Alastair J. Cunningham, Cathy Phillips, Gina A. Lockwood, David W. Hedley, and Claire V. I. Edmonds
Alastair J. Cunningham PhD is a senior scientist Cathy Phillips EdD and Claire V.I. Edmonds PhD are research associates, Gina A. Lockwood MMath is a biostatistician, and David W. Hedley MD is an oncologist all at the Ontario Cancer Institute/Princess Margaret Hospital, Toronto, Ontario, Canada.
*This study was supported financially by the J.S. Trust and the Princess Margaret Hospital Foundation. We would also like to thank the following oncologists for providing survival estimates for our subjects: Dr Rachael Barstow, Dr Rob Barton, Dr Andrea Bezjak, Dr David Bloomfield, Dr Charles Catton, Dr Pamela Calton, Dr Bernard Cummings, Dr Anthony Fyles, Dr Mary Gospodarowicz, Dr Peter Kirkbride, Dr Normand Laperriere, Dr Fei-Fei Liu, Dr Warren Mason, Dr Sue-Anne McLaughlin, Dr Malcolm Moore, Dr Amit Oza, Dr Lillian Siu, Dr Jeremy Sturgeon, and Dr Padraig Warde, all of the Princess Margaret Hospital.
Correspondence can be sent to:
Dr. Alastair J. Cunningham
Department of Epidemiology, Statistics and Behavioral Science
Ontario Cancer Institute/Princess Margaret Hospital
610 University Avenue
Toronto, Ontario, Canada
There have been numerous anecdotal claims that when patients are dedicated users of a variety of psychological self-regulation strategies, including relaxation, mental imaging, cognitive restructuring and meditation, such dedication may have a life-prolonging effect. Our aim was to test this possibility more rigorously, in patients with metastatic cancer.
A prospective, longitudinal, correlative study was carried out on 22 patients with varying kinds of medically incurable metastatic cancer. The intervention was one year of weekly group psychological therapy. Extensive verbal data (patients' written homework and therapists' notes) were collected over the year. The extent of each patient's involvement with psychological work was estimated following a qualitative analysis of these data. Patients were classed as showing high, moderate, or low involvement on the basis of a quantitative rating of categories defined by the analysis. These three subgroups did not differ significantly in their expected median survival duration as estimated from independent quantitative predictions by a large panel of oncologists who analyzed the patients' medical charts at time of study entry.
A significant relationship was found between degree of involvement in psychological work and survival duration. Results are presented as Kaplan-Meier survival curves (Fig. 2; P = 0.006, Log Rank test) and as a graphic display of the median survival of each of the three groups (Fig. 3). The main likely confounders (medical status, age, quality of life, and attendance at therapy) were similar across subgroups and did not change the relation between psychological work and survival duration. Limitations in the design are discussed. However, the strong effects observed support clinical observations that dedicated involvement in psychological self-regulation may prolong the life of some patients with metastatic cancer. © John E. Fetzer Institute 2000
The idea that life may be prolonged in some cancer patients as a result of the practice of psychological self-help techniques has been the subject of numerous anecdotal claims in the trade press over the last few decades (Simonton et al. 1980; Siegel 1986). While such claims may be a source of hope to patients, health care professionals are often concerned that this hope may be unfounded (Doan & Gray 1992). The existing scientific literature on this subject is inconclusive, and consists of two main streams. The first is a series of older reports attempting to correlate patients' attitudes and use of psychological techniques with their subsequent survival (Simonton et al. 1980; Newton 1982; Meares 1980). In all of these studies, the psychological characterization of patients was minimal or impressionistic and medical documentation was lacking; thus, it could not be ruled out that those patients who survived longest had the least serious illness. For example, Newton (1982) compared the survival of patients who attended 10 or more hypnosis-based behavioral training sessions with those who attended more than 3 but less than 10 sessions, and reported that those attending more sessions lived longer. A second, small literature also exists on so-called "remarkable survivors," describing qualities common to patients who appear to have outlived their expected duration of survival (Kennedy et al. 1976; Achterberg et al. 1977; Roud 1986; Pennington 1988; Huebscher 1992; Berland 1995). Psychological descriptions of varying thoroughness have been offered in these studies, but they uniformly lack adequate medical documentation, and have the added disadvantage, common to retrospective analyses, that one cannot assess the frequency with which similar psychological qualities were exhibited by other patients who did not survive longer than expected.
