JAMA’s publication of the NIH’s Trial to Assess Chelation Therapy (TACT) has reignited a heated debate about the trial. The TACT investigators have generously agreed to respond to questions posed by CardioExchange’s Harlan Krumholz. The TACT investigators who participated in this interview are Gervasio A. Lamas, Christine Goertz, Robin Boineau, Daniel Mark, Richard L. Nahin, and Kerry L. Lee. (To read CardioExchange’s news story on the study, click here.)
Krumholz: How was the time to stop the study selected? Statistical significance seems to have been reached just as the study was halted.
TACT Investigators: The TACT protocol always specified that the last patient enrolled in the trial should have 1 year of follow-up (see design paper). Also, as clearly explained in the TACT report in JAMA and amplified in the Supplementary Online Appendix, the decision to enroll 1700 patients occurred in 2009, well before the end of the trial. Thus, the trial results were significant based on (a) enrolling the specified number of patients, (b) following them for the prespecified period of time, and then (c) closing the trial and analyzing the data according to the prespecified statistical plan. The timing of when the study stopped was pre-planned — it had nothing to do with whether or when statistical significance was achieved.
The primary endpoint has been criticized as “soft.” How was it chosen, and has it been used in other studies?
The primary endpoint consisted of time to the first occurrence of the following major adverse cardiovascular events: death, MI, stroke, coronary revascularization, or hospitalization for angina. Combined endpoints are used when the most important endpoint to provide proof of efficacy, all-cause mortality, is likely to occur so seldom that an impractical and unaffordable number of patients and patient-years of follow-up would be needed to have enough statistical power to detect a difference between groups. In a stable coronary disease population, the incidence of mortality is low, so a combined primary endpoint is necessary for most studies. As we were planning the study, we selected a population very similar to that of the WIZARD study, which used a primary endpoint similar to ours. Other studies with similar combined endpoints, particularly including revascularization, are JUPITER and ACCELERATE.
How does the unexpectedly large number of patients who withdrew consent affect the study’s findings?
Simply put, patient withdrawal limits the observation time for the patient, as well as the likelihood that an event will take place and be detected. Consequently, it limits statistical power to detect a difference between groups. In TACT, more placebo patients than active-infusion patients withdrew consent. This complicates the situation, as more actively treated patients are likely to have their endpoints counted, and there was less follow-up time to detect and count placebo events. Thus, the effect on the study is to push the results toward the null. Yet, in spite of this rather daunting problem, the intent-to-treat analysis of TACT trial was still positive.
The editors at JAMA asked us to perform sensitivity analyses in which we would make assumptions regarding what might have happened to the missing patients and how that might have altered the study results (see eTable 8 of the online Appendix). I quote the conclusion in p. 32, although the full text and table are well worth reading.
“Based on the other data observed in the trial and because the baseline risk factors of the patients who withdrew consent were very similar in the two arms of the trial, the most plausible scenarios in eTable 8 are those where the percentages of events among the withdrawn or lost patients in the two arms are nearly equal or slightly favor the active arm. However, the comparison of the two arms remains significant at the 0.036 level if the relative increase of events in the active arm is as much as 20% higher than in the placebo arm, and even generally if the percentage of events in the active arm is 25% higher than in the placebo arm. The hazard ratio for all of these scenarios remains in the range of 0.80 to 0.84, the p-values are quite robust, and significance of the treatment effect is maintained, not only for the scenarios for the withdrawn or lost patients that would be considered most plausible, but also for scenarios that are unfavorable to EDTA chelation.”
Was the study design changed midway through the trial?
The basic study design was never changed. The sample size was changed while maintaining statistical power by lengthening follow-up time. Thus, the key protocol feature, 85% power to detect a 25% difference, was unchanged throughout the study. Changes in sample size are fairly common in clinical trials. Other recent clinical trials studying coronary disease patients, such as OAT and FREEDOM, reduced the sample size due to difficulty in implementation or recruitment. Their results have not been questioned.
