Does the degree of efficacy is depend on the time at which it is measured? Apparently so. The case of psychological support and breast cancer longevity again.
After an original 1989 report of positive effects on metastatic breast cancer, by 2006- 7 the majority of RCTs on such effects had settled the issue in the negative. This was only after 20 years of repeated research grants and RCTs based on hunches and feelings that somehow emotional support really affected the course of cancer. Investigations continued despite analyses showing the few original positive studies had been so flawed in design or defective in reported details, that they should have been dismissed and perhaps excluded from systematic reviews. (Spiegel D, Bloom JR, Kraemer H, Gottheil E. Psychological support for cancer patients, Lancet ,1989 Dec 16;2(8677):1447., Fawzy FI, Fawzy NW, et al. Malignant melanoma. Effects of an early structured psychiatric intervention, coping, and affective state on recurrence an survival 6 years later. Arch Gen Psychiatry. 1993 Sep;50(9):681-9.)
But to advocates, conflicting results served as motive to prove the claims by repeating the studies for 20 years, “doing them right this time.” As of mid-2008, consensus was the issue was still “negative.” Now another study, claimed to be positive, makes the news.
Here is the part of the abstract of the latest study, published in the December 15 issue of the journal Cancer:
METHODS. A total of 227 patients who were surgically treated for regional [stage 11-111] breast cancer participated… Patients were randomized to Psychologic
Intervention plus assessment or Assessment only study arms. The intervention
was psychologist led; conducted in small groups; and included strategies to
reduce stress, improve mood, alter health behaviors, and maintain adherence to
cancer treatment and care….
RESULTS After a median of 11 years of follow-up, disease recurrence was reported to occur in 62 of 212 (29%) women and death was reported for 54 of 227 (24%) women. Using Cox proportional hazards analysis, multivariate comparison of survival was conducted. As predicted, [emphasis added] patients in the Intervention arm were found to have a reduced risk of breast cancer recurrence (hazards ratio [HR] of 0.55; P 5.034) and death from breast cancer (HR of 0.44; P 5.016) compared with patients in the Assessment only arm. Follow-up analyses also demonstrated that Intervention patients had a reduced risk of death from all causes (HR of 0.51; P 5.028).
CONCLUSIONS. Psychologic interventions as delivered and studied here can
improve survival. Andersen et al, Cancer 2008;113:3450–8.
What does the statement mean that as predicted, the support group had a reduced risk of recurrence and death? There was too little evidence on which to predict a positive outcome at the time of the trial’s conception. (In a later interview, the senior author made this contradicting statement: “Nobody was more surprised with the results than we were,” said study lead author … in an interview with Medscape Oncology.” (http://www.medscape.com/viewarticle/583889) How is one to interpret the authors’ conclusions? Despite the fact that realistically, most investigators have an outcome in mind, such a prediction (as expected) is a bit odd.
Nevertheless, let us assume that bias did not affect patient assignment or affect objective outcomes such as recurrence and death. One nevertheless – in order to gain perspective on the results – can examine the raw data.
Each arm of the study contained slightly over 100 subjects. At the median of 11 years recurrences were 33 in the control group, 29 in the support group, a difference of 4; deaths from breast cancer were 19 in the support group, 25 in the controls, a difference of 6. I submit that the difference on both counts (and they are not independent outcomes) was not practically significant. A difference of initial assignment of 1 or 2 cases from the support to the control group would have altered the outcome significantly.
The authors presented the results in a manner that tends to magnify differences. They compared the differences in number of recurrent disease patients rather than relaying the differences between the overall percentage of patients. They concluded that psychosocial support reduced the risk of recurrence by 56 percent, and reduced the risk of death by 45 percent. That is quite a difference, but is similar to a result stated as a risk of death being reduced 50 percent by reducing the number of deaths from 2 to 1.
The graphed results showed 75 treatment subjects free of disease and 65 control patients disease free – a difference of four and six patients in a pool of slightly less than 200 who completed the study – hardly impressive – at least not as impressive as reported.
Furthermore, in the treatment group 10 patients were lost to follow-up, while in the controls 5 were lost. Given the small numbers being dealt with here, the difference leaves yet another potential for an unbalanced result – due to dropouts. Unbalanced dropouts leave unanswered the questions as to why a patient would leave a treatment arm of a study – knowing one had already had “the maximum”? Would one want to seek treatment elsewhere when suspecting a recurrence? The significance of the results seem to diminish with each question.