In more recent times, results from a number of randomized controlled trials, initially carried out to investigate the impact of psychological therapy of various kinds on various aspects of quality of life, have subsequently been scrutinized for evidence of effects of the therapy on survival. Some positive results have been noted (Spiegel et al. 1989; Richardson et al. 1990; Fawzy et al. 1993; Ratcliffe et al. 1995), and some negative (Linn et al. 1982; Ilnyckyj et al. 1994). However, the three published trials that have been designed specifically to test the hypothesis that group psychological therapy might prolong life (one case control study, Morganstern et al. 1984, and two recent randomized trials, Cunningham et al. 1998 and Edelman et al. 1999) gave null results. It is apparent that further, and preferably larger, randomized trials will be needed to determine whether psychological therapies affect the mean survival of groups of cancer patients.
The assumption underlying attempts to prolong life through psychological and spiritual self-help work is that the mind may affect the rate of cancer progression via neurophysiological mechanisms. It seems now generally accepted that the rate of progression of a cancer is influenced by regulatory Methods mechanisms in the tissue (Schipper et al. 1995). To bring about the necessary neurophysiological changes by voluntary action, considerable psychological change may be required. We hypothesized that the size of effects on survival would be proportional to the amount of effort and change patients made. A subgroup making strong efforts of this kind might not be detectable in a comparison of means (as in a randomized trial) but might be demonstrated with a correlative design. In effect, this shifts the focus of the investigation from the presence or absence of an intervention to what individual patients do with it, that is, on the relationships between individual behaviors or attitudes and outcome.
In an attempt to overcome the limitations of the earlier anecdotal reports (which were also basically correlative in nature), we used a prospective, longitudinal design, with careful medical documentation of the medical status of patients at the time of study entry. To obtain a detailed psychological description of our subjects, and one that made minimal assumptions about the qualities that might prove important, we carried out a qualitative analysis of extensive data (patients' written homework and therapists' notes) collected weekly over a year for a relatively small number of patients (22). Relevant qualities defined by this analysis (see Fig. 1) were then subjected to a rating process to convert the qualitative data to a score. Specifically, a panel of psychologists assessed each subject's verbal and written material and then applied numerical estimates to denote the apparent strength or intensity of the relevant qualities. This was related to survival, as described below.
A detailed account of this experiment, together with analyses of the relationships between a number of psychological variables and survival, has been presented elsewhere (Cunningham et al. 2000). The present paper is a brief, further analysis of the psychological data from the same patients with the survival data updated by 14 months and with a different emphasis: In this paper we have divided the patients into three subgroups, according to their involvement in their self
The 22 individuals constituting the study sample (Table 1) were patients who self-selected to enroll in a group psychotherapy program, and who met the following medical criteria: diagnosis of medically incurable metastatic cancer as assessed by a panel of oncologists from the patients’ charts at the time of entry, and sufficient health to attend the therapy groups for at least two months. The most common sites of the original cancers were breast, colon or rectum, and pancreas, and the interval since diagnosis was variable. While this sample was sufficiently motivated to undertake this program, the members of the sample were consecutively enrolled over 3.5 years; that is, there was no further selection exercised by the investigators, beyond requiring that the patients met the medical criteria. The use of a prospective design is in sharp contrast to the retrospective studies reviewed above. All but 5 of the patients were female, all were between the ages of 31 and 67, and most had better than average education.
