- Email: [email protected]
Clinical Evaluation of an Individualized Risk Prediction Tool for Men on Active Surveillance for Prostate Cancer
Accepted Manuscript
Clinical evaluation of an individualized risk prediction tool for men on active surveillance for prostate cancer Joseph H. Huntley BS , Rebecca Y. Coley PhD , H. Ballentine Carter MD , Archana Radhakrishnan MD, MHS , Melinda Krakow PhD, MPH, MA , Craig E. Pollack MD, MHS PII: DOI: Reference:
S0090-4295(18)30903-8 https://doi.org/10.1016/j.urology.2018.08.021 URL 21213
To appear in:
Urology
Received date: Revised date: Accepted date:
30 May 2018 15 August 2018 17 August 2018
Please cite this article as: Joseph H. Huntley BS , Rebecca Y. Coley PhD , H. Ballentine Carter MD , Archana Radhakrishnan MD, MHS , Melinda Krakow PhD, MPH, MA , Craig E. Pollack MD, MHS , Clinical evaluation of an individualized risk prediction tool for men on active surveillance for prostate cancer, Urology (2018), doi: https://doi.org/10.1016/j.urology.2018.08.021
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Title: Clinical evaluation of an individualized risk prediction tool for men on active surveillance for prostate cancer Authors and affiliations: Joseph H. Huntley, BS; Department of Medicine, Johns Hopkins School of Medicine, Baltimore, MD, USA; [email protected]
CR IP T
Rebecca Y. Coley, PhD; Kaiser Permanente Washington Health Research Institute, Seattle, WA, USA; [email protected] H. Ballentine Carter, MD; Department of Urology, The James Buchanan Brady Urological Institute, Johns Hopkins School of Medicine, Baltimore, MD, USA; [email protected]
AN US
Archana Radhakrishnan, MD, MHS; Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA; [email protected]
Melinda Krakow, PhD, MPH, MA; National Cancer Institute, National Institutes of Health, Bethesda, MD, USA; [email protected]
M
Craig E. Pollack, MD, MHS; Department of Medicine, Johns Hopkins School of Medicine, Baltimore, MD, USA; [email protected]
CE
PT
ED
Corresponding author: Craig E. Pollack, MD, MHS Associate Professor, Division of General Internal Medicine Johns Hopkins School of Medicine 2024 E. Monument Street, Suite 2-519 Baltimore, MD 21287 Phone: 410-955-4201 Email: [email protected] Conflict of interest statement: All authors declare no conflict of interest.
AC
Acknowledgments: We would like to acknowledge funding from the Maryland Cigarette Restitution Fund. Keywords: risk assessment, prostatic neoplasms, active surveillance
ACCEPTED MANUSCRIPT
Abstract Objective: To determine whether providing individualized predictions of health outcomes to men on active surveillance (AS) alleviates cancer-related anxiety and
CR IP T
improves risk understanding.
Materials and Methods: We consecutively recruited men from our large, institutional AS program before (n=36) and after (n=31) implementation of a risk prediction tool. Men
AN US
in both groups were surveyed before and after their regular visits to assess their
perceived cancer control, biopsy-specific anxiety, and burden from cancer-related information. We compared pre-/post-visit differences between men who were and were not shown the tool using two-sample t-tests. Satisfaction with and understanding of the
ED
M
predictions were elicited from men in the intervention period.
Results: Men reported a relatively high level of cancer control at baseline. Men who
PT
were not shown the tool saw a 6.3 point increase (scaled from 0 to 100) in their perceived cancer control from before to after their visit whereas men who were shown
CE
the tool saw a 12.8 point increase, indicating a statistically significant difference between groups (p=0.04). Biopsy-specific anxiety and burden from cancer information
AC
was not significantly different between groups. Men were satisfied with the tool and demonstrated moderate understanding.
Conclusion: Providing individualized predictions to men on AS helps them better understand their cancer risk and should be considered at other clinical sites.
