It was a bad week for science-based medicine. It was a good week (sort of) for science-based medcine.
First the bad.
There has been a development that anyone who supports science in medicine and opposes quackery will likely find disturbing. Do you remember Senator Tom Harkin (D-IA)? We’ve written about him extensively over the last several months on this blog. First of all, he is the man most responsible for the creation of that government-sanctioned, government-funded bastion of pseudoscience, the National Center for Complementary and Alternative Medicine. He’s also one of the men most responsible for the Dietary Supplement Health and Education Act (DSHEA) of 1994, which has done more to protect the supplement industry from making dubious health claims than any other piece of federal legislation. More recently, Harkin has made a name for himself in the health care reform debate currently ongoing by inviting advocates of “integrative” medicine (IM), which in essence integrates quackery and the pseudoscientific with scientific medicine, to Capitol Hill as a means of trying to persuade his fellow legislators to include a CAM/IM version of “wellness” care as part of any bill that might pass this fall. In essence, he is trying to hijack any health care reform bill to include government sanction of unscientific medicine. Meanwhile, he has been chastising NCCAM because it hasn’t “validated” enough “alternative medicine” for his taste. (Actually, it’s validated none, because virtually none of it is likely to be valid.)
This is the man who, according to reports, will almost certainly be taking over the chair of the Senate Committee on Health, Education, Labor, and Pensions (HELP) after the death of its former chair, Senator Edward Kennedy. This committee is among the most important for government health policy and will be in the thick of the final negotiations and battles over any health care reform that may arise from Congress this fall.
The existence of powerful supporters of pseudoscience in the highest eschelons of government has real consequences. As I’ve described before, NCCAM, being based entirely on studying highly–even ridiculously implausible–notions about disease and how to treat it, has resulted in the infiltration of quackery into academia, where ideas once rightly dismissed as quackery are respectfully given deference and studied as though they were anything other than Tooth Fairy science, a process that Dr. R. W. Donnell has amusingly termed “quackademic medicine.” One result was the expenditure of $30 million on an unethical, poorly designed, and corrupt trial of chelation therapy for cardiovascular disease. Another result was an even more unethical trial of an even more scientifically implausible remedy for a deadly cancer. Although the fact that the trial was even done is a horror, at least last week we finally found out the results, which had been suppressed for nearly four years, namely that this protocol is not just useless, but worse than useless. It’s a Pyrrhic victory for science-based medicine and cold comfort to patients with pancreatic cancer who may have continued to use this protocol during those four years, but at least we finally know.
Let’s take a look at the study. But first, a little background.
PANCREATIC CANCER AND THE GONZALEZ PROTOCOL
Pancreatic cancer sucks.
There’s no other way to put it. It’s one of the most lethal, if not the most lethal, cancer there is. There are several reasons for this. Less than 5% of all patients diagnosed with pancreatic cancer are alive five years after diagnosis. To put it another way, pancreatic cancer is the tenth most commonly diagnosed cancer but number four in the list of cancer killers. That’s because most (at least 80%) are diagnosed with unresectable and/or metastatic disease, for whom surgery cannot be performed. Given that the only currently known possible chance of long term survival in pancreatic cancer comes from a complete surgical resection of the cancer with negative surgical margins (i.e., no tumor at the margins of the surgical specimens and a rim of normal tissue between the margin and the tumor), any pancreatic cancer patient who is not a candidate for surgery has incurable disease. Of the minority of patients who do have their cancer completely resected surgically, the five year survival rate is better, perhaps in the range of 15-20% or so, but still the vast majority will be dead within five years, usually much less. Moreover, known as a pancreaticoduodenectomy or Whipple procedure, the surgery necessary to remove a pancreatic cancer in the head of the pancreas (the most common location) is a huge operation that involves removing the head of the pancreas and the duodenum and then reconstructing the connections between the bile and pancreatic ducts and the GI tract and establishing continuity between the stomach and small intestine. It’s a tour de force operation that often takes 8 hours or more and is fraught with the potential for complications, both short term and long term. However, for someone with a potentially resectable pancreatic cancer in the head of the pancreas, it is the patient’s only hope. Even so, after surgery, median survival times still only range from 12 to 19 months.