Some people have raised ethical questions about the study. Specific comments include failure to conduct the investigation in accordance with the signed statement and investigational plan; to report promptly to the IRB all unanticipated problems involving risk to human subjects or others; to prepare or maintain adequate case histories with respect to observations and data pertinent to the investigation; and to keep good records on investigational drug disposition with respect to dates, quantity, and use by subjects. Can you set the record straight on these issues?
The above comments and questions link to an FDA site visit to a TACT site. The principal finding was that the delineation of responsibility log was not filled out correctly. The study coordinator who was obtaining consent was not listed on the log. This led to a daisy chain of findings reported by the FDA. I want to point out that at no time were patients endangered. A corrective action plan involving retraining and site visits was put in place. The FDA’s concerns were resolved, and the site continued study activities. We invite you to read in detail the online content in Dr. Bauchner’s editorial, which describes many of these issues in much greater detail.
Much of the criticism and controversy surrounding TACT was originated by several small, self-appointed groups with a history of aggressive opposition to CAM practices and to TACT in particular, as they feel that 1) CAM practices should not be studied; 2) if studied and the results are negative, the trial should be condemned as a waste of money; and 3) if studied and the results are positive, the trial should be condemned as incompetently and unethically carried out, and the results discounted. Obviously, TACT has received the 3rd response. We ask our cardiology colleagues to look at the study critically, without emotion, and ask themselves if they would feel the same way about our results if the words EDTA chelation never appeared, and instead “stem cell” or “new statin” appeared. We believe that the debate should focus on the unexpected biological activity of chelation therapy rather than on spurious allegations that TACT investigators were involved in willful wrongdoing that affected the results of the trial.
A major comment concerns the components of the chelation therapy, specifically about the inclusion of procaine and heparin and their possible effects on cardiovascular outcomes. Why were they included, and do you think they had an effect?
The NIH RFA to which we responded had gone through NHLBI and NCCAM Councils and called for a definitive trial of EDTA chelation therapy, as it was currently being implemented in clinical practice. When we looked into the infusions, we found that they included many compounds, not just EDTA. Therefore, to achieve a result that would be consistent with how chelation therapy is actually used in practice, we chose to mimic precisely the most prevalent infusion in use. We have no reason to believe that a small amount of procaine or 2500 U of unfractionated heparin once weekly would affect outcomes to the extent that we found in TACT.
Another comment is that the placebo solution contained 1.2% glucose in order to match the osmolarities of the control and experimental solutions. Some people think that might have contributed to worse outcomes in the control group. What is your view of that possibility? What options did you consider for the placebo infusion?
We wanted to keep the placebo solution as simple as possible and not introduce an unexpected risk or benefit. Normal saline fit the bill, as we excluded patients who had active heart failure. The amount of glucose in 500 mL of 1.2% is 1.2 grams X 5 = 6.0 grams of glucose weekly. It is not plausible that this would lead to a greater coronary risk in diabetic patients. We also considered other iso-osmolar solutions such as D5W but felt these would have introduced a greater sugar load. Another option would have been a solution that encompassed all the ingredients except EDTA. However, as stated above, our intent was to investigate a treatment currently in use and determine whether it was safe and effective, rather than deconstruct the solution. The present design allows us to draw those conclusions.
Some people have claimed that unblinding might have biased the trial. Is there evidence of unblinding? Here is what one person wrote: “First of all, the chelation mixture is not stable and therefore must be mixed at the local site. The ascorbic acid, which must be injected into the mixture, is yellow in color and highly viscous. The work around for this problem is to cover ascorbic acid, chelation mixture and placebo syringes and bags with tinted translucent tape and to add concentrated dextrose solution to the placebo syringe to make it as viscous as the ascorbic acid.”
All necessary precautions were taken to ensure blinding was preserved in the study. Both the placebo and active infusions were shipped in identical refrigerated containers. Neither EDTA nor ascorbic acid is stable if shipped mixed with the other 8 components of the chelation solution. Thus, the shipped and refrigerated pack contained one syringe with EDTA (or placebo), one syringe with ascorbic acid (or placebo), and a bag for intravenous infusion with all the other components already mixed (or normal saline only).
