Frequently Asked Questions

Did the FDA miss this? (show/hide)

I believe that the FDA failed to consider both the unexpected, unexplained data, and the alternative explanation for the trial results which stems from it. This belief is founded on having read the 8 Provenge advisory committee documents from 2007 and the transcript of that event, as well as the approximately 160 FDA documents released after the approval of Provenge. In none of these did I find either mention, or discussion, of several critical anomalies and questions.

These include:

  • Whether removing 2/3 of the cells from the placebo arm but not from the Provenge arm can legitimately be termed “placebo” (irrespective of assumptions regarding how benign this may be)
  • What proportion of circulating PBMCs are removed during the leukapheresis, to thereby give context to consideration of losing 2/3 of these from the placebo group (cell counts are only ever expressed as just that, never as percentages of baseline counts)
  • Whether any studies have ever been done in elderly patients or cancer patients regarding the impact of significant repeated removal of PBMCs.
  • Identifying the unexpected observation that Provenge appears only to work in patients >65, and not <65 (the inverse of what would be anticipated a priori), and hence discussing what might have caused this counterintuitive result.
  • Identifying the unexpected observation that placebo patients <65 lived (a very statistically significant) 11 months longer than placebo patients >65, and hence scrutinizing the intervention received by the placebo arm to see if some part of it that was assumed not to have any effect might have caused this.


Why was this missed by the FDA? (show/hide)

Obviously I cannot know why, but I can offer informed speculation and make some reasonable suggestions regarding extrinsic and intrinsic factors, all of which probably played some role.

Intrinsic (show/hide)

  1. The ultimate source of all the problems here is the approval of a very flawed trial design. A true placebo would have consisted of giving placebo patients a fake leukapheresis (running the machine without extracting any cells), and 3 days later giving them a saline infusion. Using opaque pouches would maintain the blinding for both the treating doctor, and the patient (the leukapheresis was done at a separate location.

    The trial design was approved in 1999 by someone who is no longer at the FDA. After this original mistake, nobody ever revisited the fundamental trial design to question it. Not only would another FDA member be questioning their colleague’s judgment, but it would be politically fraught to tell Dendreon that their trial design had been unapproved and they would have to start over.

    Although removing 2/3 of placebo patient’s cells clearly introduces a second variable, the assumption of the person approving the trial design (and almost everyone that has scrutinized the trial design since) was that this was unimportant and would have no effect. Since this appears to be true for the patients under 65 in the trial, and since there is no age restriction on blood donation, it is understandable that this assumption was made, given the reason for the removal and freezing of the placebo cells (see “2. Why was the study designed like this” below).

  1. Multiple individuals involved with different branches of analysis.
    • The problematic age data is only found in the statistical reviewer’s documents. However, if the statistician did not have expertise in prostate cancer trials (I have no idea if he did) then he might have assumed it reasonable that old patients were dying sooner than younger patients in the placebo group. 
    • If the statistician didnt have expertise in immunology and particular cancer immunotherapy (again, I have no idea if he did or not), then he might not have noticed and flagged to his colleagues how unusual and unexpected it was that Provenge appeared only to be working in older patients; exactly the patients in whom vaccines do not usually work. 
    • If his colleagues were not reading the stats doc that he produced with a fine comb, they could easily have missed the one chart which showed the unusual age data.

  1. No immune aging expertise on evaluation team
    • Although the field of immune aging or immunosenescence is now well established and a robust field of active research, in 1999, when these trials were approved, it was still a relatively young field.
    • As I discovered during my research, many immunologists have no knowledge of immune aging, and have some conviction that losing this number of cells could not possibly be harmful until referred to the immune aging literature.
    • However, as the data from this trial might demonstrate (if the Alternative Explanation is correct), what can be said of the young immune system cannot be said of the elderly immune system, and particularly not of the immune system after the age of 65 or the immune system of a patient with metastatic cancer.
    • Without such expertise, assuming the removal of 2/3 of cells from placebo patients to be benign (and therefore not an imbalance to worry about, though still not a defensible placebo design) could easily have passed unquestioned.

