The effect of disclosing life expectancy information on patients’ prognostic understanding: secondary outcomes from a multicenter randomized trial of a palliative chemotherapy educational intervention

Abstract

Context

Many advanced cancer patients have unrealistic prognostic expectations.

Objectives

We tested whether offering life expectancy (LE) statistics within palliative chemotherapy (PC) education promotes realistic expectations.

Methods

In this multicenter trial, patients with advanced colorectal and pancreatic cancers initiating first or second line PC were randomized to usual care versus a PC educational tool with optional LE information. Surveys at 2 weeks and 3 months assessed patients’ review of the LE module and their reactions; at 3 months, patients estimated their LE and reported occurrence of prognosis and end-of-life (EOL) discussions. Wilcoxon tests and proportional odds models evaluated between-arm differences in LE self-estimates, and how realistic those estimates were (based on cancer type and line of treatment).

Results

From 2015–2017, 92 patients were randomized to the intervention and 94 to usual care. At baseline most patients (80.9%) wanted “a lot” or “as much information as possible” about the impact of chemotherapy on LE. Among patients randomized to the intervention, 52.0% reviewed the LE module by 2 weeks and 66.7% by 3 months — of whom 88.2% reported the information was important, 31.4% reported it was upsetting, and 3.9% regretted reviewing it. Overall, patients’ LE self-estimates were very optimistic; 71.4% of colorectal cancer patients estimated >5yrs; 50% pancreatic patients estimated >2yrs. The intervention had no effect on the length or realism of patients’ LE self-estimates, or on the occurrence of prognostic or EOL discussions.

Conclusions

Offering LE information within a PC educational intervention had no effect on patients’ prognostic expectations.

INTRODUCTION

Prognostic misconceptions are pervasive among patients with advanced, incurable cancers.^1,2 In a study of 299 patients with refractory solid malignancies, half overestimated their life expectancy (LE) in excess of two years and, and nearly a third by five years or more.³ Unrealistic expectations impede patients’ ability to come to terms with terminal illness¹ and have been associated with receipt of intensive medical care near end-of-life (EOL).^4–6

The high prevalence of prognostic misconceptions belies the fact that a large majority of patients, spanning all phases of illness, want to be informed of their prognosis.^1,7–10 In the setting of upfront palliative chemotherapy (PC) decisions, approximately 80% of patients desire explicit information about its potential impact on LE.⁷ Beyond being fundamentally important for informed consent,^11,12 promoting realistic prognostic expectations from the outset of PC may help patients make important life decisions such as whether to continue working, how to prepare financially, and how to communicate with family and friends.

Analyses of audio-recorded oncologic consultations suggest that the survival benefits of PC are discussed much less frequently and in vaguer terms than reportedly desired by patients.^13–17 Disclosing a shortened LE is one of oncologists’ most difficult communication tasks,^18,19 particularly in the setting of a recent diagnosis when relationships are just developing. Patients often have high expectations of chemotherapy² and may feel disillusioned to learn that its survival benefits are modest. Furthermore many patients have conflicted feelings toward prognostic disclosure,^20,21 leaving oncologists uncertain of when and how to broach the subject.¹⁸

There is surprisingly little experimental evidence that disclosing LE information as part of PC consent (or in any other setting) supports realistic expectations or is acceptable to patients. Observational research has consistently linked prognostic disclosures to improved understanding;^1,22,23 however, these intuitive associations may be biased by patients’ underlying illness acceptance and receptivity toward these discussions. Although most patients report wanting maximal prognosis information, oncologists justifiably worry that discussing LE at the outset of PC may be unnecessarily upsetting, ineffective, and ultimately regretted.¹⁹

Reasoning that gaps in prognostic communications may impede informed decision-making, a disease and regimen-specific PC educational intervention was developed to provide comprehensive information regarding treatment risks and benefits, including an optional module with explicit LE information. This report presents secondary outcomes from a multicenter randomized trial in which patients starting PC for advanced gastrointestinal cancers were randomized to receive, or not receive, this educational intervention. We describe the proportion of patients who chose to review LE information and their reactions, as well as the impact of the intervention on patients’ prognostic expectations, and on the occurrence of prognostic and EOL communications.

