Patient Preferences in Treatment Choices for Early-Stage Lung Cancer

Abstract

Background

Decision-making for lung cancer treatment can be complex because it involves both provider recommendations based on the patient’s clinical condition and patient preferences. This study describes the relative importance of several considerations in lung cancer treatment from the patient’s perspective.

Methods

A conjoint preference experiment began by asking respondents to imagine that they had just been diagnosed with lung cancer. Respondents then chose among procedures that differed regarding treatment modalities, the potential for treatment-related complications, the likelihood of recurrence, provider case volume, and distance needed to travel for treatment. Conjoint analysis derived relative weights for these attributes.

Results

A total of 225 responses were analyzed. Respondents were most willing to accept minimally invasive operations for treatment of their hypothetical lung cancer, followed by stereotactic body radiation therapy (SBRT); they were least willing to accept thoracotomy. Treatment type and risk of recurrence were the most important attributes from the conjoint experiment (each with a relative weight of 0.23), followed by provider volume (relative weight of 0.21), risk of major complications (relative weight of 0.18), and distance needed to travel for treatment (relative weight of 0.15). Procedural and treatment preferences did not vary with demographics, self-reported health status, or familiarity with the procedures.

Conclusions

Survey respondents preferred minimally invasive operations over SBRT or thoracotomy for treatment of early-stage non-small cell lung cancer. Treatment modality and risk of cancer recurrence were the most important factors associated with treatment preferences. Provider experience outweighed the potential need to travel for lung cancer treatment.

Patients choose among medical treatments based on their personal value systems and their providers’ recommendations. Therapeutic decisions for patients with lung cancer involve trade-offs between risks and benefits associated with treatment, often involving a combination of operative treatment, chemotherapy, and radiation therapy. For most patients with clinical stage I lung cancer, treatment involves operative treatment alone through a lobectomy, segmentectomy, or other sublobar resection [1]. Despite current recommendations for minimally invasive approaches to lung cancer operations [2], and despite studies demonstrating lower morbidity and mortality for minimally invasive lung resection, especially for high-risk patient populations [3–6], many surgeons have been slow to adopt minimally invasive approaches in favor of thoracotomy [7]. Moreover, up to 45% of patients who are offered operative treatment for either biopsy-proven or suspected early-stage disease may reject surgical treatment for their own reasons [8, 9].

In addition, alternative therapies may be appropriate for patients who refuse operative treatment or those who are considered to be at high risk for surgical resection. One such therapy, stereotactic body radiation therapy (SBRT), has emerged as an effective treatment for stage I lung cancer in patients considered to be at high risk or medically inoperable. However, precise definitions of “high risk” vary and should be considered on an individual basis. Several recent clinical trials of SBRT and sublobar resection for lung cancer have defined medically inoperable and high risk according to specific criteria. These include the American College of Surgeons Oncology Group (ACOSOG) Z4032 and Z4033 studies, as well as the Radiation Therapy Oncology Group (RTOG) 0236 and 0813 studies [10–13].

To date, few studies have been conducted to determine patient preferences regarding alternative treatment options for early-stage lung cancer. One method to do so is conjoint analysis, a statistical technique using stated preferences. This methodology has historically been used in market research to demonstrate people’s willingness to make trade-offs among the characteristics, or attributes, of a particular good or service. It is also used to estimate whether an attribute is important and to demonstrate the relative importance of different attributes [14]. Conjoint analysis has been successfully applied to health care in a variety of ways, determining preferences at the individual patient level, in randomized controlled trials, and in entire communities. In this study, participants made trade-offs across the 3 alternative treatments for lung cancer, which also varied across 4 treatment-related attributes: risk of complications, risk of recurrence, provider volume, and distance needed to travel for treatment.

Patients and Methods

A conjoint paired preference task provided the input to estimate the relative weights of 5 attributes, or considerations, in the treatment of early-stage lung cancer. The Duke University Medical Center Institutional Review Board deemed the protocol (Pro00050511) to be exempt from full board review; the survey and research protocols were approved without the need for informed consent from survey participants. The complete survey is presented in the Appendix.

