in the past 30 years. As you know, the Advanced Ovarian Trialists Group meta-analysis did not even demonstrate a significant ( p = 0.23) difference in the trials comparing platinum combinations to single agent alkylators, and two very large trials could not demonstrate significant differences between single agent platinum as compared to platinum-Taxol. One of the reviewers for my JNCI manuscript objected that the superiority of platinum regimens for advanced ovarian cancer is an achievement of the empiric clinical trials system. Based on what? Yes, advanced ovarian cancer patients are probably living longer today than in the pre-platinum era. But the contribution of platinum to this improved survival has been by no means proven. In addition to the negative conclusions of the largest and best meta-analysis which has been published, we also know that platinum dose-intensity is not a significant variable in survival, while the year of study publication is a significant variable and maximum cytoreductive surgery is the most powerful determinant (Bristow, et al. J.Clin.Oncol. 20:1248,'02).
Thus, historical comparisons of "platinum-era" therapy to "pre-platinum era" therapy are very misleading. The best evidence is the meta-analysis of randomized trials, and that showed no difference. For second and third line therapy, there has likewise been no important progress. What we have is a very large number of "rigorous," "well-designed," and certainly well-analyzed studies which have shown virtually nothing of any importance, save for the probable equivalency of cisplatin and carboplatin.
In contrast, here is what my (rejected) paper showed:
1. Complete, detailed description of a public domain assay system based on measuring cell death of three dimensional microclusters of fresh ovarian tumor biopsies in short term culture in anchorage-independent conditions, which had a 96% evaluability rate (>98%, including only cytologically positive specimens), testing a median of 20 drugs and combinations, with two different cell death endpoints, with drugs tested at two concentrations.
2. A clear-cut, highly significant relationship between tumor differentiation and platinum resistance, with poorly-differentiated tumors being most sensitive, well-differentiated tumors most resistant, and moderately-differentiated tumors falling in between. These relationships were shown, independently, for both cisplatin and carboplatin.
3. A clear-cut, highly significant relationship between prior treatment status and platinum resistance, with previously-treated tumors being much more resistant than previously-untreated tumors. This was also shown for cisplatin and, independently, for carboplatin. In my opinion, this is now, by far, the most highly validated system for studying the circumvention of platinum resistance in human adenocarcinomas. It also affords a relatively simple, rapid method for reproducing and confirming the clinical and biological relevance of the method by other laboratories.
4. A clear-cut relationship between platinum resistance and long term, overall patient survival, which was striking in the case of poorly differentiated tumors. This was shown both prospectively (assay cutoffs defined and reported in advance of treatment) and retrospectively (through statistical, objective recalculation of cutoffs, based on the greatly enlarged database generated by testing the 450 fresh ovarian tumors reported in the study). This survival correlation held only for previously-untreated patients. There was no significant correlation between platinum resistance and survival in previously-treated patients. These latter associations were also independently demonstrated for cisplatin and also for carboplatin.
Now, as much as I would have wanted to have data regarding tumor stage, effectiveness of tumor debulking, patient performance status, treatment actually administered, etc., these data were not available. I do have official pathology reports of all specimens, patient age (the results hold significantly for subgroups of patients < 65 and > 65, though I did not include this subgroup analysis in the already lengthy paper). As described above, I tried my best to do this study through the GOG, but they turned me down cold.
O.K., so the study is flawed because I couldn't do an elegant, multivariate analysis. Yes, it is remotely possible that the platinum assay-sensitive tumors were optimally debulked, while the platinum-resistant tumors had gross residual disease. But why in the world should an assay result correlate with how much tumor was left behind? Nor is there reason to suspect that the assay-sensitive tumors are more likely to be very early stage tumors (remember that all of these were $1000 - $2000 tests, where the patients agreed in advance to be financially responsible, and that the tumors were submitted because someone cared enough about getting the information to generate this level of expense in sending the tissue to us). I doubt seriously that any early stage tumors were sent to us and, anyway, why in the world should the stage of the tumor be correlated with our assay results? And, even if it were correlated, would this not be a very important finding to ovarian cancer biology? That early stage tumors were more platinum sensitive than late stage tumors? So, either way, the observation of the association of in vitro cell death in the presence of platinum with the long term survival of the patient is very important.
It's worth noting that in the three recent published studies correlating assay results with patient survival (in which a multivariate analysis was possible), there was no association between assay results and clinical factors such as cytoreductive surgery, etc. (Taylor, et al. Eur.J.Gynaec.Oncol. 22:278,'01; Konecny, et al. Gynecol.Oncol. 77:258,'00; Holloway, et al. Gynecol.Oncol. 87:8,'02).