METHODS

Study Setting, Participants, and Enrollment

Here report secondary outcomes from a multicenter randomized study that enrolled patients at five major US cancer centers: Dana-Farber Cancer Institute (Boston, MA), Beth Israel Deaconess Medical Center (Boston, MA), Hellen Diller Comprehensive Cancer Center at University of California-San Francisco, University of North Carolina Lineberger Comprehensive Cancer Center (Chapel Hill), and Virginia Commonwealth University Massey Cancer Center (Richmond). Main outcomes of the trial have been previously reported.²⁴ Patients were eligible if they were considering or starting first or second-line PC for metastatic colorectal cancer, locally advanced or metastatic pancreatic cancer. These disease were chosen because they are the second and third leading causes of cancer death in the US,²⁵ and because they represent one advanced cancer with a relatively favorable prognosis, and one with an extremely poor prognosis. Exclusion criteria included any plan for curative-intent surgery, age <21, inability to speak English, delirium or dementia. A trained research assistant approached eligible patients after receipt of an initial PC recommendation and no later than two weeks after chemotherapy initiation. Patients on first-line PC who consented beyond this window were also permitted to enroll, but they were not randomized until faced with a second-line chemotherapy decision. The trial was registered at clinicaltrials.gov (NCT02282722). The study was approved by the IRBs at all participating institutions.

Randomization and interventions

After participants signed written informed consent, they were randomized 1:1, without blinding, to usual care or the intervention, stratified by line of therapy. Participants were given standard chemotherapy educational materials at the discretion of their treating oncologist. Participants randomized to the intervention were given the study booklet tailored to their cancer type and regimen by the research assistant, with a weblink/password to the corresponding video, which could be reviewed in clinic or from home. Review of the intervention materials was neither incentivized nor mandated.

Intervention overview

We previously developed a suite of PC educational tools for common regimens used to treat advanced colorectal cancer (FOLFOX and FOLFIRI, with or without bevacizumab) or advanced pancreatic cancer (FOLFIRINOX, gemcitabine, and gemcitabine plus nab-paclitaxel). Development occurred through a highly iterative stakeholder driven research process built upon multiple layers of patient and clinician review. Each tool was regimen and disease-specific, consisting of a color printed booklet (19–23 pages) and a companion video (22–29 minutes). Videos featured oncologists and oncology nurses reviewing core information about the regimen including infusion logistics, its purpose, potential benefits, side effects with self-management advice, and treatment alternatives. Each video incorporated interviews with 3–5 patients describing their authentic experiences on treatment, reflecting a range of experiences with respect to toxicity and treatment benefits. These tools were designed to as educational resources to be distributed to patients following a treatment recommendation, and were not meant to replace patient-clinician discussions, nor were they meant to be decision aids. The theoretical framework, development, and initial testing of the intervention has been previously reported.²⁶

At the strong recommendation of patient-stakeholders, LE information was presented using an “opt-in” approach. Within each video an oncologist explained the potential of chemotherapy to prolong life, explained how LE information might be helpful, oriented patients to the optional LE module (via an embedded hyperlink), and normalized the decision to review or skip the module. LE statistics were presented in two opposing pages of each booklet, sealed by a sticker, and preceded by an introduction to the content. Survival statistics (extrapolated from medians and variance from clinical trial and population data) were presented with words and graphics, using lay-friendly explanations. As an example of the types of statistics presented, colorectal cancer tools reviewed median survival (“2 to 2 ½ years”), and approximate survival rates at 1 year (75%), 2 years (50%), and 5 years (10%)for patients not amenable to curative resection.^27–30 Pancreatic cancer tools presented separate survival statistics for metastatic^31,32 and locally-advanced disease.^33–35 Interpatient variability and the imprecise nature of LE estimates were emphasized. Empathic statements were integrated, and patients were encouraged to discuss the information with their oncologist. A sample booklet can be found in appendix A; a sample video can be found at https://vimeo.com/387430409.

Assessments

Participants completed surveys at three timepoints: at baseline, two weeks (window 2–4 weeks following treatment initiation), and three months (window 8–12 weeks). Surveys could be self-administered or administered by the RA (in person or via phone).