Because the outlook for pancreatic cancer, particularly unresectable pancreatic cancer is so grim (the median survival has barely budged from less than six months for decades. Median survival for untreated metastatic pancreatic cancer is on the order of 3-4 months, although gemcitabine chemotherapy regimens combined with radiation) can result in median survivals of six months or more. For locally advanced pancreatic cancer that cannot be resected but has not metastasized, the median survival is on the order of 6-12 months depending on the study. Thus, we can rightly say that pancreatic cancer is one of those cancers for which science-based medicine has frustratingly little to offer that can cure it. That’s not to say that science-based medicine doesn’t have a lot to offer for palliation, but no one wants just palliation. We all want to live to a ripe old age, not just have our pain and nausea palliated for a few months before cancer claims us. Even scarier is that pancreatic cancer usually produces few or no symptoms until it is fairly advanced. Usual symptoms include vague upper abdominal discomfort, loss of appetite, and post-prandial nausea from intermittent gastric outlet obstruction, you know, the sorts of symptoms that nearly of us have from time to time and that primary care doctors see in their practice every day. By the time a pancreatic cancer causes severe pain or obstructs the bile duct leading to jaunice, it’s usually unresectable or metastatic. (Often the reason it causes such severe pain is because it invades a plexus of nerves just posterior to the pancreas.)
It is the very deadliness of pancreatic cancer and the lack of effective life-saving or life-prolonging treatments for it that make pancreatic cancer a ripe condition for quackery. Rising above most other quackeries to attract a lot of attention about a decade ago is a quackery known as the Gonzalez protocol. It is described on Dr. Nicholas Gonzalez’s website as involving dietary changes, supplements, the replenishment of pancreatic proteolytic enzymes, and “detoxification,” including coffee enemas. It is not an easy therapy to undergo. For example, Dr. Gonzalez states:
Overall, cancer patients will consume 130-175 capsules a day, including nutrients as well as enzymes. Non-cancer patients might consume in the range of 80-100 capsules a day, the exact number depending on their health status and medical problems.
I know of no science-based cancer protocol that requires a patient to consume 150 pills a day. There are also the dietary alterations that can be quite hard to follow, as well as the frequent coffee enemas. All in all, the Gonzalez protocol is an arduous regimen for a debilitated pancreatic cancer patient to follow. Still, for some reason, it gained some popularity in the “complementary and alternative medicine” (CAM) community, so much so that, based on poorly designed case series of eleven patients, Dr. Gonzalez managed to get a clinical trial funded by the NIH to study his method versus standard chemotherapy, a sordid story that Dr. Atwood has chronicled in detail in a long series of blog posts. That trial ended in 2005.
So why is it 2009 before we have the results of this trial, which show that the Gonzalez protocol is worse than useless? The trial, begun in 1999, is known formally as Prospective Cohort Study of Gemcitabine Versus Intensive Pancreatic Proteolytic Enzyme Therapy With Ancillary Nutritional Support (Gonzalez Regimen) in Patients With Stage II, III, or IV Adenocarcinoma of the Pancreas. The results were finally reported in an E-pub ahead of print manuscript for the flagship journal of the American Society of Clinical Oncology, the Journal of Clinical Oncology, and entitled Pancreatic Proteolytic Enzyme Therapy Compared With Gemcitabine-Based Chemotherapy for the Treatment of Pancreatic Cancer, written by John A. Chabot, Wei-Yann Tsai, Robert L. Fine, Chunxia Chen, Carolyn K. Kumah, Karen A. Antman, and Victor R. Grann.
A CLINICAL TRIAL STUDYING FAIRY DUST VERSUS CHEMOTHERAPY
One reason why we emphasize prior probability so much here at SBM is because it matters. Nowhere have I seen it matter quite as much as in the genesis of a trial of the Gonzalez protocol for pancreatic cancer. The reason is that the “scientific” rationale behind these megadoses of supplements and pancreatic enzymes plus “detoxification” by coffee enemas is what Harriet Hall terms “Tooth Fairy science.” The Gonzalez treatment is basically a modification of a protocol known as the Kelley Treatment, which in turn was very similar to the Gerson protocol. Chabot writes in the introduction to this study:
The Scottish embryologist John Beard first proposed pancreatic proteolytic enzyme treatment in 1906 and soon after published a monograph, entitled The Enzyme Therapy of Cancer. In 1981, Nicholas Gonzalez began to evaluate the use of proteolytic enzyme therapy. Twelve years later, in 1993, he was invited to present a series of cases at the National Cancer Institute (NCI), which led him to undertake a case series of alternative medical therapy that included proteolytic enzymes, diet, nutritional supplements, and detoxification procedures. Among 11 patients with inoperable, biopsy-proven, stages II to IV pancreatic adenocarcinoma, he reported 81% survival at 1 year and 45% at 2 years. Four of the 11 patients survived for 3 years.