ACCEPTED MANUSCRIPT
Introduction Over 180,000 men are diagnosed with prostate cancer every year in the United States.1 An large proportion of all new prostate cancer diagnoses are indolent tumors that do not require immediate curative intervention.2 Yet, the majority of men with favorable risk
CR IP T
cancer will choose surgical removal or radiation of the prostate,3 which poses
considerable risk of permanent side effects including incontinence and impotence, and can be physically, emotionally, and financially burdensome.4
AN US
Despite evidence that active surveillance (AS) is a safe alternative to curative
intervention for favorable risk prostate cancer, uncertainty regarding the underlying prostate cancer precludes many men and their physicians from electing AS as a rational
M
strategy for managing their cancer.5,6 Further, previous literature has reported that up to 20% of men drop out of AS without evidence of disease progression,7–13 with the
ED
uncertainty involved in being on AS and fear of cancer spreading as primary drivers of
PT
their decision-making.7,14
CE
Data-driven tools that help men in AS understand their underlying risk of developing more aggressive prostate cancer may alleviate anxiety and improve adherence to AS.
AC
Despite the potential for these approaches, such tools have not, to our knowledge, been evaluated in clinical care. We sought to investigate whether a tool that provides individualized predictions for risk of developing more aggressive cancer15,16 is associated with lower self-reported worry about prostate cancer among men in AS. We hypothesized that men who received individualized risk predictions would report feeling
ACCEPTED MANUSCRIPT
more control (i.e., less worried) over their cancer relative to men who did not receive predictions, and would not feel more burdened by cancer-related information.
Patients and Methods
CR IP T
The Johns Hopkins Institutional Review Boards determined this project to be a quality improvement effort that did not constitute human subjects research.
Individualized risk prediction tool
AN US
The tool for individualized risk prediction for men in the Johns Hopkins AS program has been previously reported and its predictive validity assessed.15,16 Briefly, the underlying model combines data on men’s repeated prostate-specific antigen measurements
M
(PSA) and prostate biopsies within a Bayesian hierarchical model to predict the underlying grade of an individual’s prostate cancer, given clinical observations of similar
ED
patients. The model requires at least two PSA observations and one biopsy after diagnosis to generate an individualized prediction. When men are new to AS, their
PT
predicted pathological Gleason scores are closer to those of the population; however,
CE
as men receive more PSA tests and biopsies, the model generates predictions that
AC
allow providers to better determine how men vary from the population average.
The tool presents the probability that the individual’s next biopsy will be upgraded, as well as 5- and 10-year cure rates, biochemical recurrence rates, and metastasis rates after definitive treatment (Figure 1 and Appendix).
ACCEPTED MANUSCRIPT
Sample selection Recruitment was performed in two phases. Men in the Johns Hopkins AS program17 return to Johns Hopkins every six to twelve months for evaluation. Men who returned for a follow-up visit between 9 February 2017–4 April 2017 were eligible to be included in
CR IP T
the control phase and were consecutively recruited. These men received standard of care and were counseled about the risks associated with AS but were not shown
specific predictions from the tool during their visits. Men who returned for a follow-up visit between 20 September 2017–8 February 2018 were eligible to be included in the
AN US
intervention phase. These men were shown individualized predictions from the tool during their visit, discussed the predictions with their provider, and were given a print copy. Men in the study met with one of two AS providers who worked closely together.
M
The time between the control and intervention phases was spent integrating the tool into routine clinical care. Men who were enrolled in AS for a short time and had few PSA
ED
measurements or biopsies did not have enough data for accurate risk prediction and were excluded from the study sample. All eligible men agreed to participate. There were
CE
PT
no men who were included in both the control and intervention phases of the study.
Data collection
AC
Men were given paper surveys to complete before and after meeting with their provider. They were encouraged to complete the second survey immediately after their visit, but were given the option to return it by mail. A follow-up call was made to men who elected the mail-in option to ensure that they completed the survey within three days after their
ACCEPTED MANUSCRIPT
visit. Sociodemographic and clinical variables were extracted from the existing AS database.
