Lee SJ, Fairclough D, Antin JH, Weeks JC. Discrepancies Between Patient and Physician Estimates for the Success of Stem Cell Transplantation. JAMA. 2001;285(8):1034–1038. doi:10.1001/jama.285.8.1034
Author Affiliations: Department of Adult Oncology, Dana-Farber Cancer Institute, Boston, Mass (Drs Lee, Antin, and Weeks); and AMC Cancer Research Center, Denver, Colo (Dr Fairclough).
Context Stem cell transplantation is associated with considerable morbidity
and mortality. The extent to which patients and their physicians correctly
estimate these risks is unknown.
Objective To measure the expectations of patients and physicians prior to stem
cell transplantation and correlate them with actual outcomes after transplantation.
Design Prospective cohort study with baseline questionnaire administered July
1996 through November 1999 and follow-up to May 2000.
Setting Tertiary care transplant center in the United States.
Participants Of 458 surveys mailed, evaluable returned surveys were included for
313 autologous and allogeneic stem cell transplantation patients and their
Main Outcome Measures Patient and physician expectations prior to transplantation (measured
on 6-point Likert scales) of treatment-related mortality, cure with transplantation,
and cure without transplantation; actual treatment-related mortality and disease-free
survival among patients with at least 1 year of follow-up after transplantation
(n = 263).
Results Both patients and physicians were fairly accurate in estimating treatment-related
mortality when actual mortality was less than 30%. However, in situations
in which mortality was higher than 30%, such as with allogeneic transplantation
for intermediate or advanced disease, physician expectations were lower, while
patients remained optimistic. Similarly, physicians provided lower estimates
of disease-free survival in cases of intermediate or advanced disease while
patient expectations remained high and constant regardless of disease stage.
Conclusions Patients and their physicians have the most concordant and accurate
expectations when the outcome of stem cell transplantation is likely to be
favorable. However, patients with more advanced disease fail to recognize
the higher risks associated with their situations.
Models for optimal shared (or informed) medical decision making encourage
open discourse between patients and physicians regarding the pros and cons
of treatment options.1- 4
However, whether patients actually emerge from these discussions with accurate
expectations for the chosen therapy has rarely been studied. Hematopoietic
stem cell transplantation (SCT) for hematologic diseases offers an ideal opportunity
to study this question in a high-stakes setting since SCT offers the potential
of life extension or cure, but at substantial risk of treatment-related death.
We measured the expectations of 313 patients and their physicians prior to
transplantation, then conducted follow-up longitudinally to compare expectations
with actual outcomes.
Patients were enrolled at the Brigham and Women's Hospital and the Dana-Farber
Cancer Institute. Eligibility criteria included age older than 18 years, ability
to read English, and planned autologous or allogeneic transplantation within
3 months of study enrollment but not less than 1 week. Patients were contacted
by telephone by a study investigator (S.J.L.) to introduce the study, then
mailed questionnaires, consent forms, and self-addressed, stamped envelopes.
Some patients could not be contacted due to logistic reasons. Others elected
not to participate in the study or did not return baseline questionnaires
and thus were not enrolled. Baseline surveys were administered July 1996 through
November 1999. Results reported here include follow-up through May 1, 2000,
although data collection continues. The data used for this analysis were collected
as part of a much larger, prospective, longitudinal questionnaire study evaluating
expectations and quality of life.
Disease status was categorized at the time of study enrollment into
3 groups. Early stage patients were those with leukemia in first complete
remission, stable phase chronic myelogenous leukemia, or aplastic anemia.
Intermediate patients had leukemia in second or greater complete remission,
accelerated phase chronic myelogenous leukemia, chemotherapy-responsive lymphoma
or myeloma, or had myelodysplastic syndrome, which was either untreated or
in remission. Advanced patients had relapsed or refractory diseases.
The specific conditioning regimen and graft manipulation, if any, was
according to protocol requirements and attending physician or patient preference.
Patients were to receive autologous or allogeneic grafts. Graft-vs-host disease
prophylaxis consisted of T-cell depletion or methotrexate and cyclosporine.