Survey respondents were members of an online panel recruited by Amazon Mechanical Turk (MTURK), an open online platform through which “workers” can be hired to perform tasks such as surveys or beta testing [15]. The MTurk interface, hosted and governed by Amazon, brings together “requesters” who post tasks that can be done by computer and “workers” who can browse the listing of tasks and complete them for varying amounts of compensation [16, 17]. Respondents in this study were required to confirm that they were 40 years of age or older and had a self-reported history of either current or former tobacco use. Respondents with a smoking history were intentionally recruited for the study. This was not to produce a bias against nonsmokers but to provide a more realistic cohort of respondents who would potentially be offered SBRT because of medical or pulmonary comorbidity. After introductory questions regarding the respondents’ overall state of health as well as specific cardiac and pulmonary diagnoses, they reported their use of pulmonary-related medications and supplemental oxygen. Respondents were then asked to imagine that they had a “spot” on their lung, with a subsequent biopsy result confirming lung cancer.

The respondents learned about open operations, minimally invasive operations, and focused radiation therapy as alternative ways to treat their cancers. The descriptions elaborated on a number of clinical considerations for each treatment, including a description of the treatment itself, procedure duration, hospitalization, and recovery time at home. These characteristics are shown in Table 1. Respondents indicated what they liked and disliked about each and then were asked to indicate their relative preferences across the 3 treatments.

Table 1.

Defined Characteristics of the 3 Procedures

Procedure Characteristics	Open Operation	Minimally Invasive Operation	Focused Radiation
Length of procedure	3 h	3 h	1 h outpatient visit
Days in hospital	6 nights	4 nights	5 daytime visits within 1 wk (no nights)
Recovery time at home	6–8 wk	3 wk	1 wk

Once respondents had a sense of each of the treatments, a trade-off task asked for relative preferences between the options. For the conjoint exercise, these options presented with 4 other characteristics. These varied across tasks: risk of cancer returning (10%, 15%, and 20%), risk of complications (5%, 15%, and 20%), provider annual volume (3, 12, and 50 cases), and driving time (0.5, 1.5, and 3 hours). Attribute levels for risk of major complications and chance of recurrence were rounded estimates of values published in previous studies [6, 12, 18, 19]. Figure 1 gives examples of the preference tasks presented to the respondents. Across 12 tasks, the attributes paired with the procedures were varied to determine conjoint weights that would best predict each respondent’s valuations. After completing the survey, respondents were provided with a bar code to initiate payment of $1.00 through the Amazon MTurk site.

Fig 1.

Examples of trade-off tasks. The items describe tasks presented to the survey respondents.

Sawtooth Software, (Sawtooth Software, Inc, Orem, UT) using hierarchical Bayes modeling, generated the individual-level conjoint weights that most closely reproduced each respondent’s preferences. IBM SPSS Statistics (SPSS Inc, Chicago, IL) generated the analyses.

Results

A total of 246 individuals with a self-reported age greater than 40 years and a smoking history participated in the survey. Low-quality responses were excluded on the basis of 2 criteria: completion time and response reliability. Individuals who scored less than 5% for either criterion or less than 10% on both criteria were excluded from further analysis. Of the 246 responses collected, 225 were retained for analysis (91.4% of respondents). Tables 2 and 3 provide their baseline demographic and self-reported medical histories. The majority of survey respondents were between 51 and 60 years of age (165 of 225 [73.3%]). There were 119 women (52.9%) and 106 men (47.1%). The majority of respondents were white (189 of 225 [84.0%]). The median time to complete the survey was 12 minutes, 51 seconds (range 4–55 minutes) for 53 questions and 8 conjoint tasks.

Table 2.

Respondent Demographics

Characteristic	(N = 225)	%
Age, y
41–50	21	9.3
51–60	165	73.3
61–70	30	13.3
≥71	9	4.0
Sex
Female	119	52.9
Male	106	47.1
Race
White	189	84.0
Black or African-American	14	6.2
Hispanic	11	4.9
Asian	6	2.7
Other	5	2.2
Education level
Less than high school	2	0.9
High school diploma or GED	27	12.0
Some college	97	43.1
4-year college degree	68	30.2
Master’s degree	26	11.6
Doctoral degree	5	2.2
Employment status
Full-time employed	102	45.3
Part-time employed	36	16.0
Unemployed seeking work	27	12.0
Disabled	18	8.0
Retired	35	15.6
Other	7	3.1
Household size
Lives alone	54	24.0
Lives with 1 other individual	104	46.2
Lives with 2 other individuals	41	18.2
Lives with 3 other individuals	26	11.6
US geographic region
Northeast	40	17.8
Southeast	62	27.6
Midwest	48	21.3
Southwest	53	23.6
Northwest	20	8.9
Other	2	0.9
Community type
Large urban	43	19.1
Small urban	42	18.7
Suburban	95	42.2
Rural	44	19.6
Missing data	1	0.4
Annual household income
Less than $20,000	42	18.7
$20,000–$39,999	79	35.1
$40,000–$59,999	41	18.2
$60,000–$79,999	30	13.3
$80,000–$99,999	21	9.3
$100,000 or more	12	5.3

GED = general educational development.