  1. No decline in white blood cell levels of placebo patients
    • The ways in which the elderly immune system could be particularly affected by repeated lymphocyte removal are countless, and most of these would not be seen in declines in white blood cell levels. This box from our JNCI paper mentions a few:
      • Decreased naive T cell levels 
      • Collapse in diversity of both naïve and memory T cell subcompartments
      • Decreased T cell responsiveness resulting from both decreased expression of CD28 and proliferative exhaustion
      • Increased numbers of circulating natural killer cells with reduced responsiveness and cytotoxicity per cell
      • Impaired differentiation of CD34-positive cells into mature dendritic cells
      • Disruptions in T-cell/B-cell interactions 
      • Alterations in immune cell trafficking as a consequence of widespread changes in cytokine and chemokine signaling
      • The Background Science primers on immune aging detail the myriad ways that the decline in immune function in the elderly all occurs within the context of steady white blood cell counts. The mechanisms to preserve overall t-cell levels are strong and do not decline with age or immune health, so lymphocyte counts do not provide useful information on the consequences of most kinds of immune manipulation.

  1. 2007 panel given incomplete information to discuss
    • The 2007 advisory committee were not provided with the data on age stratification from the trials, so their expertise was partly wasted. The age stratification data is critical to lead one to questioning the placebo design. Neither the adcom members, nor the broader scientific community had access to this data.
    • The 2007 adcom were not given the data from which cell counts in the leukapheresis and final products could be ascertained (their copy was also redacted).
    • Some of the answers provided during Q&A to the specific topic of cell composition of the product appear incongruous.
    • The document provided by Dendreon contained a typo, expressing the volume of blood being leukapheresed as “1.5-2 L”, rather than “1.5-2 body volumes”
  1. Leukapheresis product is only a small pouch of fluid shipped to Dendreon for manufacture. It is non-obvious to any casual observer that such a small volume of liquid could contain over 95% of the lymphocytes from the donor's blood system. There was no report of cell #s involved until well into the trials.

Extrinsic (show/hide)

  1. A group (supported and run in large part by investors) whose charter states that they were created for the implicit purpose of lobbying and conducting litigation to seek the approval of Provenge, pursued an overwhelming campaign of exactly such lobbying and litigation against both the FDA and many individuals involved with the approval process from June 2007 until today. This could not have fostered an environment conducive to optimal decision-making.
  2. The same group was successful in co-opting the voices of congressmen and senators to pressure first the FDA, and then Medicare, towards approval and coverage. 
  3. The problematic age data, and data on cell counts were not public, so the FDA was unable to benefit from public input
    • While the FDA has many excellent clinicians and scientists on its staff, it cannot possibly deploy experts from every potentially relevant speciality to scrutinise clinical data, particularly in new and rapidly-evolving areas of research such as cancer immunotherapy. For every expert within the agency, there are hundreds, if not thousands outside its walls. Although most public companies are quite secretive with their data, they are obliged under laws governing disclosure of information material to their shareholders, to disclose such data to the public. This includes serious adverse events, and should also include significant unexpected results.
    • Typically every piece of data relating to any pre-approval drug, especially any anomalous human results, would be thoroughly debated and scrutinised by the global scientific and medical communities long before the FDA is called upon to make its evaluation and ruling.
    • Since the data on age stratification and on cell counts were never available publicly, the FDA could not benefit from this massive resource in this instance. The importance of immune aging expertise in considering these trials made it particularly unfortunate that the FDA could not benefit from public debate in this instance.