The dichotomy between the "investigator" mentality versus "discoverer" mentality is that the former is more interested in exhaustive analysis of data to conclusively test hypotheses of often trivial importance, while the later is more interested in opening windows leading to new knowledge, even if the initial studies cannot neatly distinguish between all possible explanations of the findings. Yes, it is possible that there may be some, unknown and non-obvious, artifact to explain the results, but is this the most likely explanation? Especially considering the clear-cut relationship also between assay results and prior treatment status? Should it not be sufficient simply to clearly note this limitation of the study (as I did) and let the readers draw their own conclusions?
In my opinion, there is so much in this study which could serve to drive two different fields forward, that it is astounding to me that the value of the work was not appreciated. Certainly, investigators in the fields of cell culture assays, drug resistance circumvention, and development of surrogate markers for drug resistance (e.g. gene expression and/or tissue immunohistochemical microarrays) would find this work to be very worthy of consideration. In my opinion, clinical investigators in ovarian cancer are brain dead, in a creative sense, if they are incapable of appreciating the potential implications of this work. To dismiss it entirely, simply because I wasn't able to perform a multivariate analysis of factors extremely unlikely to in any way alter the conclusions of the study, is, to me, completely unfathomable (and enormously frustrating, given my prodigious effort to do the study through the GOG in the first place).
The above is a very lengthy introduction. I have similar data correlating assay results with long term patient survival in the case of platinum resistance in non-small cell lung cancer, cyclophosphamide resistance in breast cancer, and 5FU resistance in colorectal cancer. I had planned on writing a series of papers describing these correlations. I want to point out also that the technology used is entirely non-proprietary and public domain. I wrote such a long methods section because I wanted to provide a procedure manual so that other laboratories could reproduce and build upon my work.
I have been performing 40 to 50 assays per month (my maximal capacity, given the amount of work I put into each one) for more than 10 years now. I have prolonged and saved many lives and helped even more patients avoid toxicity from inactive treatment regimens. I have also, not infrequently, made life difficult for clinical oncologists by identifying as "best" regimens those which are financially disadvantageous to the oncologists. This is a serious impediment to progress, but this will hopefully change as reimbursement moves oncologists away from being retail pharmacists and back toward being physicians.
What I haven't been able to do is to get formal clinical trials done. I've tried very hard (e.g. VA CSP # 280, multiple myeloma; ECOG #535, NSCLC; GOG # 118, ovarian cancer). I clearly lack the organizational and communication skills to be successful as a clinical investigator. But I've done something original and important and it deserves to see the light of day in the form of the type of clinical trials which everyone demands (e.g. see the suggestion of reviewer #2 of my JNCI manuscript, who says that the only thing worth publishing relating to the field is a three-armed prospective, randomized trial). I'm willing to do it. I'm wanting to do it. I'm waiting to do it. But I can't do it without help. So I come to you; the person most responsible (if inadvertently and unintentionally so) for what has come to be my life's work. And one of the most esteemed ovarian cancer investigators in the world. What can I do? What should I do? Can you help get the studies done which everyone keeps saying should be done?
I hope that your viewpoint on all of this is not clouded by the NCI's NSCLC studies. Those were studies based on subculturing tumors, and they were a test not of the assays themselves but of the feasibility and suitability of using tumor subcultures for this testing. In the NCI's NSCLC studies, 100% of tumors were subcultured; the specimen evaluability rate was < 25%; the tumors were tested as monolayers, which has been shown not to correlate with in vivo resistance in murine tumor models, while testing the same tumors as three dimensional spheroids does correlate (Teicher, Kerbel, et al...do a PubMed search on "MULTICELLULAR [titl] AND RESISTANCE [titl]," and you will pull up a couple of dozen papers which describe this phenomenon). We test tumors as multicellular, floating spheroids. We throw away the single cells (exactly opposite to Salmon's/Von Hoff's assays). Our evaluability rate is 95%, as compared to <25%. We test true fresh tumors and not subcultures. We rigorously standardize conditions, use two different but complementary cell death endpoints, and we base our classification of results on very well defined training set databases. I spend 100% of my time doing these assays, and I've been doing them, largely full time, since 1979. It is completely different than the NCI methodology, yet most oncologists don't understand the difference between cell proliferation assays versus cell death assays, much less understand the nuances of assay technology.
So can you help? Or introduce me to someone who can help?
cc: Barnett S. Kramer
nb: Dr. Young ultimately sent me a polite reply, in which he suggested applying to the NIH for a research grant.