The experiment was conducted in a large metropolitan cancer hospital. As described in more detail in the earlier account (Cunningham et al. 2000), the intervention was a year of professionally
Collection and quantitative rating of psychological data:
The database consisted of written homework provided by the patients (often 100 pages or more over the year) and detailed notes by the therapists made during each session and at interviews (typically 3) with each individual throughout the year of contact. Using standard techniques of grounded, qualitative analysis (Schatzman 1991) of the combined homework and therapists' notes, a number of psychological qualities ("categories") were defined, and related to one another in a model. This analysis is more fully described elsewhere (Cunningham et al. 2000). Sixteen of these categories were considered particularly relevant to the patients' involvement in self-help work. They fell under four headings, or themes, which are shown in Figure 1: (1) the patients' appraisal of the threat posed by the cancer and of the need to make adaptive changes; (2) the patients' intrinsic ability to act, or openness to change (as opposed to a defensive clinging to old habits); (3) the patients' expressed motivation for exploring new patterns of behavior; (4) the actual extent and dedication of the patients' practice of psychological strategies. We developed quantitative ratings, on a scale of 1 to 5, for each of the 16 component categories, with written descriptions of patterns that would qualify as "1", " 3", or "5." An assessment of the strength or intensity of that category could now be made for each subject. As an example, in the category of "dedication to self-help work," a rating of "1" was given when the subject was not at all committed and one of "5" when the work was the top priority in life, occupying several hours a day. The ratings were assigned following a team discussion of the data among 3 to 4 psychologically trained raters, 2 of whom knew the patients and 1 to 2 who did not; all had to be convinced that the data supported the rating. The sum of these 16 ratings was termed "involvement in self-help," and was expressed as a percentage of the possible total score. A limitation of this exploratory study was that the ratings were done a year or more after enrolling the subjects, and many patients were known to be dead before their data was rated.
For statistical analysis, the subjects were ranked in order of their involvement in self-help, and divided, by rank, into 3 approximately equal subgroups. Those with the top 8 scores (ranging from 95 to 61%) were termed "high" involvement (H) individuals; they made healing the top priority in their lives, and typically devoted 2 or more hours each day to practicing the techniques. The next 7 (58 to 46%) were designated "medium" (M), all demonstrating some degree of commitment to the methods, but less than the H group. The lowest 7 (45 to 34%) were designated "low" (L) and showed relatively little involvement in psychological techniques for helping themselves, being from the start either relatively apathetic or dismissive of the idea that they could influence the course of the disease.
In line with the literature on the relationship of quality of life and related variables to outcome (Coates et al. 1992; Wislott et al. 1997), we administered the Functional Living Index for Cancer (Schipper et al. 1984), a global quality of life instrument scored on 22, 7-point Likert scales, to 17 of the 22 subjects at entry into the study. It was not obtained for 2 patients in each of the H and L subgroups and 1 in the M subgroup. This index was part of a larger psychometric package described elsewhere (Cunningham et al. 2000).
To counter the possibility that any differential survival of the subgroups was caused by differences in severity of disease, it was obviously important to show the equivalence of the groups in terms of their main prognostic factors at the time of entry. We therefore obtained, from members of a panel of oncologists, estimates of the probable survival duration of each subject, based on blind chart review of the patients' clinical histories and prognostic factors at the time of study entry. The study population was heterogeneous, and the panel therefore had to be satisfied that they had received on each individual patient all relevant prognostic factors (culled by an oncologist co-investigator-DH-who had not met the patients and was blind to their progress). From 9 to 14 independent predictions for each patient were obtained. A median value for each patient was calculated from the estimates. (More details of this process are given in Cunningham et al. 2000).
To validate the accuracy of the panel, a separate study was conducted in which a sample of deceased patients, who had not received the psychological intervention, was randomly selected from the hospital's records (Barbera et al, submitted for publication). These patients were selected to approximately match the diagnoses and age range of the study population, and their individual survival durations were predicted in the same way as with the study's patients and compared to their actual survival (shown in Figure 3, alongside the predicted and observed survival of the patients in the study).
The significance of differences between patient characteristics of the three sub-groups in age, duration of attendance at the therapy, and quality of life as assessed by the Functional Living Index for Cancer (FLIC), was tested using Kruskal-Wallis tests, and the medical panel estimates of survival were compared using a Savage test (see Table 1).