Outcome measures
CR IP T
Our main outcome was change in men’s perceived control over their prostate cancer, measured by the validated Cancer Control scale18 administered before and after men’s visits. This is a 5-point Likert scale which uses five items to determine how worried
individuals are about their cancer spreading or becoming more aggressive, and how
AN US
confident they are that their treatment option works and their cancer is under control. Responses were scaled to fall between 0–100 and adjusted so that higher scores
M
reflected improved control.
We also aimed to determine whether men felt overwhelmed by cancer-related
ED
information. This was measured by a modified version of the Cancer Information Overload scale,19,20 which was composed of five items and scored according to a 5-
PT
point Likert scale. Lastly, we measured anxiety specific to the prostate biopsy through a
CE
modified, two question version of the PROMIS Emotional Distress – Anxiety Short Form 7a.21 Men responded on a 5-point Likert scale about their anxiety due to the prostate
AC
biopsy and its physical consequences (see Appendix). These two scales were asked on pre- and post-visit surveys.
For men in the intervention phase, we also measured understanding of the tool’s predictions by asking four multiple choice questions about predictions that were
ACCEPTED MANUSCRIPT
presented explicitly by the tool, as well as four questions (5-point Likert) about perceived usefulness and understandability of the tool’s predictions (see Appendix). Responses to questions about the explicit predictions were scored for accuracy. These
CR IP T
questions were asked on the post-visit survey only.
Covariates
On the pre-visit surveys, we measured medical decision autonomy and numeracy as covariates. Medical decision autonomy was determined by presenting men with a line
AN US
numbered from 0 (complete patient autonomy) to 10 (complete physician autonomy) and asking them to place a mark on the line that best represented their preference regarding the decision to receive a prostate biopsy. Numeracy was measured by asking
M
men generally how difficult it was for them to understand medical statistics, followed by
ED
one basic multiple-choice question (3 options) about medical statistics.22
From the AS database and electronic medical records system, we extracted
PT
sociodemographic characteristics including race (White or non-White), age (<75 yrs. or
CE
≥75 yrs.), relationship status (married or not married), education level (Master’s or less than Master’s), and employment status (full-time or not full-time) for all study
AC
participants. We also determined whether each participant met all pre-specified eligibility criteria (Gleason <7, positive cores <3, positive core involvement <50%, PSA density <0.15 ng/mL/cm3) before joining AS, how many years they had been in AS, and their change in PSA since diagnosis.
ACCEPTED MANUSCRIPT
Finally, for men in each recruitment phase, we used the tool’s predictions to calculate the average probability of a grade group increase from a same-day biopsy, the average probability of men’s current grade group matching their predicted pathologic grade group, and the average chance of being cured (i.e., no biochemical recurrence or
CR IP T
metastasis) of prostate cancer 5 and 10 years after a prostatectomy.
Statistical analysis
We used descriptive statistics to summarize our sample’s sociodemographic and clinical
AN US
characteristics, predictions from the tool, and other covariates. Descriptive p-values were calculated via χ2 tests and two-sample t-tests to assess differences in characteristics between patients enrolled in the control and intervention phases. To
M
estimate the impact of the tool’s predictions on the outcomes measured both pre- and post-visit, for each measure we subtracted the change in score (from pre- to post-) of
ED
the control group from the change in score of the intervention group. Two-sample t-tests were used to determine whether the changes from pre- to post-visit significantly differed
PT
between the two groups. Due to the small sample size and similarity in characteristics
CE
between the control and intervention groups, we did not perform multivariable analyses.
AC
Analyses were performed using Microsoft Excel (2016) and R version 3.3.2.23
Results
Sociodemographic and clinical characteristics Overall, 67 men were included in the study including 36 in the control group and 31 in the intervention group (Table 1). Nineteen men (28.4%) were seen by one provider, and
ACCEPTED MANUSCRIPT
48 (71.6%) were seen by the other provider. Men were mostly White (87%), less than 75 years old (58%), married (91%), and had a Master’s degree or higher (57%), without significant differences between groups. Most men in both groups met the pre-specified Johns Hopkins criteria to join AS (76%), were enrolled in AS for about seven years, and
CR IP T
had a slight but statistically insignificant PSA increase since diagnosis (1.1 ng/mL/cm3).