Details of the transplant regimens and graft-vs-host disease prophylaxis have
been previously published.5- 10
Each patient's physician determined the appropriateness and type of
transplantation and was responsible for communicating the risks and benefits
of the procedure. However, educational practices were not controlled, and
individual physicians determined the depth and format of medical information
discussed. All patients signed informed consent documents for transplantation
that provided only general information as required by the institutional review
board, not quantitative estimates of morbidity or mortality.
Patients were surveyed by mail before transplantation to record expectations
for the outcome and collect additional sociodemographic characteristics. Predictions
about cure with and without transplantation were captured by specific reference
to personal outcome (What do you think is your chance of being cured with
a stem cell transplant? and What do you think is your chance of cure without
a stem cell transplant?), whereas the question about treatment-related mortality
invoked population estimates (What percentage of people do you think die during
the transplant hospitalization or from complications within the first year?).
Responses were measured on 6-point Likert scales offering paired verbal and
numeric information (eg, almost all [>90%], most [75%-90%], the majority [50%-74%],
less than half [25%-49%], few [10%-24%], or almost none [<10%]).
The physician of each enrolled transplant patient was asked to estimate
the patient's chance of cure with and without transplantation and treatment-related
mortality within the first year. Response categories were identical to patient
options. Physician information was collected as soon as possible after patients
returned their surveys and prior to transplantation.
The SAS system was used for all statistical calculations.11
Syngeneic patients were combined with the autologous group for analysis. Responders
were compared with nonresponders using χ2, Mantel-Haenszel χ2, and Wilcoxon rank sum tests with P<.05
declared as significant.
Correlation between patient and physician expectations was evaluated
using Bowker 2-tailed tests of symmetry and weighted κ statistics. Mean
expectations for treatment-related mortality and cure were calculated for
each subgroup using the midpoint (eg, 37%) when response options were ranges
of probabilities (eg, 25%-49%). Discrepancies between estimates for chance
of cure with transplantation were quantified by calculating the minimal difference
between response options selected. For example, if a patient thought he/she
had a 75% to 90% chance of cure while his/her physician reported 25% to 49%,
we assigned a conservative discrepancy of 26% (75% minus 49%).
Analysis of treatment-related mortality and disease-free survival was
limited to the 263 patients who received transplants before May 1999 and for
whom 1-year follow-up was possible. Disease-free survival at 2 years was calculated
using the Kaplan-Meier method with surviving patients censored if in remission
at last follow-up.
A total of 458 surveys were sent to patients who underwent transplantation,
and 320 were returned of which 313 (68%) were evaluable. Characteristics of
the population are shown in Table 1.
Responders were more likely to be white than nonresponders (95% vs 88%; P = .004) but were otherwise similar in sex, disease stage,
and type of transplant procedure performed (including donor type, graft-vs-host
disease prophylaxis, and degree of match, if allogeneic).
Patient and physician estimates of chance of cure with transplantation,
chance of cure without transplantation, and 1-year treatment-related mortality
were compared with each other and with actual outcomes. There were 15 physicians
who saw at least 2 patients. Four physicians accounted for 64% of the patients
(11%-25% each). We compared these physicians with each other and with the
remaining 11 physicians; no evidence for a physician effect was seen.
Figure 1 shows concordance
of paired patient-physician expectations. Agreement occurred only 17% to 52%
of the time, and overall correlation was poor (weighted κ, 0.08-0.22).
Patients were significantly more optimistic than their physicians on all measures
(P≤.001). Anticipated chance of cure with transplantation
showed the greatest discordance with 78% of patients estimating their chance
of cure with transplantation higher than their physicians. In fact, 48% of
the population indicated an anticipated cure rate at least 2 prognostic categories
higher than their physician, corresponding to an absolute discrepancy of 16%
to 82% (median, 26%) using conservative assumptions.
Table 2 shows patient and
physician mean estimates compared with actual treatment-related mortality
according to type of transplantation and disease stage. Discrepancies between
estimated and actual mortality rates for each group can be seen by comparing
percentages across rows. When treatment-related mortality was lowest, as for
autologous transplants, related-donor T-cell–depleted procedures, or
related and unrelated donor, methotrexate and cyclosporine transplants performed
for early stage diseases, predictions were most concordant with actual outcomes.