This is the very example of an implausible hypothesis. True, it’s not as implausible as homeopathy (few hypotheses are), but it goes against everything we know about cancer in general and pancreatic cancer in particular. There is no evidence that pancreatic enzyme deficiency has anything to do with pancreatic cancer, for example. Nor is there any evidence that this discription of the rationale behind the Gonzalez protocol, cribbed straight from the NCI website, has any relationship to reality:
Two major concepts underlie use of the Gonzalez regimen in cancer treatment. The first concept is that the pancreas, like the liver, is an organ that performs a detoxification function and that pancreatic enzymes help the body eliminate toxins and help normal cells repair damaged cells. The second concept is that cancer and most other human illness are related to physiological imbalances created by environmental toxins either consumed in food or contacted in the environment.[2,3]
Proponents of the Gonzalez regimen believe that toxins from sources such as processed foods and environmental pollution are responsible for human cancers. These toxins are thought to accumulate in tissues and over time create imbalances in the autonomic nervous system, diminish the normal immune response, and give rise to cellular damage that can lead to cancer.[2,3] If these toxins could be neutralized and eliminated from the body, proponents believe, both early and established cancers would be halted, and general health would be restored.[2,3]
Here’s the problem. These “toxins” are never identified, nor is there any evidence that the Gonzalez regimen actually removes them. It’s not that environmental exposures don’t have an effect on cancer susceptibility. Smoking can cause lung cancer and, ironically enough, increase the risk of pancreatic cancer as well. “Detoxification” quackery like the Gonzalez protocol takes science and turns it into Tooth Fairy science by giving near magical characteristics to these “toxins.” In any case, cancer is primarily a genetic disease. Even heavy smokers who smoke for 50 years “only” have about a 25% lifetime risk of developing lung cancer, which means more smokers don’t get lung cancer than do. In any case, think about it: Let’s say there is this host of unnamed, undefined “toxins” that give you pancreatic cancer. Remember, the toxins from tobacco smoke that predispose to lung cancer are largely known and quantifiable. Why would one think that coffee enemas would remove those “toxins”? Certainly there is no scientific basis to think that the special diet, the dozens of capsules of supplements and pancreatic enzymes, or the other aspects of the regimen would “detoxify” anything. Basically, the Gonzalez protocol appears to derive from a prescientific notion of disease that is almost religious in its nature blaming “contamination” as the cause of all disease and that one must “purge oneself” of this “contamination” to cure the disease.
But, say Gonzalez supporters, what about the case series of 11 patients with stage II to IV pancreatic cancer from the 1990s reporting 81% survival at 1 year and 45% at 2 years, with 4 of the 11 patients surviving for 3 years? As Dr. Atwood pointed out, this was a nonconsecutive case series, a so-called “best case” series. Basically, it’s a cherry-picked series, and I’ve criticized “best case” series before because they in essence intentionally look at outliers for whom the therapy may or may not have made a difference. Some patients with pancreatic cancer do survive longer than a year. Indeed, actor Patrick Swayze is one such patient, who has now been alive over a year and a half since being diagnosed with stage IV pancreatic cancer with liver metastases. Without knowing the denominator (i.e., how many patients Gonzalez treated to produce those 11 patients with significantly better than average survival for pancreatic cancer), his case series is meaningless. Indeed, back in my old Usenet days in the late 1990s and early 2000s, I would frequently comment that if I were to apply for a grant at the NIH to fund a clinical trial based on such a thin gruel of preliminary data, my application would have found its way to the cylindrical file. The trial for the Gonzalez protocol did not. Selection bias is a wonderful thing if you’re peddling unscientific cancer therapies, where patients who actually could undergo the rigorous Gonzalez protocol would be more likely to be doing better and thus more likely to continue to do better than average, therapy aside.