Though only men in the intervention phase were shown their risk predictions, the two groups of men were also similar in the various predictions generated from the tool, with
AN US
less than one in ten (8.6%) predicted to be placed in a higher risk grade group from a same-day biopsy, more than two-thirds (70%) predicted to have the same pathologic grade group as their current grade group, and nearly all predicted to be cured of
ED
after receiving a prostatectomy.
M
prostate cancer both five (96% average chance) and ten years (94% average chance)
Both groups of men showed a moderate preference for their physician to make
PT
decisions about receiving a prostate biopsy (7.1 out of 10), and answered the medical
CE
statistics question with roughly equal accuracy (71% overall) (Table 1).
AC
Primary and secondary outcomes Using the Cancer Control scale before and after AS follow-up visits, we found that perceived cancer control increased in both groups from pre- to post-visit (control ∆ = +6.3, intervention ∆ = +12.8), with a greater increase seen among the intervention group (intervention ∆ - control ∆ = +6.5, p=0.04). The intervention group had lower pre-visit
ACCEPTED MANUSCRIPT
cancer control scores than the control group (65.6 vs. 76.8) while still having roughly equal post-visit scores (78.4 vs. 83.1) (Table 2). Pre- to post-visit changes in anxiety specific to the biopsy and its physical consequences was not significantly different between the control and intervention groups (-0.5 vs. -0.3, p=0.54), nor were the
CR IP T
changes in feelings of being overwhelmed by prostate cancer-related information (-0.1 vs. +0.1, p=0.40).
Men in the intervention group appeared to have an adequate understanding of the tool’s
AN US
predictions, answering roughly half of the questions correctly overall (46%) (Table 3). More than three-quarters (82%) agreed or strongly agreed that they understood the tool’s predictions, almost all (97%) felt that the predictions helped them better
M
understand their cancer risk, and more than three-quarters reported that the predictions gave them a better comprehension of the chance of an upgraded biopsy (77%) and
PT
Discussion
ED
more insight into whether they should get a biopsy (77%).
CE
Through this evaluation of the impact of our individualized risk prediction tool on men in AS at our institution, we identified several key findings. Compared to men who were not
AC
shown the tool’s predictions during their visit, men who were shown their predictions reported a larger increase in perceived control over their cancer without feeling overwhelmed by the amount of cancer-related information. Men exposed to the tool had a satisfactory understanding of its predictions, and mostly felt that the predictions helped them understand their cancer risk and aided their decision about whether to get
ACCEPTED MANUSCRIPT
a biopsy. These findings highlight the potential for using individualized tools to better inform current and eligible AS patients about their disease risk while not exacerbating health-related stress and feelings of being overloaded by cancer information.
CR IP T
Despite evidence of disease progression being the primary reason that men discontinue AS to pursue curative treatment, a large proportion drop out, at least in part, to their perception that surveillance is riskier than curative treatment.24 Prior research with
patients from a multisite group suggests that men on observation felt significantly less
AN US
control over their cancer (53.1 out of 100 on Cancer Control scale) compared to men who chose surgery (73.2), radiation (67.1), or hormone therapy (63.9).25 In contrast, the level of control of the Johns Hopkins AS cohort was high even prior to the follow-up visit,
M
which may be due to several factors including the program’s long-standing nature and strict inclusion criteria. It is notable that despite these high levels of control at baseline,
ED
which may lead to ceiling effects,26 we observed a significant increase in cancer control
PT
for men who were shown their risk predictions.