When treatment-related mortality was higher (>30%), as for patients with intermediate
and advanced disease or those undergoing unrelated donor, T-cell–depleted,
allogeneic procedures, patient (and to a lesser extent, physician) predictions
greatly underestimated actual mortality.
Table 3 shows the expectations
for disease-free survival with or without SCT, stratified by type of transplantation
and disease stage. Most autologous patients had intermediate stage disease
(90 intermediate, 4 advanced). Results for allogeneic patients are presented
by disease stage only because expectations and disease-free survival were
not affected by donor status or type of graft-vs-host disease prophylaxis.
For each group, patient expectations were high and overly optimistic. Physician
expectations were closer to actual outcomes and differed by stage of disease
(P < .001) with lower estimates assigned to intermediate
and advanced patients.
We found discrepancies between patients and their physicians regarding
anticipated and actual outcomes of SCT. Estimates were most concordant and
correct when observed outcomes were favorable. However, as the likelihood
of poor outcomes increased, physicians adjusted their expectations while patients
Discrepancies between patient and physician expectations have been observed
in several other settings and attributed to both patient and physician factors.12- 15 Our
findings show that while most transplant patients are aware they have some
chance of treatment-related death and relapse, those at highest risk do not
appreciate the true likelihood of these possibilities. It is easy to imagine
that as a prognosis worsens, both patients and physicians may avoid discussions
involving discouraging statistics, instead focusing on the curative potential
of the procedure. Clinicians may be concerned that full awareness of grim
prognostic information will destroy hope and increase distress, possibly decreasing
a patient's will to live.12,16
To prevent this, they may selectively withhold or minimize some details, use
qualitative terms (such as frequent or possible) instead of numeric figures,14,16,17 emphasize optimistic
information, or suggest to patients that grim prognostic estimates apply to
populations and may not be meaningful to them as individuals.16,18,19
Patients considering transplantation may have powerful psychological incentives
to view their prognoses with an optimistic bias, and it is conceivable that
any attempts by physicians to force patients to acknowledge unfavorable statistics
may be met with anger, denial, or mistrust, in addition to anxiety and despair.
However, there may also be costs of allowing discrepancies between expectations
and reality to persist. At one level, complete respect for the principle of
patient autonomy requires that each patient be given the means to make medical
decisions for themselves unless they explicitly assign this role to someone
else. In some cases, an unabridged understanding of the risks might change
treatment decisions (alternatives include chemotherapy only or best supportive
care). Even if treatment choice is not altered, a more realistic understanding
of prognosis could encourage attention to financial and family arrangements,
affect personal preparations prior to transplantation, or prepare families
for the most likely outcome, which is death. Indeed, there is evidence to
suggest that concordance of expectations with actual outcomes promotes better
There are a number of limitations to this study and its findings. Most
importantly, this study was conducted at a single institution with a high
percentage of well-educated and white patients, and other centers may have
different populations and procedures for counseling patients. Nonetheless,
we suspect that our observations might be relevant to other transplant programs
and even other situations in medicine where patients accept substantial risks
in the hope of a cure. Second, we began our observation period after the decision
to undergo transplantation had already been made. We did not study the process
leading up to those expectations. In particular, physician-patient interactions
were not observed, so we do not know how prognostic information was communicated.
Finally, we did not measure other parameters that would allow us to assess
whether overly optimistic expectations meaningfully affect treatment choices,
are necessary to preserve hope, interfere with realistic preparations prior
to transplantation, or adversely affect adjustment following transplantation.
These hypotheses will be addressed in future studies.
Our findings should caution clinicians that many patients proceeding
to SCT harbor quite optimistic expectations regarding anticipated survival,
cure, and recovery that may not be supported by actual outcomes. How should
a clinician respond to these data? One practical suggestion is that physicians
try to ascertain their patients' expectations early in the decision-making
process and identify reasons for any misperceptions. To the extent that patients
simply lack information, this would be an opportunity for education. If more
active avoidance or denial is present, greater care may be required to ensure
that these defensive mechanisms are not compromising the patient's ability
to make appropriate treatment choices and to be prepared for transplantation.