Why was the trial designed like this in the first place? (show/hide)
    • PFS was the original endpoint. In 1999 when the trials were designed, docetaxel had not yet shown a 2.9 month survival benefit which lead to its approval by the FDA in 2004. Salvage therapy (Frovenge) was deemed necessary to entice patients to participate in the trial (and since phII data lead some to believe Provenge showed efficacy, it was argued by some that it would be unethical not to offer placebo patients the opportunity to receive Frovenge). Removing 2/3 of the cell lots from placebo patients was introduced as the way in which Frovenge could be manufactured without requiring another 3 leukapheresis procedures to manufacture treatment for these patients. After the endpoint was changed to OS from PFS the trial design should have been changed (since unblinding a study midway - let alone as early as patients in these trials were unblinded - let alone when subsequent therapies are not randomized - is not generally viewed as a legitimate blind), but it was not. 
    • Placebo cell loss was assumed to be inconsequential. But in science you cannot make assumptions without strong evidence to justify it. Sugar pills are used as placebos because doctors know that a few grams of sugar have little effect on most diseases. However, lactose is not used as a sugar for placebos due to the prevalence of lactose intolerant study subjects for whom the “placebo” would have quite tangible effects. It is likely that the magnitude of the cells being removed from the blood was not appreciated.
        1. The volume of liquid from the leukapheresis machine in which the white blood cells are shipped to Dendreon is just a small pouch. It is non-obvious to think that this small volume could contain over 90% of the patients baseline circulating mononuclear cells.
        2. It is likely that no data on cell counts, nor on cell losses to manufacturing would have been available to the FDA at the time the trial design was approved. It is possible that if they had been aware of the numbers they might have considered more carefully their assumption regarding the benign consequences of the repeated losses in each arm.


Is there any proof that repeated loss of PBMCs is harmful to cancer patients over 65? (show/hide)

There is proof from human studies that significant loss of lymphocytes is harmful. These have been conducted in somewhat different populations (Studies showing that after immunodepletion resulting from either Campath-1H treatment in arthritis or chemotherapy in various cancers, the repopulation of blood results in diminished naive CD4+ T cell populations with both reduced functionality, and diminished receptor diversity), and any direct read-across must be caveated, but we must do the best with the data we have, since conducting new studies on this issue would be unethical. My co-authors and I were unable to find any evidence that the repeated loss of PBMCs is safe in cancer patients over 65 (or healthy elderly). The Provenge trial data generated and support the hypothesis that it could be harmful to older patients, and in-house data which is private to Dendreon may contain more information which could inform this matter. 

The evidence from the human studies (described in the video) is also supported by 4 bodies of proof from other studies:

  • A very large body of proof that the elderly immune system is more fragile and less responsive/effective than that of younger patients.
  • A large body of proof that the immune system of cancer patients is more fragile and significantly altered from that of healthy people of the same age.
  • Many studies indicating that vaccines do not work in patients >65
  • Many detailed biochemical studies showing that age 65 is associated with a precipitous collapse in the diversity (and therefore responsiveness) of both naïve and memory T cell populations. T cells have a central role in cancer immunity.
  • Studies showing that (in sheep, since it would be unethical to conduct such studies in humans) the lymphocytes in the blood and the lymphocytes in the lymph are not a freely interchanging pool, and therefore cannot be considered a single pool when assessing the significance of the number of cells being removed from the blood.

Hypotheses ALWAYS come before proof. Hundreds of the most important discoveries in science have originated in just this way. A significant unexpected result is observed which generates a hypothesis which is tested experimentally. Either the hypothesis is proven, or another hypothesis to explain the unexpected results is generated. To simply ignore the unexpected results just results in remaining ignorant for longer. In medicine this ignorance usually results in the delayed discovery of a useful therapy or the delayed discovery that a current practice is harmful. To name a few examples:

Beneficial therapies discovered by investigating unexpected observations:

  • In 1928, Alexander Fleming observed that there were “holes” cleared of bacteria in the layer of staphylococci he was growing in petri dish which he had inadvertently left on his workbench before going on holiday. The “holes” investigated and identified to be patches of mould; a rare strain of Penicillium notatum  which produces penicillin.  Thus antibiotics were first discovered.

  • In the 1700s, Edward Jenner wondered why, at a time when smallpox was rampant, milkmaids seemed not to be susceptible to the disease. He hypothesized not only that the cowpox the milkmaids had previously suffered from gave them resistance to smallpox, but also that cowpox could be transmitted from one person to another as a deliberate mechanism of protection. So he took the puss from a fresh cowpox lesion on the hand of a milkmaid and put it into a cut on an 8-year-old boy. 2 months later he inoculated the boy with puss from a fresh smallpox lesion, and no disease developed. After years of rejection by the medical community, Jenner’s discovery led to the birth of vaccines.