CE
Patients face a flood of cancer recommendations that may become ambiguous over time and discourage them from seeking cancer-related information.27,28 However, we did
AC
not find that exposure to the tool’s predictions changed men’s feelings of being overwhelmed by information. This finding may have been influenced by our sample’s relatively high level of education.29 Considering that patients with low numeracy are less inclined to choose AS initially30 and correspondingly may be more likely to drop out soon after starting, a crucial next step will be to replicate this study with a less educated
ACCEPTED MANUSCRIPT
sample. If less educated patients are overwhelmed by additional risk information, it may be necessary to reevaluate how detailed the provided information should be and the best ways to communicate risk information in order to influence patients’ decision-
CR IP T
making.31
For individualized risk prediction tools to be effective, patients need to understand what the predictions mean. In AS specifically, patients who feel better informed about their prostate cancer are less conflicted over the efficacy of AS32 and less fearful of disease
AN US
progression.33 Men in our sample who learned about their predictions considered the predictions to be most useful in helping them understand their risk of currently having aggressive prostate cancer and least useful in helping them decide whether to get a
M
biopsy. One reason that men may find the predictions less useful for the biopsy decision is because they rely on their provider’s opinion. Men in the intervention group on
ED
average reported a preference for their provider to make treatment decisions, and
PT
accordingly may have felt that the biopsy decision was not theirs to make.
CE
Men in our study have been enrolled in AS for roughly seven years on average (range 1 to 16). In a large Swedish AS population followed for five years, men discontinued AS
AC
due to personal reasons mostly by two years after starting.11 Since most men we surveyed were past this critical two-year period, it is unlikely that many of them were at a high risk of dropping out of AS due to personal reasons such as anxiety. Future studies that specifically target and follow men who recently joined AS will be an important step to evaluate the generalizability and usefulness of the tool. It is plausible
ACCEPTED MANUSCRIPT
that even without many data points to individualize risk predictions, men who know that they are being followed with this data-driven approach may be more likely to remain in AS. Moreover, it is necessary to evaluate whether the provision of these types of tools
CR IP T
lead to increased uptake of AS in the first place.
The results from our study should be considered with several limitations in mind. First, we collected data from a relatively small sample at a single site, thereby limiting the generalizability of the study. It is possible that patients at our institution have different
AN US
levels of perceived control than patients at other institutions, leaving room for more or less improvement from the risk prediction tool. Calibrating the underlying statistical model to produce accurate risk predictions using different cohorts and deploying it in
M
additional clinical sites is an important next step. Because patients at our institution are unlikely to be representative of other AS patients, the model cannot be simply
ED
transported to other institutions. We are currently working on a model that combines data from data from several AS cohorts and adjusts for known differences while sharing
PT
information where appropriate, such as how prostate volume and age affect PSA. Our
CE
cohort may have also been unique in their high average level of education, which may have improved their understanding of the risk predictions. Second, our outcome
AC
measures were slightly modified based on existing, validated measures to more closely align with the cancer risk being presented. Third, there was a nearly six-month period between control and intervention phases and more than one provider who counseled men about their risk over the course of the study. Because our AS program and its providers are well-established without other program changes during the study period,
ACCEPTED MANUSCRIPT
we do not expect that secular trends or provider variation were significant. Fourth, we did not include patient comorbidity in the study. Patients’ health history may influence how they perceive risk of their cancer and of continuing in AS and will be important to investigate in future work. Finally, we surveyed men immediately after their visit in an
CR IP T
attempt to capture their feelings at their most salient. It is possible that doubt and
anxiety could peak later and that men’s understanding of the tool’s predictions could diminish.
AN US
Conclusion
This study underscores the potential value of using a statistical tool to help men in AS better understand their cancer risk, allowing men in AS and their providers to be more
M
confident and informed about the most suitable way to manage prostate cancer. It will be important to investigate the extent to which tools such as this help retain men in AS
ED
and, even earlier in the process, help men decide to elect AS as their initial prostate
CE
References
PT
cancer management strategy.