Covariates and outcomes

Clinical characteristics were abstracted from the medical chart at baseline. Participants self-reported race/ethnicity, education, and difficulty paying bills.³⁶ Participants were also asked how much information they wanted during treatment decision-making about how chemotherapy might impact the length of their life. Response options ranged from “none” to “as much information as possible” on a 4-point Likert scale.

Review of and reactions to prognosis information

On both follow-up assessments participants randomized to the intervention were asked whether they had reviewed the optional prognosis chapters of the booklet or video. Participants responding “yes” rated their agreement with the following statements on a five-point Likert scale: “The information [reviewed in the booklet/video LE module] was important for me to know,” “my doctor had already discussed most of this information with me,” “I found this section [of the booklet/video] to be upsetting,” and “I wish I had not [read/watched] this section.” For simplicity we report participants’ first assessment of the prognosis information, defaulting to assessments of the printed materials for those who rated both.

Prognostic understanding

At 3 months participants were asked “Every person is different and their situation is unique. If you had to make a guess — based on what you have learned about your cancer, your cancer treatment, and what you know about yourself — how long do you think that you have to live?” with response options of less than one year, more than one but less than 2 years, more than 2 but less than 3 years, more than 3 but less than 5 years, more than 5 but less than 10 years, and more than 10 years. Because of the heterogeneity in prognoses of colorectal and pancreatic cancers, we assessed how realistic a participants’ LE self-estimate was based upon their underlying diagnosis and line of therapy [supplemental figure 1]. Extrapolating from survival curves reported in key colorectal^27–30 and pancreatic cancer^31–35 PC clinical trials and SEER data,²⁹ we developed algorithms that designated participant’s estimate as “extremely optimistic” if less than 10% of patients in their situation would be expected to survive that long (i.e. reflecting the extreme right side of a survival curve); “optimistic” if approximately 10–25% of patients in their situation would be expected to survive that long; and “realistic” if 25–75% would be expected to survive that long (i.e. spanning the median and interquartile range). All other estimates were characterized as “pessimistic;” however, this was collapsed with “realistic” for analysis because of the rarity of pessimistic estimates..

Prognosis and EOL communications, and EOL care preferences

At 3-months participants were asked “has your doctor or anyone on your healthcare team ever discussed how long you might expect to live with your cancer?” (yes/no). From the SUPPORT study, participants were asked if they preferred care focused on life-extension or comfort.⁶ They were asked “have you spoken with your doctor about your wishes for your care in the future should you become more seriously ill?,”³⁷ whether they had discussed these wishes with their healthcare proxy, and whether they had discussed financial preparations with their family (yes/no).

Analysis

We present the proportions of participants in the intervention arm who reviewed the optional LE modules and their reactions. Differences between study arms in dichotomous outcomes were analyzed using Fisher’s exact tests; bivariate logistic regression estimated odds ratios and 95% confidence intervals., Wilcoxon rank sum test evaluated between-arm differences for ordinal categorial variables (i.e. LE self-estimates and LE optimism as reported on the 3 month survey), an appropriate method for analyzing non-parametric ordinal data with small sample sizes. Because the Wilcoxon test does not provide a statistic reflecting effect size, proportional odds models were used to calculate odds ratios and 95% confidence intervals for ordinal categorical variables. Primary analyses are reported according to the intention to treat (ITT) principle. Pre-specified subgroup analyses explore intervention effects separately for participants with pancreatic cancer, and those with colorectal cancer and are hypothesis generating. Exploratory post-hoc analyses explored intervention effects among the control group versus those who reviewed the optional LE component of the intervention; exploratory post-hoc analysis also compared LE self-estimates of patients randomized to the intervention who chose to review the optional LE module and those who did not. P values are two-sided and considered significant if <0.05. All analyses were performed using R version 3.5.1, R Foundation for Statistical Computing (Vienna).

RESULTS

Study participants and follow-up

From 2015–2017, 216 patients consented and were enrolled on the study, 200 were randomized, 186 of whom completed at least one survey assessment and were included in the analytic cohort. Completion rates for the 2-week and 3-month assessments were 83% and 77.4%, respectively [Figure 1].

Figure 1.