Table 3.

Self-Reported Health History and Smoking Status

Characteristic	(N = 225)	%
Smoking status
Current smoker	106	47.1
Former smoker	119	52.9
Smoking exposure (both current and former smokers)
Less than one-half pack/d	67	29.8
One-half–1 pack/d	97	43.1
1–2 packs/d	49	21.8
2 or more packs/d	12	5.3
Years of smoking
Less than 10 y	41	18.2
10–20 y	65	28.9
20–30 y	58	25.8
More than 30 y	61	27.1
Missing data	0	0.0
Smoking cessation information
Less than 1 y previously	4	3.4
1–15 y previously	65	54.6
15–30 y previously	38	31.9
More than 30 y previously	12	10.1
Previous medical conditions
Asthma	58	25.8
Emphysema/COPD	18	8.0
Pulmonary fibrosis/ILD	4	1.8
Sarcoidosis	3	1.3
Pulmonary hypertension	9	4.0
None of the above	150	66.7
Medication use
Inhaled steroids	27	12.0
Other inhalers	27	12.0
Other asthma medications	14	6.2
Oral steroids	5	2.2
Combination inhalers	10	4.4
Pulmonary hypertension medication	8	3.6
Supplemental oxygen	6	2.7
Previous hospitalization for lung condition
Never	179	79.6
Once	31	13.8
2–4 times	12	5.3
5 or more times	2	0.9
Missing data/not sure	1	0.4
Current overall health
Excellent	5	2.2
Very good	55	24.4
Good	103	45.8
Neither good nor poor	39	17.3
Poor	23	10.2
Very poor	0	0.0
Current health problems	(None/Some/Major)	(%)
Mobility	77.8/22.2/0
Self-care	93.8/6.2/0
Normal activities	76.0/20.0/4.0
Pain	52.9/44.0/3.1
Anxiety or depression	60.9/35.1/4.0

COPD = chronic obstruction pulmonary disease; ILD = interstitial lung disease.

Of the survey respondents, 106 (47.1%) were current smokers, whereas 119 (52.9%) were former smokers. Smoking amounts varied, with 28.0% of the respondents reporting tobacco use of more than 1 pack of cigarettes per day. Of the former smokers, 65 of 119 (54.6%) reported quitting within 15 years of taking the survey; 38 of 119 (31.9%) quit between 15 and 30 years previously, and 12 of 119 (10.1%) quit smoking more than 30 years before taking the survey.

Compared with typical patients with lung cancer, survey respondents were relatively healthy. The most commonly reported pulmonary diagnosis was asthma (58 of 225 [25.8%]). Similarly, a minority of respondents reported the use of inhalers, steroids (oral or inhaled), or supplemental oxygen (61 of 225 [27.1%]) to treat any pulmonary condition. Only 29 of 225 patients (12.9%) reported any history of undergoing invasive pulmonary or cardiac procedures, ranging from bronchoscopy to lung resection to cardiac bypass operations or valve repair/ replacement. Although few respondents reported their health as excellent (5 of 225 [2.2%]), many saw it as very good (55 of 225 [24.4%]) or good (103 of 225 [45.8%]).

The conjoint task generated estimates of the weights for the 5 attributes. When compared with each other, the individual attributes did not have equal weights. Both treatment type and risk of cancer recurrence were equally most important (Fig 2). Provider volume was next most important, followed by risk of treatment complications. The distance necessary to travel for treatment was least important of the 5 attributes tested.

Fig 2.

Relative weights of attributes determining respondent conjoint exercise preferences.

After the conjoint exercise, respondents were asked about their reaction to elaborated versions of the 3 procedures, which included information about hospitalization and expected recovery. Respondents were most willing to accept minimally invasive operations for treatment of their hypothetical lung cancer (mean score, 4.07 of 5; SD, 1.12), followed by focused radiation or SBRT (mean, 3.40; SD, 1.34). Respondents were least willing to accept thoracotomy (mean, 2.88; SD, 1.30) based on the treatment description alone. Even when the described treatments were presented with estimated complications and risk of locoregional recurrence, respondents chose minimally invasive operations (72.0%) over SBRT (22.7%) and open operations (5.3%). Although these results show strong differences in aggregate willingness to accept each of the treatments, they also reveal important heterogeneity: 22.7% of respondents would accept SBRT over either of the surgical options, whereas 32.4% would accept either operation over SBRT. Only a few respondents diverged from the aggregate preferences, with 5.3% saying they would accept open operations over either of the other treatment options and 5.4% saying they would be least willing to accept video-assisted thoracoscopic surgery (VATS).