The median survival times of the group were compared using the Log Rank test. The Kaplan-Meier survival plot is presented in Figure 2. Cox regression was used to test for an observed survival difference between the three groups while controlling for age, duration of attendance, baseline FLIC, or medical panel estimates of survival. Only one variable was controlled for at a time because of limited sample size.
The comparison of subjects' median age, attendance, baseline FLIC total scores, and medical panel estimates of survival is given in Table 1. None of these variables differed statistically across groups.
Figure 2 shows the Kaplan-Meier survival plots for the 3 subgroups (P = 0.006 by Log Rank test). Figure 3 plots the median survival times and interquartile range from the Kaplan-Meier survival analysis along with the corresponding median medical panel estimates of survival for visual comparison. The predicted and observed survival times are also given for the separate group of deceased patients who were assessed in a similar manner. As Table 1 reports, 3 of the H subgroup were alive at the time of writing (and therefore the upper bound of the interquartile range of the H group has not yet been reached), and 1 of the M subgroup, while all the L subgroup were dead. Two patients in the H group have had complete remissions of metastatic disease lasting 6 years (one with breast cancer, one with malignant melanoma). The great range of survival in the H and the M subgroups is indicated in the figure, and contrasts with the limited range in the L group and in the medical panel's estimates of survival.
The 3 groups remained statistically different in survival outcome in the Cox regression analyses when controlled for age, attendance, baseline FLIC, or medical panel estimates of survival (P < 0.01 in all cases).
This study is an investigation of a phenomenon often observed by clinicians (including ourselves) concerned with helping cancer patients cope better with their disease: Strong personal involvement of patients in psychological work appears to be associated with longer survival duration. To document this, we did not choose a traditional, clinical trials approach, where the focus is on the intervention, and where psychometric questionnaires are used to characterize patients psychologically. Instead, our focus was on what the individual patients did, and how this related to survival, for which a correlative design was needed. Rather than assume we could capture relevant subject behaviors by a set of predetermined questions in standard psychometric tests (which have, by and large, previously failed to consistently predict survival duration), we conducted an exhaustive, case by case, qualitative analysis of data obtained weekly, over a year, from and about each patient. This process developed a very detailed and rich description of the experimental subjects' behaviors and attitudes that no objective psychometric testing can achieve, but at the price of limitation in the numbers of individuals who could be studied.
While our study thus falls into the tradition of correlative descriptions of the mind-cancer link, a number of improvements over previous accounts of this kind have been incorporated. The first of these is the prospective, longitudinal design, with subjects in therapy, that is, rather than looking back at what survivors claimed to have done, we followed all the eligible individuals who were enrolled in the therapy over a period of time, and documented their attitudes and behaviors as they were struggling with their disease, relating these characteristics to eventual outcome. The second improvement was thorough medical documentation. The eligibility of patients to participate in the study was assured by medical chart review (most of the early studies relied on patient self-report), and the actual median survival of each subgroup was compared with predictions made through blind rating by an expert panel. The third improvement was the qualitative analysis followed by a quantitative rating of the strength of characteristics related to the involvement of patients in psychological self-help.
The main finding of the study is that the data appear to show an association, not necessarily causal, between such involvement and longevity.
Since this conclusion is likely to be controversial, it is important to examine threats to its validity. The most important limitation of the experiment is that the assessment of involvement had to wait upon collection of the data, and thus could not be done until after patients had completed the intervention, and at a time when many were known to be dead. While we took all possible care not to be biased by this knowledge, we are obviously open to the allegation that our psychological ratings were biased in this way. In future experiments of this kind, with the relevant attributes already defined, ratings can be done before survival outcomes are known.