CONSORT Diagram

Patient Characteristics

Participants on this study were largely White, and highly educated, with 57.6% having at least a college degree [Table 1]. Overall, 36.6% had pancreatic cancer and 63.4% had colorectal cancer; 79.7% were considering first-line and 19.3% second-line chemotherapy. ECOG performance status was slightly better among intervention than control participants (p = 0.05); however, other characteristics did not differ significantly. When asked baseline how much information they wanted about how chemotherapy might influence the length of their life, 64.5% of patients (118/186) wanted “as much information as possible,” 16.4% (30/186) wanted “a lot,” 23% (12.6) wanted “a moderate amount,” and 6.6% (12/186) wanted little or no information.

Table 1:

Participants characteristics

Baseline Characteristics	Overall trial population (N=186)			Subset responding to 3-month survey (N=144)
	Intervention (%) N=92	Usual Care (%) N=94	Total (%) N=186	Intervention (%) N=73	Usual Care (%) N=71	Total (%) N=144
Age; mean (SD)	59.4(12.1)	59.2(13.0)	59.3(12.6)	59.71(11.64)	58.10(13.25)	58.92(12.44)
Male	53(57.6)	54(57.4)	107(57.5)	39(53.4)	42(59.2)	81(56.2)
Race/ethnicity
White	84(91.3)	76(82.6)	160(87.0)	66(90.4)	58(82.9)	124(86.7)
Black	6(6.5)	11(12.0)	17(9.2)	5(6.8)	9(12.9)	14(9.8)
Asian/other	2(2.2)	5(5.4)	7(3.8)	2(2.7)	3(4.3)	5(3.5)
Advanced cancer diagnosis
Colorectal cancer	59 (64.1)	59 (62.7)	118 (63.4)	47(64.4)	45(63.4)	92(63.9)
Pancreatic cancer	33 (35.9)	35 (37.2)	68 (36.6)	26(35.6)	26(36.6)	52(36.1)
Line of therapy
1st line	71(78.9)	74(80.4)	145(79.7)	55(77.5)	57(81.4)	112(79.4)
2nd line	19(21.1)	18(19.5)	37(20.1)	16(22.5)	13(18.6)	29(20.6)
Chemotherapy regimen
FOLFOX +/− BEV	29(31.5)	31(33.0)		25(34.3)	24 (33.8)	49(34.0)
FOLFIRI +/− BEV	30(32.6)	28(29.8)	58(31.2)	22(30.1)	21(29.6)	43(29.9)
FOLFIRINOX	22(23.9)	22(23.4)	44(23.6)	17(23.3)	20(28.2)	37(25.7)
Gemcitabine	0(0)	2(2.1)	2(1.1)	0	0	0
Gemcitabine + nab-paclitaxel	11(12.0)	11(11.7)	22(11.8)	9(12.3)	6(8.5)	15(10.4)
Education
Grade school or less	5(5.5)	8(8.6)	13(7.1)	3(4.2)	5(7.1)	8(5.6)
High school, some vocational school or college	29(31.9)	36(38.7)	65(35.3)	25(34.6)	25(35.7)	50(35.2)
College degree or higher	57(62.6)	49(52.7)	106(57.6)	44(61.1)	40(57.1)	84(59.1)
Lives alone	12(13.0)	16(17.0)	28(15.1)	10(13.7)	10(14.5)	20(14.1)
Difficulty paying bills (very/extreme)	6(6.5)	9(9.6)	15(8.1)	6(8.4)	5(7.2)	11(7.8)
ECOG Performance Status *
0	44(57.9)	31(39.7)	75(48.7)	37(56.1)	29(45.3)	66(50.8)
1	26(34.2)	42(53.4)	68(44.1)	25(37.9)	33(51.6)	58(44.6)
2	6(7.9)	5(6.4)	11(7.1)	4(6.1)	2(3.1)	6(4.6)
Comorbidities
0	33(35.9)	37(39.3)	70(37.6)	26(35.6)	24(33.8)	50(34.7)
1	53(57.6)	44(46.8)	97(52.1)	43(58.9)	37(52.1)	80(55.6)
2	3(3.3)	10(10.6)	13(7.0)	2(2.7)	8(11.3)	10(6.9)
3+	3(3.3)	3(3.2)	6(3.2)	2(2.7)	2(2.8)	4(2.8)

Characteristics are presented for the overall study population (N=186), and for the subset of the population (N=144) that responded to the 3-month survey which assessed secondary outcomes presented in table 2.