The ascorbic acid solution has a pale yellow color, which upon mixing becomes indistinguishable from the clear saline placebo solution. The syringes containing ascorbic acid or placebo were covered in translucent yellow adhesive tape, thereby blinding the pale yellow color of ascorbic acid. In addition, ascorbic acid, in the concentration provided by the manufacturer, is viscous. The placebo ascorbic acid solution contained enough 50% dextrose to mimic the viscosity of the active ascorbic acid solution.
In his editorial, Steve Nissen complained that the sponsors had access to the results during the course of the study. Is this atypical for an NIH trial?
The TACT trial was conducted under a policy that allowed NHLBI and NCCAM program staff access to unblinded data. At all times there was a firewall between trial decision makers who were blinded and NHLBI staff who were unblinded, to prevent the possibility of improper influence by unblinded staff. NCCAM and NHLBI staff were unblinded primarily to assess safety issues with the intervention, not efficacy.
The goal of the Program Officers and other NIH personnel is to safeguard the integrity of the trial, not to sway the results either way. There is no indication of improper influence exerted by NHLBI or NCCAM program staff who saw unblinded data.
What was the result of the Office of Human Research Protections (OHRP) reports? Were there serious ethical breaches in this trial?
No serious ethical breaches occurred in the trial. In 2008 a self-appointed group of individuals, none of whom can be considered clinical trialists, with a record of strident opposition to CAM and to the study of CAM wrote to OHRP to complain about the scientific basis of TACT, the consent form, the investigators, and whether patients were properly notified of new developments. Given the misinformation disseminated on the internet and in print media, we want to emphasize that the study was never suspended by OHRP, NIH, or the DSMB. The study management team halted infusions for 1 week while the OHRP documents were examined and a plan of action developed. Infusions restarted a week later. New enrollments restarted 4.5 months later after a new consent form was approved and a Dear Patient letter was sent. The OHRP investigations closed on October 2009. Greater detail can be found in the online eAppendix that accompanies Dr. Bauchner’s editorial in the same issue of JAMA.
Was there evidence that the CAM sites cheated?
There is no evidence whatsoever that CAM sites cheated. In fact, we compared the effect of study therapy on patients enrolled in CAM vs. non-CAM sites (bottom of Figure 2 in the paper). The p for interaction was not significant.
Were there investigators who had violated the law – and how might unethical behavior by individuals who strongly believed in CAM have altered the trial? Is this any different from other trials where investigators are invested in the success of the intervention – or did this trial have specific issues that must be considered?
All investigators had an unrestricted license to practice medicine in their states, and they received human-subjects training, protocol training in person and online, research training, IRB approval, in-person site visits, and electronic data monitoring by the Data Coordinating Center at Duke.
All investigators brought strengths and weaknesses to TACT. Chelation sites had the clinical experience and infrastructure to administer chelation. Conventional sites had research experience but did not have the clinical experience and infrastructure to administer chelation. The purpose of training the sites was to develop a cohesive team with blended abilities. The lack of interaction of site-type with treatment bears this out.
Physicians who have an opinion and are invested in a treatment or technology frequently are open-minded enough to participate in clinical trials. If this were not the case, then clinical trials comparing PCI or surgery to each other or to medical therapy could never take place.
In their paper, the TACT investigators have taken a very cautious perspective, defending the trial but arguing that it should not influence practice. What was the point of doing the trial if a positive result is not enough to influence practice?
We were cautious scientists when we started, and we are cautious scientists now. We do not believe that we have enough data yet to recommend the routine use of chelation therapy for all post MI patients. We are continuing our analyses, now of the vitamin randomization and of the 4 factorial groups. When we finish and publish, we will be happy to revisit this question with you.
Did patients enrolled in TACT choose chelation above standard, evidence-based post-MI therapies?
No. In fact, the statistically and clinically significant benefit detected in post MI patients occurred in addition to maximal, but imperfect, real-world use of evidence-based post MI medications. In addition, chelation therapy did not interact with these medications, possibly suggesting a new mechanism of efficacy. This is an exciting finding that deserves further exploration.
- Gervasio A. Lamas MD
- Christine Goertz DC PhD
- Robin Boineau MD MA
- Daniel Mark MD MPH
- Richard L. Nahin PhD MPH
- Kerry L. Lee PhD