  • Dozens of drugs were discovered by investigation of accidental observations or side-effects during trials for other indications. Partial list of serendipitous drug discoveries (show/hide)

Compound

Nature of accidental discovery OR original indication

Acetanilide

Tested as internal antiseptic (instead of naphthalene)

Acetylsalicylic acid

Irreversible enzyme inhibitor (vs. salicylic acid prodrug)

Aminoglutethimide

Breast cancer treatment (instead of antiepileptic)

Amphetamine

Stimulant (instead of nasal decongestant)

Chloral hydrate

Prodrug of trichloroethanol (instead of chloroform)

Chlordiazepoxide

Tranquiliser (unexpected chemical rearrangement)

Chlorpromazine

Neuroleptic (tested to prevent surgical shock)

Cinnarizine

Cardiovascular (predominant to antihistamine) activity

Cisplatin

Cytotoxic effect of electrolysis product

Clonidine

Antihypertensive (instead of nasal decongestant)

Cromoglycate

Antiallergic (accidental formation of chromone dimer)

Cyclosporine

Immunosuppressant (instead of antifungal)

Dichloroisoprenaline

Beta blocker instead of bronchodilator

Dicoumarol

Anticoagulant found via fatal poisoning of cattle by moldy hay

Diethylstilbestrol

Estrogenic impurity of anol dimerisation product

Diphenhydramine

Prevention of travel sickness instead of allergy treatment

Diphenoxylate

Antidiarrhea agent instead of analgesic

Disulfiram

Hypersensitivity to alcohol instead of antiparasitic

Ether

Had anesthetic effect at an inhalation party

Etomidate

Anesthetic instead of chemotherapeutic activity

Griseofulvin

Growth inhibition of fungi on certain soils

Guanethidine

Antihypertensive instead of antitrypanosomal

Haloperidol

Neuroleptic instead of analgesic activity

Heparin

Deterioration of lipid coagulant unmasked anticoagulant activity

Imipramine

Antidepressant instead of neuroleptic activity

Iproniazid

Antidepressant instead of tuberculostatic activity

Isoniazid

Tuberculostatic activity of organic intermediate

Levamisole

Immunomodulating (instead of antiparasitic) agent

Lithium carbonate

Antidepressant activity of lithium salt of urate

Lysergide (LSD)

Hallucinogenic (instead of cardiovascular) activity

Meprobamate

Tranquiliser (instead of muscle relaxant)

Merbaphen

Diuretic activity of antisyphilitic agent

Methaqualone

Hypnotic instead of antimalarial activity

Mifepristone

Antiprogesterone (instead of glucocorticoid)

Minoxidil

Promoted hair growth in addition to antihypertensive activity

Naftifine

Antifungal; rearrangement product of CNS drug

Nalorphine

Respiratory depressant (and opiate) instead of respiratory stimulation

Nitrogen mustard

Cytotoxicity observed after ship bombardment

Nitroglycerin

Antianginal activity (headache upon inhalation) of high explosive

Nitrous oxide

Anesthetic effect in accidental wounding in laughing gas session

Norethynodrel/mestranol

Estrogenic impurity in the first oral contraceptive

Penicillin

Antibiotic activity of Penicillium infection

Pethidine (meperidine)

Morphine agonist (instead of spasmolytic)

Phenylbutazone

Anti-inflammatory activity of solubility enhancer

Phenolphthalein

Laxative, first tested as label for cheap wines

Praziquantel

Antiparasitic agent (instead of antidepressant)

Prednisone

Bacterial oxidation roduced highly active analog

Propafenone

Antiarrythmic (instead of beta-blocker)

Sildenafil citrate

Anti-impotence agent instead of cardiovascular drug

Sulphamidochrysoidine

Prodrug of sulfanilamide (active only in vivo)

Sulfonamides, various

Diuretic and antidiabetic side effects

Tamoxifen

Antiestrogenic activity of cis-isomer

Urethane

Hypnotic activity (instead of alcohol prodrug)