^*p 0.05 for the overall trial population; otherwise differences between study arms were non-significant

Review and reactions to optional LE information

By the 2-week assessment, 52.0% (39/75) of patients randomized to the intervention had reviewed the optional LE information, with 48.0% (36/75) reviewing LE information within the booklet and 28.0% (21/75) within the video. By 3-months, 66.7% (54/81) of patients responding to at least one follow-up survey had reviewed the optional LE information, including 64.2% (52/81) reviewing LE information within the booklet and 30.9% (25/81) within the video. Among patients who reviewed LE information, 88.2% agreed or strongly agreed it was important to know; 31.4% agreed that it was upsetting, 73.1% agreed or strongly agreed that their doctor had already discussed most of the information, and only 2 patients regretted reviewing it (Figure 2).

Figure 2: Perceptions of life expectancy (LE) information among 54 participants who chose to review an optional prognosis module included as a component of the intervention.

At the 2-week and 3-month survey participants who reviewed the optional LE module of the intervention were asked to rate their agreement with 4 statements about the module. If a participant reviewed the LE module more than once, this figure presents Journatheir firstassessment,defaultingto rating of the LE booklet module if they also rated the LE video module.

Patients’ LE self-estimates, and impact of the intervention on patients’ expectations

Overall, patients were very optimistic regarding their LE. Among the 77 colorectal cancer patients willing to estimate their LE, 37.7% estimated >10 years and 62.3% estimated >5 years, with 77.9% categorized as “optimistic” or “extremely optimistic.” Among the 46 pancreatic cancer patients willing to estimate their LE, 20.4% estimated >5 years and 50% estimated >2 years, with 50.0% categorized as “optimistic” or “extremely optimistic.”

Patients’ LE self-estimates did not differ between the intervention and control arms in the overall study population (Table 2); however, among the pancreatic cancer subgroup, patients randomized to the intervention arm estimated their LE to be shorter than those in the control arm (OR, 5.29; 95% CI, 1.73–17.74), and their estimates appeared to be more realistic (OR, 5.99; 95% CI, 1.80–22.12). In contrast, among the colorectal cancer subgroup patients randomized to the intervention arm appeared to have shorter LE self-estimates as compared to the control arm (OR 0.44; 95% CI, 0.19–0.99), and their estimates were less realistic (OR, 0.33; 95% CI, 0.12–0.84).

Table 2.

Participants’ self-estimates of life expectancy, optimism regarding life expectancy self-estimates, and communications regarding prognosis and end-of-life (EOL) care preferences

Outcomes	Overall Study Cohort (N=144)				Colorectal Cancer Subgroup (N=92)				Pancreas Cancer Subgroup (N=52)
	Intervention N=73*	Control N=71*	OR (95% CI)	p-value**	Intervention N=47	Control N=45	OR (95% CI)	p-value	Intervention N=26	Control N=26	OR (95% CI)	p-value
Self-estimate of life expectancy
>10 years	17(28.8)	18(29.0)	0.98 (0.52–1.85)	0.96	16(42.1)	13(33.3)	0.44 (0.19–0.99)	0.05	1(4.7)	5(21.7)	5.29 (1.73–17.74)	<0.01
5–10 years	12(20.3)	10(16.1)			12(31.6)	7(17.9)			0(0)	3(13.0)
3–5 years	5(8.5)	8(12.9)			3(7.9)	4(10.3)			2(9.5)	4(17.4)
2–3 years	9(15.3)	8(12.9)			6(15.8)	4(10.3)			3(14.3)	4(17.4)
1–2 years	9(15.3)	13(21.0)			1(2.6)	9(23.1)			8(38.1)	4(17.4)
<1 year	7(11.9)	5(8.1)			0(0)	2(5.1)			7(33.3)	3(13.0)
Declined to estimate	14	9			9	6			5	3
Optimism regarding life expectancy self-estimate
Extremely optimistic	31(54.4)	31(50.8)	0.95 (0.47–1.90)	0.89	29 (76.3)	20 (51.2)	0.33 (0.12–0.84)	0.02	3(14.3)	12(52.2)	5.99 (1.80–22.12)	<0.01
Optimistic	7(12.2)	11(18.0)			4 (10.5)				3(14.3)	4(17.4)
Realistic/pessimistic	19(33.3)	19(31.1)			5 (13.2)	12 (30.8)			15(71.4)	7(30.4)
Patient-doctor discussions
Life expectancy	45(64.2)	36(51.4)	1.69 (0.82–3.54)	0.17	30(66.7)	21(47.7)	2.17 (0.86–5.65)	0.09	15(60.0)	15(57.7)	1.10 (0.31–3.88)	1.00
EOL preferences	16(22.5)	10(14.3)	1.69 (0.82–3.54)	0.28	8(17.4)	3(6.8)	0.285 (0.62–17.87)	0.20	8(32.0)	7(26.9)	1.10 (0.31–3.88)	0.76
Patient-caregiver discussions
EOL preferences	54(85.7)	48(82.8)	1.25 (0.42–3.79)	0.80	33(82.5)	27(75.0)	0.60		21(91.3)	21(95.4)	0.51 (0.0110.45)	1.00
Financial preparation	51(70.8)	49(70.0)	1.04 (0.47–2.28)	1.00	30(47.6)	28(25.0)	1.00		21(80.7)	21(80.7)	1.00 (0.20–5.06)	1.00
Preferences for life-prolonging versus comfort-oriented care
Extend life	43(67.1)	44(66.7)	1.02 (0.46–2.27)	1.00	28(70.0)	30(69.7)	1.00		15(62.5)	14(60.9)	1.07 (0.28–4.07)	1.00
Relieve symptoms	21(32.8)	22(33.3)			12(30.0)	13(30.3)			9(37.5)	9(39.1)