Comment

This was a pilot study exploring patient preferences for alternative treatments for early-stage lung cancer. Together, the attributes examined comprehensively describe patient considerations in lung cancer treatment. The study did not attempt to assess cost-effectiveness, because several studies have demonstrated that SBRT may be more cost-effective than surgical treatment in “marginally operable” patients [20, 21]. We purposely excluded treatment costs in this study to avoid any confounding effect. Moreover, a significant proportion of oncology patients do not consider out-of-pocket costs when making decisions about cancer treatment, nor do they wish for their providers to do so in making medical decisions [22].

With results comparable to other sampling methods, MTurk has been used successfully in the social sciences as well as in clinical research [23, 24]. Although the general age and demographics of “Turkers” is younger than that of most patients at risk for lung cancer, we did not find differences in preferences between those older and sicker and younger and healthier respondents. Further, the smoking status of our survey population is similar to that of current smokers in the United States, with 27.1% of our study population reporting smoking more than 1 pack of cigarettes per day [25].

Study participants were least willing to accept thoracotomy and lung resection for treatment of their lung cancers. This is not surprising given the known benefits of VATS versus thoracotomy regarding perioperative complications, pain, hospital length of stay, and patient recovery time [6, 26]. Considered together, the expressed patient preferences, the demonstrated oncologic efficacy of VATS, and the lower treatment-related morbidity relative to thoracotomy all support the notion that surgical candidates with early-stage lung cancer should undergo a minimally invasive surgical approach.

Study participants were most willing to accept minimally invasive operations, regardless of demographics, smoking history, or health status. At present, it is unclear what proportion of patients with borderline operability are discounted for operative treatment and referred directly for SBRT without actually consulting with a thoracic surgeon. Given the participants’ strong preference for surgical resection, even in light of potential procedure-related complications, consideration of SBRT should be done in a multidisciplinary setting with thoracic radiation oncologists and thoracic surgeons alike.

The estimates for length of procedure, hospital duration, and home recovery for the surgical treatments were based on the clinical experience of 3 coauthors (T.A.D., M.G.H., and B.C.T.); those associated with SBRT are based on the practices of our colleagues in Radiation Oncology (D.Y. and C.K., mentioned in acknowledgements). Supporting literature provided the basis for the estimates of treatment-related risk and cancer recurrence. These estimates were not exact replicates of the literature; to facilitate the respondents’ choices, the respondent must be able to appreciate a reasonable difference between attribute levels.

As in other studies of patient preferences, the attributes examined did not have equal weights [27]. Treatment type and risk of cancer recurrence were equally important and substantially more important than the other characteristics. The fact that provider volume was next most important suggests that patients might prefer to be treated by someone specializing in thoracic oncology and lung cancer, rather than someone with a variety of clinical interests.

Among the attributes studied, the fact that travel distance for treatment was least important suggests that patients are willing to travel to obtain what they perceive to be the best care for their lung cancer. Although it was the least important of the attributes studied, one should not consider it to be unimportant. Other studies in the literature suggest that distance traveled for treatment does affect patient decision-making regarding cancer care, and that some patients are not willing to travel for specialist care [28, 29]. In contrast, our results suggest that although still a factor in patient decision-making, travel distance is less important than treatment type and outcomes in this study population.

It is surprising that the “sicker” respondents did not state a preference for SBRT, a less-invasive treatment modality than lung resection by either VATS or thoracotomy. We had expected that respondents would realize from the descriptions of the 3 treatments that SBRT would be more appropriate for respondents with poorer performance status. However, there was no correlation between self-reported health status and their preference judgments.