A side issue is that the authors state in their rationale for doing the trial that stress is associated with increased death rates, so that perhaps reducing stress would modify that trend. But both arms of this study had 5 deaths from other cancers (3) and other causes (2) – evidence against that rationale. But the authors did not comment on it.
The illustrations of the results added to magnification. The two graphs showing recurrence rates and death rates showed only the scale from 65 percent to 100 percent, thus magnifying the visual impression of the differences (e.g., 75 v 65.) The two lines of each graph would have been portrayed much closer together if shown on a scale from 0 to 100 percent. And again, a randomization assignment difference of just one or two subjects in either scale would have made the curves close to identical (a difference of from 4 to 3 or 2.)
The small difference between the two groups should point one toward examining any differences in initial risk factors for recurrence such as receptor status, stage, type, grade, and size, any of which could affect outcome. My first run through showed such minor differences that did not draw attention. However a second look for minor differences showed several that favored the treatment group; more benign cell type and more benign grade. They differed by several percent (2-3.) Receptor status, size were close to equal, and node status ( 2 v 3) favored the controls. None varied more than for about 2-4 subjects, but the differences in grade and cell type characteristics could well have explained part of the difference in outcomes.
The authors presented the current status of the question (effect of psychosocial support on cancer progression) as conflicted in the literature. In fact, once the various reports were evaluated for quality one could not conclude efficacy. The first two studies purporting to show positive effects of psychological support on breast cancer (Spiegel et al, Lancet 1989, Fawzy et al, Ann Psychiat 1995) were analyzed and found to have shared problems of abnormal control groups, retrospective analysis, and small numbers (Fox BH, A hypothesis about Spiegel et al.’s 1989 paper on Psychosocial intervention and breast cancer survival. Psychooncology. 1998 Sep-Oct;7(5):361-70. Sampson W, Errors in Alternative Medicine Literature. (Abst) Proc. AAAS 1997; Sampson W, Cancer and the Mind… Seminars in Oncology (Suppl), Dec 2002.)
In their discussion, the authors speculated on mechanisms for their positive findings, but did not explore the above artifacts, nor other explanations of chance findings. They speculated about peripheral possibilities for which little to no significant evidence exists – immune function, cytokine secretions, and other speculative intermediate mechanisms. That is as much as I could glean from the actual report. Others may see more in it, or may disagree. But the story goes on.
A Reuters dispatch on the report played up the quoted survival difference, as follows:– Psychological group therapy for women with breast cancer may help them not only to cope better with their disease but also live longer… The idea that such therapy can extend survival in cancer patients has been controversial for two decades. Past studies have yielded conflicting results.
… After 11 years, the women who participated in the group therapy were 56 percent less likely to die of breast cancer and 45 percent less likely to have their cancer return, the researchers wrote in the journal Cancer. “– and that’s a huge health outcome,” [the author] who helped lead the therapy groups, said in a telephone interview. By repeating the magnified 5 and 56 percent presentation, Thus the reporter leads a casual reader to believe there is a highly significant effect of psychology support on cancer recurrence and survival.
Medscape offered a summary for continuing medical education with highlights, author quotes, and the following conclusions:
For cancer patients, studies have shown that stress-related psychosocial factors have led to higher cancer incidence in initially healthy people, poorer survival in patients diagnosed with cancer, and higher cancer mortality rates. Psychologic interventions demonstrated a reduction in the risk for breast cancer recurrence, death, and risk for death from all causes.
Offered for continuing medical education (CME,) the article had no reference to the analyses of the subject of stress and cancer by James Coyne PhD of Columbia University, who analyzed the claims and found them nearly baseless, nor was there reference to the fact that the literature on psych support on cancer survival is not just conflicted; the positive evidence is flawed, and the major initial positive studies were repeated several times with negative results (Coyne JC, Palmer SC. Does Psychotherapy extend survival? Some methodological problems overlooked. J Clin Oncol. 2007 Oct 20;25(30):4852-3; )
Failure of replication is evidence of disproof. Unless rules of science have been changed…
Neither the Reuters nor the Medscape report included the statement by Michael Stefanek, PhD, spokesman for the American Cancer Society, which co-funded the study: “We should not conclude that psychological interventions increase survival among women with early stage breast cancer. The results were quite modest, […] and may well have been due to small differences in medical attention received between the groups [….]
Our reasons may differ, but the conclusion is the same.