It could also be argued-less justifiably, we believe-that "involvement" indirectly reflected health status, that is, that the H group were initially much healthier than the L, and thus more able to "get involved." However, when quality of life (FLIC), attendance, age, and medical estimates of survival were controlled for in the analysis, the subjects' degree of psychological work remained significantly related to their survival. Beyond this, the principal counter to the possibility that covert health status influenced involvement in self-help is that median estimates of survival for the 3 groups did not differ significantly when made by the expert medical panel. The estimates are unlikely to be biased in favor of the hypothesis since they were based on medical data collected before the therapy began. Furthermore, as seen in Figure 3, the interquartile ranges of the survival predictions of the same medical panel overlapped with the interquartile range of the observed survival in the separate study of deceased nontherapy patients. In the present study the median survival of the "High" involvement subjects are far apart from and do not overlap with the panel's median prediction of survival. The "Low" involvement group has a survival median that is very close to the predicted survival median, and there is great overlap in the ranges. The "Medium" group appears to have some separation, although the ranges do overlap slightly.
The attendance figures (Table 1) also speak against gross differences in health status of the patients at study entry: All were relatively well, fully capable of practicing their psychological strategies, and, with 1 exception, remained so for at least 6 months (the exception, a member of the L group, died at 4.5 months).
Involvement ratings were, in any case, based on the nature of statements, not their quantity. Thus, the L patients did not lack time to achieve a high involvement rating. Instead, they tended from the start to discount the idea that their efforts could make a difference, whereas those in the H group all accepted it unquestioningly and worked in a dedicated fashion throughout the year. The M group was intermediate in this way.
It is important to note also that, while the group of patients in this study were self-selected for qualities of motivation and good physical functioning at enrollment, the conclusions do not depend on a comparisons with the population at large but on comparisons amongst those within the study. The generalizability of conclusions is, however, limited to subjects with similar characteristics.
The results we have obtained support the contention of Pincus (Pincus 1997; Pincus & Callahan 1995) that valuable data may be obtained by studying a consecutive series of patients seen in the course of regular clinical practice, provided necessary documentation is kept. They are also reminiscent of the work of Ornish et al. (1990) with patients with coronary occlusions; in his experiments, there was a similar dose-response relationship between the slowing down or reversal of further occlusions and degree of involvement in a self-help program.
This has been an exploratory study, aimed at defining relevant psychological categories and making an initial estimate of their relationship to survival. In future work of this kind it should be possible to improve the design in various ways. The most important modification will be to do the rating as early as possible, to minimize any confounding between medical status and psychological self-help work. It will probably remain necessary, however, to observe the participants for about 2 months in therapy before many of the salient characteristics can be rated. A second improvement might be to use raters who are blind to the patients' medical variables and not exposed to the inferences that the therapists make from their clinical contacts. This would remove potential bias from knowing health status, and it would objectify the analysis. It might also prove to weaken the association; we used the therapists on the rating team so as to take advantage of their intimate knowledge of the subjects. A larger N would also enhance confidence in the effect.
One advantage of the present correlative approach is that useful research can be done with small numbers of patients. A homogenous population of patients may be useful to minimize the contribution of medical variables, although using the survival predictions of an expert medical panel as a covariate allows one to compare members of a group of patients with different prognoses. The method could prove useful in conditions other than cancer.
Our experiment is not concerned to show whether or not the therapy is critical; it could be argued that the psychological nature of the subjects at entry is decisive, without the therapy (although such a view would attribute much more prognostic importance to patients' psychology than usual). Our hypothesis, from watching hundreds of patients use these strategies over many years, is that enhanced survival depends both on psychological attributes at entry (especially openness to change and motivation) and on application of the techniques taught in the therapy. A qualitative analysis of this phenomenon (in preparation) will document the profound changes that can take place when a patient becomes involved. Confirmation of the need for therapy may require a randomized trial with motivated patients.