AC
1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2016: Cancer Statistics, 2016. CA Cancer J Clin. 2016;66(1):7-30. doi:10.3322/caac.21332 2. Mohler J, Bahnson RR, Boston B, et al. NCCN clinical practice guidelines in oncology: prostate cancer. J Natl Compr Cancer Netw JNCCN. 2010;8(2):162-200. 3. Cooperberg MR, Broering JM, Carroll PR. Time Trends and Local Variation in Primary Treatment of Localized Prostate Cancer. J Clin Oncol. 2010;28(7):1117-1123. doi:10.1200/JCO.2009.26.0133 4. Resnick MJ, Koyama T, Fan K-H, et al. Long-term functional outcomes after treatment for localized prostate cancer. N Engl J Med. 2013;368(5):436-445. doi:10.1056/NEJMoa1209978
ACCEPTED MANUSCRIPT
AC
CE
PT
ED
M
AN US
CR IP T
5. Dall’Era MA, Albertsen PC, Bangma C, et al. Active surveillance for prostate cancer: a systematic review of the literature. Eur Urol. 2012;62(6):976-983. doi:10.1016/j.eururo.2012.05.072 6. Fenton JJ, Weyrich MS, Durbin S, Liu Y, Bang H, Melnikow J. Prostate-Specific Antigen-Based Screening for Prostate Cancer: Evidence Report and Systematic Review for the US Preventive Services Task Force. JAMA. 2018;319(18):1914-1931. doi:10.1001/jama.2018.3712 7. Berger ZD, Yeh JC, Carter HB, Pollack CE. Characteristics and experiences of patients with localized prostate cancer who left an active surveillance program. The Patient. 2014;7(4):427-436. doi:10.1007/s40271-014-0066-z 8. Lang MF, Tyson MD, Alvarez JR, et al. The Influence of Psychosocial Constructs on the Adherence to Active Surveillance for Localized Prostate Cancer in a Prospective, Population-based Cohort. Urology. 2017;103:173-178. doi:10.1016/j.urology.2016.12.063 9. Patel MI, DeConcini DT, Lopez-Corona E, Ohori M, Wheeler T, Scardino PT. An analysis of men with clinically localized prostate cancer who deferred definitive therapy. J Urol. 2004;171(4):1520-1524. doi:10.1097/01.ju.0000118224.54949.78 10. Venderbos LDF, van den Bergh RCN, Roobol MJ, et al. A longitudinal study on the impact of active surveillance for prostate cancer on anxiety and distress levels. Psychooncology. 2015;24(3):348-354. doi:10.1002/pon.3657 11. Loeb S, Folkvaljon Y, Makarov DV, Bratt O, Bill-Axelson A, Stattin P. Five-year Nationwide Follow-up Study of Active Surveillance for Prostate Cancer. Eur Urol. 2015;67(2):233-238. doi:10.1016/j.eururo.2014.06.010 12. Martin RM, Gunnell D, Hamdy F, Neal D, Lane A, Donovan J. Continuing controversy over monitoring men with localized prostate cancer: a systematic review of programs in the prostate specific antigen era. J Urol. 2006;176(2):439-449. doi:10.1016/j.juro.2006.03.030 13. Dall’Era MA, Cooperberg MR, Chan JM, et al. Active surveillance for early-stage prostate cancer: Review of the current literature. Cancer. 2008;112(8):1650-1659. doi:10.1002/cncr.23373 14. Xu J, Neale AV, Dailey RK, Eggly S, Schwartz KL. Patient perspective on watchful waiting/active surveillance for localized prostate cancer. J Am Board Fam Med JABFM. 2012;25(6):763-770. doi:10.3122/jabfm.2012.06.120128 15. Coley RY, Fisher AJ, Mamawala M, Carter HB, Pienta KJ, Zeger SL. A Bayesian hierarchical model for prediction of latent health states from multiple data sources with application to active surveillance of prostate cancer. Biometrics. 2017;73(2):625-634. doi:10.1111/biom.12577 16. Coley RY, Zeger SL, Mamawala M, Pienta KJ, Carter HB. Prediction of the Pathologic Gleason Score to Inform a Personalized Management Program for Prostate Cancer. Eur Urol. 2017;72(1):135-141. doi:10.1016/j.eururo.2016.08.005 17. Tosoian JJ, Trock BJ, Landis P, et al. Active surveillance program for prostate cancer: an update of the Johns Hopkins experience. J Clin Oncol Off J Am Soc Clin Oncol. 2011;29(16):2185-2190. doi:10.1200/JCO.2010.32.8112 18. Clark JA, Bokhour BG, Inui TS, Silliman RA, Talcott JA. Measuring patients’ perceptions of the outcomes of treatment for early prostate cancer. Med Care. 2003;41(8):923-936. doi:10.1097/01.MLR.0000078147.80071.78