Valproic acid

Anticonvulsant; started as solubility enhancer for various drugs

Warfarin

Low acute toxicity of rat poison in failed suicide attempt

Source: Kubinyi, H. (1999) Chance favors the prepared mind – from serendipity to rational drug design. J Receptor and Signal Transduction Res 19: 15-39  & Sneader, W. (1996) Drug Prototypes and Their Exploitation



Harmful practices discovered by investigating unexpected observations:

  • The drug approval process is highly sensitised to adverse effects of drugs as revealed in clinical trials. But sometimes a negative effect is limited to a specific population, and observations that a certain practice or drug is safe in most people, does not constitute the basis for dismissing unusual observations in one subgroup as unrelated to the intervention. When it comes to adverse effects in specific subpopulations, perhaps no drug is more notorious than thalidomide with its teratogenic properties, resulting in its worldwide ban in 1961. Once the populations at risk had been fully researched and identified, the drug was reintroduced with tight controls in 1997 with FDA approval for erythema nodosum leprosum and later for multiple myeloma.
  • Although the observation that hepatitis appeared to be transmitted by transfusions had gradually emerged over many years, and in 1943 P Beeson published the classic description of transfusion-transmitted hepatitis, it was not known how or why the cause and effect were related. Only with decades of research were first antigens, then the three types of hepatitis virus (A, B, then C), and finally screening tools for the blood identified.

“The exceptions to any rule are most interesting in themselves, for they show us that the old rule is wrong. And it is most exciting, then, to find out what the right rule, if any, is.  The exception is studied, along with other conditions that produce similar effects. The scientist tries to find more exceptions and to determine the characteristics of the exceptions, a process that is continually exciting as it develops. He does not try to avoid showing that the rules are wrong; there is progress and excitement in the exact opposite. He tries to prove himself wrong as quickly as possible”

   -- Richard P. Feynman



Is there an upper age limit on blood donation? (show/hide)

In the US there is no upper age limit for whole blood donation. However, most other Western countries have imposed such limits, ranging from 65 years (UK and the Netherlands), to 68 (Germany), 70 (Australia) and 71 (Canada).

  • Whole blood donation is typically 450ml, and is recommended not to be repeated within 8 weeks (US) to 16 weeks (UK).  Since the average adult has 4.5-5.5 liters of blood, this represents 8-10% of the total volume ( vs the 90% of PBMCs (with partial return of <12%) removed 3 times in 4.5 weeks from the placebo group in the Provenge trials).
  • Leukapheresis to remove the same type of blood cells as in the Provenge trials is only performed in response to occasional, discrete requests to provide cells for immunology research. Leukapheresis (typically for granulocytes, a different subset of white blood cells to the PBMCs extracted in the Provenge trials), is very rarely performed as a blood donation process.

Searching the literature I could not find any trials (whether prospective, retrospective, randomized or epidemiological) that had been published which investigate the long term consequences of repeated blood donation in donors aged 65 or older.  There are several studies that report adverse reactions (occurring immediately following blood donation) in elderly donors. These have been used to inform changes in the upper age limits for donation, but do not inform the question at hand.




Responses to Counter-Arguments (against the alternative explanation)

"This is a retrospective analysis, and therefore invalid" (show/hide)

Inevitably someone will suggest that our observations on the perplexing age data are invalid because the analysis was not pre-specified. This argument is sometimes appropriate, but it is woefully insufficient here.