So what was this trial? These were the objectives:
- Compare the survival of patients with stage II, III, or IV adenocarcinoma of the pancreas treated with gemcitabine versus intensive proteolytic enzyme therapy and adjunctive dietary and nutritional support.
- Compare the quality of life in patients treated with these regimens.
The two arms of the study were described thusly:
This is an open-label study. Patients are stratified according to stage (II or III vs IV), performance status (0-1 vs 2) and nutritional status (well nourished or moderately malnourished vs severely malnourished). Patients are entered into 1 of 2 treatment arms at their choice:
- Arm I (Nutritional Arm): Patients receive pancreatic enzymes orally every 4 hours and at meals daily on days 1-16, followed by 5 days of rest. Patients receive magnesium citrate and Papaya Plus with the pancreatic enzymes. Additionally, patients receive nutritional supplementation with vitamins, minerals, trace elements, and animal glandular products 4 times per day on days 1-16, followed by 5 days of rest. Courses repeat every 21 days until death despite relapse. Patients consume a moderate vegetarian metabolizer diet during the course of therapy, which excludes red meat, poultry, and white sugar. Coffee enemas are performed twice a day, along with skin brushing daily, skin cleansing once a week with castor oil during the first 6 months of therapy, and a salt and soda bath each week. Patients also undergo a complete liver flush and a clean sweep and purge on a rotating basis each month during the 5 days of rest.
- Arm II (Chemotherapy Arm): Patients receive gemcitabine-based chemotherapy.
Quality of life is assessed at 0, 2, 6, and 12 months and then yearly thereafter.
Patients are followed at 1, 3, 7, and 12 months and then yearly thereafter.
Arm I is the very epitome of Tooth Fairy science.
A word needs to be said about the design of this study. It is not randomized, and it is not double-blinded. It was originally designed as a randomized but not blinded study, given the difficulty in blinding which group is getting coffee enemas, for example, and which were not. (I can see how placebo enemas would be a problem.) Even so, because placebo effects would not be evident in the objective outcome, namely survival, this would not be a major problem in a study of this sort, although the lack of blinding could conceivably affect the quality of life measurements. Unfortunately, investigators could only accrue three patients between 1999 and 2001; so in 2001, the trial protocol was converted from a randomized phase III design to an open label observational design. Basically, patients picked which therapy they wanted, and the investigators prospectively studied how they did. The figure below shows the ultimate flow of patients:
One serious problem with a nonrandomized design like this is the likelihood of selection bias. In this case, it would be “self-selection” bias in that investigators would have to worry whether patients who were either worse off or better off might opt preferentially for one arm of the trial over the other. I will address that point after we look at the results, which are quite striking:
At enrollment, the treatment groups had no statistically significant differences in patient characteristics, pathology, quality of life, or clinically meaningful laboratory values. Kaplan-Meier analysis found a 9.7-month difference in median survival between the chemotherapy group (median survival, 14 months) and enzyme treatment groups (median survival, 4.3 months) and found an adjusted-mortality hazard ratio of the enzyme group compared with the chemotherapy group of 6.96 (P< .001). At 1 year, 56% of chemotherapy-group patients were alive, and 16% of enzyme-therapy patients were alive. The quality of life ratings were better in the chemotherapy group than in the enzyme-treated group (P <.01).
Moreover, the quality of life, as measured by standardized surveys, was actually worse for the Gonzalez therapy group:
Patients in the two groups responded similarly to the questionnaires on quality of life before initiation of therapy, but the overall FACT-PA scores during 12 months decreased more in the enzyme group than in the gemcitabine group (Fig 3). Twenty-four percent of total measurements were missing. Quality of life scores of both groups were significantly different (P.01). During the first 6 months of the study, pain scores increased in the enzyme group, but they decreased in the chemotherapy group (P.05); however, few patients reported on use of analgesics. (Table 2).