ACCEPTED MANUSCRIPT
AC
CE
PT
ED
M
AN US
CR IP T
19. Jensen JD, Carcioppolo N, King AJ, Scherr CL, Jones CL, Niederdieppe J. The cancer information overload (CIO) scale: establishing predictive and discriminant validity. Patient Educ Couns. 2014;94(1):90-96. doi:10.1016/j.pec.2013.09.016 20. Costa DSJ, Smith A, Lim BT, Fardell JE. Simplifying the assessment of cancer information overload: A comment on Jensen et al. (2014). Patient Educ Couns. 2015;98(11):1450. doi:10.1016/j.pec.2015.04.020 21. Health Measures. List of Adult Measures. http://www.healthmeasures.net/explore-measurement-systems/promis/intro-topromis/list-of-adult-measures. Accessed February 14, 2018. 22. Lipkus IM, Samsa G, Rimer BK. General Performance on a Numeracy Scale among Highly Educated Samples. Med Decis Making. 2001;21(1):37-44. doi:10.1177/0272989X0102100105 23. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Core Team; 2016. https://www.R-project.org/. 24. Kim SP, Gross CP, Nguyen PL, et al. Perceptions of Active Surveillance and Treatment Recommendations for Low-risk Prostate Cancer: Results from a National Survey of Radiation Oncologists and Urologists. Med Care. 2014;52(7):579-585. doi:10.1097/MLR.0000000000000155 25. Clark JA, Inui TS, Silliman RA, et al. Patients’ perceptions of quality of life after treatment for early prostate cancer. J Clin Oncol. 2003;21(20):3777-3784. doi:10.1200/JCO.2003.02.115 26. Ceiling Effect. In: The SAGE Encyclopedia of Social Science Research Methods. 2455 Teller Road, Thousand Oaks California 91320 United States of America: Sage Publications, Inc.; 2004. doi:10.4135/9781412950589.n102 27. Tan ASL, Nagler RH, Hornik RC, DeMichele A. Evolving Information Needs among Colon, Breast, and Prostate Cancer Survivors: Results from a Longitudinal Mixed-Effects Analysis. Cancer Epidemiol Biomarkers Prev. 2015;24(7):1071-1078. doi:10.1158/1055-9965.EPI-15-0041 28. National Cancer Institute. HINTS Briefs Number 9: Confusion about Cancer Prevention. https://hints.cancer.gov/docs/Briefs/HINTS_Brief_9_010708.pdf. 29. Kim K, Lustria MLA, Burke D, Kwon N. Predictors of cancer information overload: findings from a national survey. Inf Res. 2007;12(4). http://students.lti.cs.cmu.edu/11899/files/cp3a_readingw1_cancerarticle.pdf. 30. López-Pérez B, Barnes A, Frosch DL, Hanoch Y. Predicting prostate cancer treatment choices: The role of numeracy, time discounting, and risk attitudes. J Health Psychol. 2017;22(6):788-797. doi:10.1177/1359105315615931 31. Hibbard JH, Peters E. Supporting Informed Consumer Health Care Decisions: Data Presentation Approaches that Facilitate the Use of Information in Choice. Annu Rev Public Health. 2003;24(1):413-433. doi:10.1146/annurev.publhealth.24.100901.141005 32. Goh AC, Kowalkowski MA, Bailey Jr DE, Kazer MW, Knight SJ, Latini DM. Perception of cancer and inconsistency in medical information are associated with decisional conflict: a pilot study of men with prostate cancer who undergo active surveillance. BJU Int. 2012;110(2b):E50-E56. doi:10.1111/j.1464-410X.2011.10791.x 33. Parker PA, Davis JW, Latini DM, et al. Relationship between illness uncertainty, anxiety, fear of progression and quality of life in men with favourable-risk prostate
ACCEPTED MANUSCRIPT
cancer undergoing active surveillance. BJU Int. 2016;117(3):469-477. doi:10.1111/bju.13099
AC
CE
PT
ED
M
AN US
CR IP T
TABLES & FIGURES
Figure 1. A set of sample predictions from the tool provided to men in the intervention group
Legend: In addition to the predicted biopsy upgrade, prognostic grade group, and 5year outcomes (above), men were shown a chart of their PSA values over time,
ACCEPTED MANUSCRIPT
predicted 10-year outcomes, an explanation of grade groups, and text explanations of
AC
CE
PT
ED
M
AN US
CR IP T
the predictions (see Appendix for full output).