  1. Advancing age is the primary cause of deterioration in the immune system. In the trial of an invasive protocol removing and manipulating immune cells, age is perhaps the single most important subset to analyse.
  2. Provenge approval was contingent upon retrospective analyses.
    • In IMPACT, the survival of patients <65 greatly favored placebo patients. Only by retrospectively pooling the patients from all 3 trials [a highly unorthodox analysis, since trials had significant differences and the FDA typically forbids any such pooling] was it possible to neutralize this observation into a more benign one in which the drug merely showed no efficacy in patients <65.
    • When first applying for FDA approval in 2007, Dendreon had two small trials as proof of efficacy. D9902A (n=98) was not statistically significant, and the significance of the survival benefit in D9901 (n=127) was the result of retrospective analysis.
  3. Extremely significant: The fact that there are four, separate, interwoven and overlapping, unanticipated observations in the age data make their significance infinitely greater, and almost impossible to attribute to chance alone.

    i.The drug does not appear to work in patients <65 – unexpected

    ii.Patients >65 in the placebo group live 11 months (median) shorter than patients <65 – unexpected

    iii.Patients >65 in the Provenge group live 5.6 months (median) shorter than patients <65 – unexpected (given the observation that the drug does not appear to be working in the younger patients, although not statistically significant due to high variability in the data)

    iv.The age group in whom the drug “works” (patients >65) live shorter than the age group in whom the drug doesn’t work – unexpected (and, to me, baffling)

    The 11 month survival difference between placebo patients >65 and <65 is an extremely statistically significant result (there is no overlap in the 95% confidence intervals). Since the placebo arm age difference is expected to be insignificant, the cause of this “age effect” cannot be separated from the alleged treatment effect.

  4. All-but pre-specified: In a teleconference with the FDA in 2006, years before the IMPACT data were analysed, Dendreon proposed submitting this subgroup analysis from the two smaller, earlier trials as part of their first application for approval. Then when submitting the IMPACT data for their second approval application, they included these subgroup analyses. In my personal opinion, specifying a subgroup several years before a trial’s data are unblinded is the next best thing to being a pre-specified analysis. In their analysis of 9901 (the first trial to be unblinded and analysed) they identified the statistical significance of the relationship between age and survival in their trials, and noted that this association had not been found by other analyses of similar populations. When trial 9902 was split into parts A and B, it would have been an opportune time to pre-specify this analysis for 9902B.
  5. The observations regarding age speak only to the question of which of the two hypotheses to explain the trial results is more probable. The balance of evidence showing harm from cell losses in other trials would not be affected even if the problematic age data were entirely ignored.


"There was no decrease in white blood cell levels detected during the trials" (show/hide)

One would not necessarily have anticipated decreases in absolute lymphocyte levels as a consequence of the trial interventions, since the immunology literature indicates that even into old age the body’s mechanisms to maintain absolute lymphocyte levels is very robust. In the elderly, however, beneath this superficial maintenance of absolute cell numbers, there are many changes in subtype and responsiveness of the cells which change and deteriorate. The Provenge trials did not measure the parameters of interest. The JNCI paper, and the video discuss the many ways in which elderly immune function could be impaired despite maintaining absolute cell numbers.



"There was no increase in infections reported in the trials" (show/hide)

There are two main reasons why this does not constitute a valid argument against the alternative explanation for the Provenge trials:

  1. As described in the paper excerpted and linked below (Dr Higano is an investigator and consultant for Dendreon, and has co-authored several of Dendreon’s publications of results), “in general, patients with cancer die of some combination of organ failure, infection, and cachexia”. They state further that “the proximal cause of death in the sipuleucel-T trials was not reported”. Therefore, we do not know if there was an increase in infections in the trials or not, because they were not reported.

                                                                         http://www.roswellpark.edu/sites/default/files/2011 Cheever et al.pdf

 

  1. Of secondary importance, this constitutes a weak argument against the immunodepletion hypothesis because the mononuclear cells removed belong mostly to the adaptive arm of the immune system.

Elderly men with prostate cancer are, for the most part, not living highly active lives outside of their homes where they would be exposed to lots of novel pathogens.

The innate immune system is the body’s first line of defense in defense against other organisms. The cells most involved in innate immunity (and those first recruited to sites of infection), namely mast cells, eosinophils, basophils and the phagocytic cells including macrophages and neutrophils are barely affected by the leukapheresis. Although dendritic cells are also involved in innate immunity and through the process of antigen presentation they serve as a link between the innate and adaptive immune systems.