All outcomes were assessed on the third study assessment, conducted between 8–12 weeks of patients’ treatment decision

^*N for the intervention arm = 73, reflecting patients randomized to the intervention who completed the third survey; N for the control arm =71, reflecting patients randomized to the control condition who completed the third survey

^**P-value: Wilcoxon Rank Sum Test was used for the ordered categorical response outcomes. Fisher’s exact test was used for the binary response outcome. The odds ratio is based on a proportional odds model for the ordered categorical response outcomes.

To explore whether limited review of LE information contributed to the null results, exploratory post-hoc analyses compared the LE self-estimates of control arm participants who the LE self-estimates of intervention arm participants who reviewed the optional LE module (Supplemental Table 1). These analyses found no difference in LE self-estimates for the overall trial population. The pancreatic cancer subgroup analysis within this analytic cohort showed similar results to the primary analysis cohort; however, findings within the colorectal cancer subgroup became non-significant. As shown in supplemental Figure 2, LE self-estimates were shorter among intervention arm participants who reviewed the optional LE module as compared to intervention arm participants who did not review the optional LE module.

Impact of the intervention on prognostic and EOL communications and EOL care preferences

The frequency of patient-physician discussions about LE and EOL care preferences did not differ between study arms. Patient-caregiver financial and EOL communications, and EOL care preferences also did not differ between study arms.

DISCUSSION

In this multicenter randomized trial, providing the option to review survival statistics within a PC educational intervention had no overall effect on patients’ LE self-estimates. The LE information was considered important by the vast majority of patients who reviewed it, and although nearly a third reported that the information was upsetting, only 3.9% regretted their decision to review it.

Potential reasons for the null results of this study include the optional format of LE information within the intervention, which was reviewed by only two thirds of patients. Making survival statistics a core feature of the intervention might have been more impactful; however, exposing patients to unwanted LE statistics is contrary to principles of patient-centered care,³⁸ and it might have caused harm. Second, prognostic understanding is shaped by many factors beyond medical facts or “what the doctor says” — including religious faith,^1,39 coping styles,⁴⁰ a desire to be “positive,” or a persons’ beliefs about their physical and inner strengths.⁴¹ Offering LE statistics is unlikely to alter prognostic beliefs rooted in these more personal factors. Third, the study schema largely relied upon patients to review the intervention independently and did not integrate this into structured clinician-patient discussions or formal teaching sessions. Finally, delivering the intervention at the time of a new diagnosis or upfront PC decision-making might not be optimal, because patients may not be ready to process LE information. Future studies could explore delivering the intervention later in the disease course, or in concert with structured clinician interactions or facilitated patient-clinician conversations.