Achterberg J, Mathews-Simonton S, Simonton OC 1977 Psychology of the exceptional cancer patient: a description of patients who outlive predicted life expectancies. Psychother Theory Res Pract 14:416-422
Berland W 1995 Unexpected cancer recovery: Why patients believe they survive. Advances Journal of Mind-Body Health 11:5-19
Cunningham AJ, Edmonds CVI, Jenkins GP et al. 1998 A randomized controlled trial of the effects of group psychological therapy on survival in women and metastatic breast cancer. Psychooncology 7:508-517
Cunningham AJ, Edmonds CVI, Hampson AW et al. 1991 Helping cancer patients cope with, and combat, their disease: Report on a group psychoeducational program. Advances Journal of Mind
Cunningham AJ, Edmonds CVI, Phillips C, Soots KI, Hedley D, Lockwood GA 2000 A prospective, longitudinal study of the relationship of psychological work to duration of survival in patients with metastatic cancer. PsychoOncology, in press
Doan BD, Gray RE 1992 The heroic cancer patient: Reflections on the relationship between illusion and mental health. Canadian Journal of Behavioral Science, 24(2):253-266
Edelman S, Lemon J, Bell DR, Kidman AD 1999 Effects of group CBT on the survival time of patients with metastatic breast cancer. Psycho-Oncology 8:474-481
Fawzy Fl, Fawzy NW, Hyun CS et al. 1993 Malignant melanoma. Effects of an early structured psychiatric intervention, coping and affective state on recurrence and survival 6 years later. Archives of General Psychiatry 50:681-689
Huebscher RR 1992 Spontaneous remission of cancer: An example of health promotion. Nurse Practitioner Forum 3:228-235
Ilnyckyj A, Farber J, Cheang MC, Weinerman BH 1994 A randomized controlled trial of psychotherapeutic intervention in cancer patients. Annals of the Royal College of Physicians & Surgeons of Canada 27:93-96
Kennedy BJ, Tellegen A, Kennedy S, Havernick N 1976 Psychological response of patients cured of advanced cancer. Cancer 38:2184-2191
Linn MW, Linn BS, Harris R 1982 Effects of counseling for late stage cancer patients. Cancer 49:1048-1055
Meares A 1980 What can the cancer patient expect from intensive meditation? Australian Family Physician 9:322-325
Morganstem H, Gellart GA, Walter SD, Ostfeld AM, Siegel BS 1984 The impact of a psychological support program on survival with breast cancer: The importance of selection bias in program evaluation. Journal of Chronic Disease 37:273-282
Newton BW 1982-83 The use of hypnosis in the treatment of cancer patients. American Journal of Clinical Hypnosis 25:104-113
Ornish D, Brown SE, Scherwitz LW et al. 1990 Can lifestyle changes reverse coronary heart disease? The Lifestyle Heart Trial. Lancet 336(8708):129-133
Pincus T 1997 Analyzing long term outcomes of clinical care without randomized controlled clinical trials: The Consecutive patient questionnaire Database. Advances 13(2):3-32
Pincus T, Callahan LF 1995 What explains the association between socioeconomic status and health: Primary medical access or mind-body variables? Advances 11(1):4-36
Pennington S 1988 Healing Yourself. Toronto: McGraw
Ratcliffe MA, Dawson AA, Walker LG 1995 Eysenck Personality Inventory L-scores in patients with Hodgkin's Disease and non-Hodgkin's lymphoma. Psycho-Oncology 4:39-45
Richardson JL, Shelton DR, Krailo M, Levine AM 1990 The effects of compliance with treatment on survival among patients with hematologic malignancies. Journal of Clinical Oncology 8(2):356-364
Roud PC 1986-87 Psychosocial variables associated with the exceptional survival of patients with advanced malignant disease. International Journal of Psychiatric Medicine 16:113-122
Schatzman L 1991 Dimensional analysis: Notes on an alternative approach to the grounding of theory in qualitative research. In: Maines DR, ed. Social Organization and Social Process. New York: Aldine De Gruyter, pp 303-314.
Schipper H, Clinch J, McMurray A, Levitt M 1984 Measuring the quality of life of cancer patients: The Functional Living Index-Cancer: Development and validation. Journal of Clinical Oncology 2:472-483
Schipper H, Goh CR, Wang TL 1995 Shifting the cancer paradigm: Must we kill to cure. Journal of Clinical Oncology, 13:801-807
Siegel BS 1986 Love, Medicine and Miracles. New York: Harper & Row
Simonton OC, Mathews-Simonton S, Creighton JL 1980 Getting Well Again. New York: Bantam Press