Pathogens that evade the innate immune response are generally rare (often due to successful vaccination programs that make use of the adaptive immune response memory), such as tuberculosis, salmonella typhi (typoid), polio, smallpox, measeles, mumps, rubella and Bacillus anthracis (anthrax). Others, such as staphylococcus and streptococcus are likely to have been encountered previously and be represented in the memory T cell compartment. It seems unlikely if these pathogens had not been encountered previously that these would be encountered for the first time by a study subject during the 6-18 month timeframe of the trial).                 



"Patients lived longer than their Halabi-Predicted Survival, so how could these treatments be harmful?" (show/hide)

Dendreon has presented the IMPACT survival data at ASCO, to FDA, to investors, and elsewhere directly accompanied by the Halabi predicted survival (HPS) for the two arms.

Survival

(months)

Halabi Predicted

Actual

Provenge (n=341)

20.3

25.8

Placebo (n=171)

21.2

21.7


These data have been used to illustrate the fact that Provenge patients lived longer than their predicted survival, whereas placebo patients did not. There are several studies and observations that speak to this point:

               i.         The Halabi model was built on learning data and tested on a validation data set generated from trials conducted long before the FDA approval of docetaxel in 2004. Very few of these patients would have received the docetaxel regimen which is now the standard of care, and showed a 2.9 month survival benefit in trials. The Halabi model is therefore typically considered a model predicting survival in the underlying, untreated disease. At a minimum the HPS in the Provenge trials should have 2.9 months added before comparing it with patients’ actual survival. [if this is done, the 21.7 month survival of placebo patients is shorter than their 25.1 month “docetaxel-modified” HPS, and the 25.8 month survival of Provenge patients is unremarkable against their 23.2 month “docetaxel-modified” HPS. However, there are several additional factors that would tend to suggest that even the “docetaxel-modified” HPS might underestimate actual survival in the Provenge subjects.

              ii.         The Halabi model was developed to predict survival in a mixed group of CRPC patients, and cannot therefore be validly applied to an exclusively asymptomatic and minimally symptomatic population. Given that the Halabi model uses neither pain, nor location of metastases as predictive factors, yet both are known as highly correlated with survival, and the IMPACT population excluded both of these patient groups, it is furthermore likely the Halabi model would systematically underestimate the survival of an asymptomatic population excluding these negative prognostic factors.

            iii.         It was shown in the original Halabi paper [http://jco.ascopubs.org/cgi/content/full/21/7/1232 ] that when testing the model in a validation population, a predicted survival of 22.8 months in the highest quartile corresponded with an actual survival of 27.2 months. The model thereby admits, and demonstrated that it increasingly underestimates survival of patients as their prognosis improves in the upper two quartiles.

 One could extrapolate from this, therefore, that the very “healthy” population enrolled into the Provenge trials, with a 20-21 month predicted survival, would be close to this upper quartile and anticipated to have an actual survival at least 3-4 months longer, all else being equal (which it is not).

             iv.         There is a phenomenon which prostate doctors call ”phase shift” by which they are referring to the general increase in survival of patients which occurs over time, but is not specifically linked to identified improvements in therapy, such as the introduction of docetaxel in 2004. Phase shift most likely results from a multitude of improvements in patients care, such as palliative care, nutrition, counseling, more judicious earlier therapies etc. Phase shift would tend to have lengthened the actual survival of patients in the decade since the Halabi model was constructed.

Observations (i) & (iv) are likely to have contributed to the results seen in the placebo arm of the GVAX trial (as discussed in the JNCI paper). Here, the placebo patients had a Halabi predicted survival of 16 months. The actual survival of these patients was 21.7 months, 5.2 months longer than the Halabi predicted survival. Given the additional influence of observations (ii) and (iii) above, it can be seen that the placebo group lived many months shorter than any reasonable updated to their HPS would have predicted. It cannot be asserted that the actual survival of the Provenge group was longer than a prognosis of their survival from a model updated with more recent and relevant data than the HPS.



"The number of lymphocytes in the blood a tiny fraction of the body’s total, so their removal is irrelevant" (show/hide)

Excerpt from video transcript (with images) providing response: Blood_Lymphocytes.pdf