Regardless of the number of medications or reported hospitalizations for either pulmonary or cardiac disease, a majority of survey respondents (72.9%) reported their health status to be “good,” “very good” or “excellent.” In clinical practice, SBRT is recommended for patients who cannot tolerate surgical resection [30]. We had anticipated that those seeing themselves as “sick” would be more likely to reject surgical treatment in favor of SBRT. Surprisingly, there was no correlation between either self-perceived status or objective health status and the respondents’ selection of VATS as their treatment of choice. Authors of a recent study of SBRT for medically inoperable lung cancer reported that 90.5% of patients reported a Zubrod score of 0 or 1, suggesting that these patients may overestimate their health status relative to that observed by providers [31]. In contrast, Lilenbaum and colleagues [32] reported that 17.4% of patients with lung cancer had a self-reported Eastern Cooperative Oncology Group performance status of 3 or 4. In our study, neither age nor self-perceived health status were predictive of choosing SBRT over VATS, raising the possibility that these factors may not be driving decision-making. The concordance or discrepancy between patients’ self-perceived health status versus that of their providers also deserves further study.

In addition, objective classifications of medical inoperability are not standardized and vary even in clinical trials of “high-risk” patients such as the ACOSOG Z4033 and RTOG 0216 trials. Furthermore, a recent study of high-risk patients, as defined by ACOSOG criteria, demonstrated no difference in early postsurgical outcomes compared with those of their “normal-risk” counterparts [33]. In practice, it is unknown what proportion of potentially operable patients are deemed “medically inoperable” in the absence of multidisciplinary evaluation.

The ACOSOG Z4099/RTOG1021 trial (NCT01336894) was launched in 2011 to compare SBRT versus surgical treatment with or without brachytherapy in high-risk operable patients [34]. If the study had demonstrated significant survival differences in a specific group or groups of patients, provider recommendations for this patient population likely would have changed. However, because the trial was closed in 2013 because of slow accrual, relative outcomes remain unknown, and thus patient preferences should be part of the treatment decision-making process for patients who are considered to be high-risk surgical candidates.

This study has several important limitations. The proportion of women in the study is comparable to the relative proportion of women and men in the United States (http://www.census.gov/prod/cen2010/briefs/c2010br-03.pdf). However, the proportion of nonwhite respondents in our survey was slightly lower than that in the general population (http://quickfacts.census.gov/qfd/states/00000.html). Racial and other demographic differences may affect patients’ treatment preferences specifically for surgical resection versus SBRT, as has been demonstrated for non-small cell lung cancer operative treatment versus no treatment, as well as for other medical conditions [8, 9, 35, 36].

Perhaps the most significant limitations of this study stem from the fact that participants were recruited through the Amazon MTurk online panel. The MTurk “workers” participate in online tasks voluntarily, and this may have resulted in some selection bias in our study. The majority of survey respondents reported that they were between 51 and 60 years of age; we did not ask for participants’ specific ages in this study. In contrast, the median age at diagnosis for invasive lung cancer in the United States is approximately 70 years [37]. In addition, despite our efforts to screen for participants with several pulmonary and cardiac comorbidities to mimic the patient population most commonly referred for SBRT over surgical treatment, the majority of our study population was relatively healthy, with 75% taking no pulmonary medications and 86% without any history of invasive pulmonary or cardiac procedures. An objectively and subjectively older and less healthy survey population may be more inclined to choose SBRT over VATS. However, respondent age was not predictive of choosing SBRT over VATS in the study population. Moreover, our subgroup analysis in which respondents with poorer health did not prefer SBRT to surgical treatment did not suggest that this would be the case.

This pilot study was conducted as an initial exploration into decision-making for patients who might be considered high risk for lung cancer operations. Participants in this study were given a hypothetical scenario in which they were diagnosed with early-stage lung cancer; it is important to recognize that the choices made in the survey might not accurately reflect their choices in a real-life scenario. Additional studies to include respondents with demographics and comorbidities more concordant with the average patient with lung cancer or patients with true lung cancer diagnoses are warranted to better understand the decision-making process for patients with significant comorbidities seeking lung cancer treatment. Also, patients make decisions on a continuum, based on information they receive and incorporate during their care visits. After a face-to-face interaction and developing a rapport with the treating physicians, patient decision-making and choices may change.

In conclusion, patients must consider several factors when choosing treatment for lung cancer. These results suggest that not all factors are equal; treatment type and risk of recurrence were the most important attributes in this study. Provider volume was more important than the risk of complications associated with the various treatments. In addition to considering clinical evidence, guidelines, and even cost, providers should be mindful of patients’ perspectives and preferences when discussing and offering treatment options for early-stage lung cancer. Because distance needed to travel for treatment was the least important attribute, further study of regionalization of lung cancer care may be warranted. Finally, this study indicates that asking patients what they think may not match evidence-based recommendations. Physicians have a responsibility to counsel patients with regard to pros and cons of each treatment decision based on the best available evidence.