We found that pancreatic cancer patients randomized to the intervention had more realistic prognostic expectations, whereas colorectal cancer patients randomized to the intervention had less realistic expectations. These findings are hypothesis generating and require confirmation. Metastatic colorectal cancer has a relatively favorable prognosis, with up to 10–20% of patients living beyond 5 years.^28,30 It is possible that these statistics could have buoyed patients’ hopes in terms of their own LE and made them less realistic. In contrast, the dismal survival of advanced pancreatic cancer leaves little room for optimism. Moreover, stigma associated with a pancreatic cancer diagnosis might have prepared patients to receive and accept prognostic estimates. Future research is required to determine whether offering LE information has a differential effect on patients’ prognostic expectations based on cancer type, and whether any other factors might moderate patients’ interpretation or acceptance of this information.

Consistent with existing research,^1,7,9,10,42 the large majority of patients wanted maximal information about how chemotherapy might impact the length of their life — which is notable given that 80% of patients were starting first-line treatment and therefore recently diagnosed. Despite reportedly wanting this information, only half of patients reviewed the optional LE module within two weeks, and two thirds reviewed it within three months of their treatment decision. Our findings highlight how stated preferences may differ from preferences revealed through behavior. Ambivalent feelings toward LE information are common, and patients understandably desire control over the timing and manner of its disclosure.^7–9 An advantage of our intervention is that it allowed patients to determine whether, when, and with whom to review prognostic information. Providing static LE statistics is not a replacement for patient-clinician communications; however, this approach may increase patients’ access to reliable information, without being subject to the many real and perceived barriers to initiating LE discussions.

Upon reviewing the optional LE module, under a third of patients reported that it was upsetting, nearly all indicated it was important to know, and only two regretted their decision to review it. Moreover, the intervention did not increase participant distress or harm satisfaction with communications around treatment decision-making (reported previously).²⁴ These findings add experimental support to several observational studies suggesting that disclosure of negative prognostic information is not intrinsically harmful.^1,42,43 Previous research on this topic has largely relied upon self-report of prognostic disclosures,^1,22,42,44 a methodology that is both subject to recall bias and incapable of determining what information was conveyed. The current study addresses these validity threats through its prospective randomized design and use of standardized LE information.

When asked about the LE modules, a surprising majority of patients agreed with the statement “my doctor had already discussed most of this information with me.” With rare exception,⁴⁵ most audio-recording studies have found explicit prognostic discussions to be absent from most oncology consultations and subsequent encounters.^{14,17,46–48} Furthermore, vague prognostic statements (e.g. chemotherapy may “buy you more time”) are shared more commonly than numerical estimates.^17,48–50 It is likely that participants considered these more general prognostic statements when answering this survey question.

Like other studies,^1,2,6,23 patients in this study were remarkably optimistic regarding prognosis. Although some research suggests that unrealistic optimism may bolster patients’ psychological wellbeing,^40,51 it has also been shown to impede EOL planning and promote intensive, burdensome care near EOL.^1,5,6,52 Patients may also make very different life decisions if they believe they many years to live versus a short time. It is worth noting that almost 90% of patients in this study were willing to estimate the length of their life, far exceeding typical response rates for similar questions.^1,2,23 The item developed to solicit LE self-estimates in this study may be a useful tool for future research.

Limitations of this study include the fact that we did not assess patients’ LE understanding at baseline and could not assess a direct effect. Moreover, only two thirds of intervention patients reviewed LE information and it is unclear how attentive they were or if they discussed it with their care team. We also did not assess alternative sources of prognostic information, such as internet searches. Finally, participants were largely White, well-educated, and received care at major tertiary care centers. Findings may not be generalizable to other settings, cancer types, treatment modalities, or patient populations.

In summary, offering optional LE information within a PC chemotherapy intervention had no overall effect on patients’ self-estimates of LE. More research is needed to understand how patients interpret LE information, how they come to understand their own prognosis, and what intervention strategies might be most effective. Future studies could evaluate the effect of enhanced interventions that supplement educational interventions with communication coaching for patients or structured procedures to facilitate patient-clinician discussions about prognosis.

Key Message.

In this multicenter randomized trial involving 186 patients with advanced colorectal and pancreatic cancers, providing optional life expectancy statistics within a palliative chemotherapy educational intervention had no overall effect on patients’ life expectancy self-estimates.