In all my years in medicine, surgery, and surgical oncology, I have never seen a study with such a striking difference in outcome between the two groups. No wonder the Office for Human Research Protections (OHRP) issued a determination letter stating that it was appropriate to terminate the study before the full 72 patients were enrolled due to the study having reached its “predetermined stopping point.” These days, clinical trials are designed with periodic assessment of results and predetermined stopping points. These stopping points are invoked to shut down the trial in the event that one group is doing so significantly better than the other that enrolling the remaining patients could not possibly change the result. The purpose of such a stopping point is to protect clinical trial subjects from being enrolled in a study that no longer has clinical equipoise; i.e., the groups can no longer be predicted to have roughly equivalent outcomes based on what we know. Usually, it’s the experimental group that does better than the control group. At least that’s the intent. Sometimes, however, as in the case of the Gonzalez protocol, it’s the experimental group that did much worse, so much so that the study was halted before reaching its full accrual.
But it’s even worse than that. Not only was the median survival of patients in the Gonzalez therapy group worse than it was for the standard chemotherapy group, it was three times worse. At one year, 56% of the chemotherapy patients were alive; only 16% of the Gonzalez protocol patients were. But it’s still even worse than that for the Gonzalez therapy. Not only did Gonzalez therapy patients do worse than those receiving standard therapy, but they did worse than the “average” pancreatic cancer patient as determined by the survival curve derived from data from the SEER Database. The most likely reason to explain such a result is that the Gonzalez therapy is not just inferior to gemcitabine but is probably completely biologically inactive against pancreatic cancer. What we are looking when we examine the survival curve for the Gonzalez protocol group is, most likely, indistinguishable from a survival curve of untreated pancreatic cancer versus treated.
Is this study a slam dunk? Is the Gonzalez therapy, for all intents and purposes, dead? For the most part, yes, but there are a few problems. Of course, the nonrandomized nature of the trial makes it possible that biases crept in. As I mentioned before, that is always a potential problem with a nonrandomized study design. Fortunately, the patients in each group ended up being reasonably well matched in age, gender, and other demographics. There was, however, at least one relevant difference that no one discussed. Indeed, I’m surprised that the investigators didn’t mention it, as it could have at least partially suggested a reason why the Gonzalez group might have done so poorly early on. It’s unlikely to have made a difference big enough to have explained the huge difference between patient groups, but it would at least have been a straw to grasp at. Specifically, the serum albumin levels in the Gonzalez group were statistically significantly lower than they were in the control group. Albumin is a rough estimate of nutrition (although there are better parameters) and we surgeons know that lower albumin levels are associated with increased early mortality in pancreatic cancer. I doubt that this is anywhere near enough to explain such a huge difference in mortality between the two groups, especially since the difference between the albumin levels in the two groups was relatively small (4.3 g/dL versus 4.0 g/dL), with the Gonzalez group still having albumin levels within the normal range, but it is one difference that caught my eye. It is also likely (but not well explained) that there were differences in supportive care, such as the rate of inserting biliary stents to relieve bile duct obstruction or aggressively treating infections such as ascending cholangitis, which is one of the most common causes of death in pancreatic cancer patients. The obstructed bile duct causes bile to back up; the bile gets infected; and the patient develops a life-threatening infection. The primary treatment is biliary drainage, either through a tube inserted endoscopically into the bile duct or inserted through percutaneously into the bile ducts in the liver. Antibiotics are necessary, too, of course, but they don’t do much good without biliary drainage. Also, if there was a difference in palliative care between the two groups, that, too, would be yet another example of the unethical nature of this trial.
The one thing that surprised me about this study was not so much that the Gonzalez group had the same mortality and median survival as, in essence, untreated pancreatic cancer patients (which they are), but rather that the survival in the gemcitabine group was so long (14 months). This is better than most studies of gemcitabine-based regimens from the time period during which patients accrued to this study (although results of more recent trials have been better). This suggests a possibility of some unexpected selection bias. Alternatively, it could be that this is simply an outlier group. However, it’s highly unlikely that it invalidates the study. Given the striking difference between the two groups and the fact that the Gonzalez group did more poorly than historical controls, from this study we can confidently say that the Gonzalez protocol is almost certainly no better than no treatment.
In fact, it’s probably worse. Go back to Dr. Atwood’s post on this issue and consider the story of an unfortunate 40 year old man who enrolled in the trial and chose the Gonzalez protocol arm. This man was told to have his fillings removed and afterward tried as hard as he could to continue the regimen to the letter and suffered horribly as he did. At one point he was even told that increasing pain might be an indication that his tumors were “dissolving.” Although it may be that the Gonzalez protocol didn’t make his pain worse, it’s quite clear that the man’s pursuing the therapy kept him from persuing effective palliative care that might have made his last few months on this earth a lot less unpleasant. Indeed, this makes me wonder what the investigators did after the trial was closed. From an ethics standpoint, every patient in the Gonzalez arm should immediately have been informed of these results and been strongly encouraged to give up the Gonzalez protocol and undergo standard gemcitabine chemotherapy.
Finally, the most disturbing issue that this trial raises is the question of why it took nearly four years after the trial was stopped to publish the results. Dr. Atwood has speculated why this might be the case, namely to cover up the results that were unfavorable to Gonzalez. There is now no doubt that this trial was completely unethical right from the very beginning, but that lack of ethics was compounded by not having the results reported right away. In this, after having had a few days to think about it, I am going to have to strongly disagree with my co-blogger Dr. Atwood when he asserted, “A compelling argument can even be made that the JCO should not have published the report—as paradoxical as that sounds.” I can understand where he’s coming from. This study clearly violates the Helsinki Declaration, to which JCO requires the clinical trials that it publishes to adhere. However, I find an it equally, if not more, compelling to take the view that the patients who suffered in the Gonzalez group will have suffered in vain if the results of this trial were not published. I also tend to take the view that shining a light on such a trial, in the form of publication, is a good thing in this particular case because it eliminates the uncertainty over whether the results were as bad as we know them to be. Let’s put it this way: If the results were not published in a peer-reviewed journal but rather announced or mentioned in a report, they would seem less credible to shruggies. It is these people who need these bad results rubbed in their noses, the better to let the stench waft into their nasal passages and make them retch. When they ask, “What’s the harm,” we can tell them in no uncertain terms what the harm is. The same would apply if they were simply announced on ClinicalTrials.gov or mentioned in a report. It’s just not the same as a peer-reviewed publication in one of the highest impact oncology journals there is. This is truly a case where the greater good can be served by disseminating this data far and wide as soon as possible by the most scientifically credible means necessary.
One thing that is most remarkable is that this study was released without any fanfare whatsoever. You can bet that, had the results been positive, it would have been trumpeted with press releases. I bet that even if the results had shown that the Gonzalez protocol was no worse than “conventional” therapy, it would have been trumpeted as “Gonzalez therapy as good as conventional chemotherapy!” to the press. Yet, here we have exceedingly striking results, and what do we hear? Nothing.
TOM HARKIN, NCCAM, AND HEALTH CARE REFORM
I now return to the issue of Tom Harkin’s likely taking over Senator Kennedy’s old chair and why it matters. The reason it matters is because Tom Harkin is not a science-based man. He is the man most responsible for the creation of NCCAM and has in 2009 been doing his best to infiltrate coverage for CAM in any government health care reform legislation that may come. Yet that doesn’t stop organizations like The Foundation for the National Institutes of Health, Research!America, and the Albert and Mary Lasker Foundation from awarding Harkin a “Champions of Research” award. While it’s true that Harkin seems a strong supporter of the NIH on the surface, at the same time he is a corrosive influence, inserting his belief in CAM/IM into the NIH and pushing to promote it. Without NCCAM, Tom Harkin, and Dan Burton, the latter of whom in particular pushed the NIH to fund the Gonzalez protocol, this pseudoscientific and highly unethical study is highly unlikely ever to have been funded.
Now Tom Harkin, who has “distinguished” himself by trying to hijack health care reform to give undeserved legitimacy to therapies like the Gonzalez protocol and even to have the government pay for such quackery, is now in charge of one of the key committees involved with government health policy. I fear he’ll do for health care reform what he did for the NIH through NCCAM.
Be afraid. Be very afraid.
John A. Chabot, Wei-Yann Tsai, Robert L. Fine, Chunxia Chen, Carolyn K. Kumah, Karen A. Antman, & Victor R. Grann (2009). Pancreatic Proteolytic Enzyme Therapy Compared With Gemcitabine-Based Chemotherapy for the Treatment of Pancreatic Cancer Journal of Clinical Oncology : 10.1200/JCO.2009.22.8429 (E-pub ahead of print).