ACCEPTED MANUSCRIPT
Table 1. Demographics and cancer-related characteristics of sample
TOTAL Demographics Race White Non-White Age <75 years ≥75 years Relationship status Married Other Education level Master’s degree or higher Less than Master’s degree Employment status Full-time Other
32 (89) 4 (11)
26 (84) 5 (16)
20 (56) 16 (44)
19 (61) 12 (39)
32 (91) 3 (9)
28 (90) 3 (10)
CR IP T
AN US
M
ED
CE
PT
Outcomes from statistical tool Upgrade probability on next biopsy, mean (SD) Estimated probability of current grade group matching grade group from same-day prostatectomy, mean (SD) Predicted cure chance 5 years post-prostatectomy, mean (SD) Predicted cure chance 10 years postprostatectomy, mean (SD)
AC
Intervention n (%) 31 (100)
p-value
0.55
Cancer-related characteristics Met all pre-specified AS eligibility criteria Years in active surveillance, mean (SD) Change in PSA since joining AS, mean (SD)
Other covariates Numeracy Finds medical statistics easy or very easy Correctly answered medical statistics question Decision autonomy, mean (SD)a a
Control n (%) 36 (100)
0.64 0.85 0.91
18 (60) 12 (40)
17 (59) 12 (41)
9 (27) 24 (73)
13 (42) 18 (58)
28 (78) 7.5 (3.6) 0.9 (3.5)
23 (74) 6.6 (4.4) 1.4 (4.0)
0.73 0.34 0.57
8.4% (3.3)
8.8% (3.7)
0.58
70.6 (18.9)
68.4 (18.0)
0.62
95.6% (2.4)
0.36
93.4% (2.3)
0.41
20 (67) 20 (67) 7.4 (2.4)
0.22 0.50 0.33
96.1% (2.1) 93.9% (2.0)
28 (80) 26 (74) 6.8 (2.4)
0.14
Measures whether patient prefers himself or his physician to make decisions regarding his prostate cancer min = 0 (complete patient autonomy), max = 10 (complete physician autonomy)
ACCEPTED MANUSCRIPT
Table 2. Comparisons of outcomes by study arm Control (n = 42)
Intervention change minus control change
p-value
+12.8
+6.5
0.04
-0.3
+0.2
0.54
+0.10
0.40
Postvisit
Change
Previsit
PostChange visit
76.8
83.1
+6.3
65.6
78.4
2.5
2.0
-0.5
3.0
2.7
1.8
1.7
-0.1
a
CR IP T
Previsit
AN US
Perceived cancer controla Biopsyrelated anxietyb Cancer information overloadc
Intervention (n = 31)
1.6
1.6
0.0
ED
M
Measures patients’ worry/uncertainty about cancer spreading or progressing, and views about efficacy of active surveillance min = 1 (not at all worried), max = 5 (very worried); scoring is flipped and scaled to fall between 0– 100, with higher scores indicating better perceived control b Measures how worried patients are about the biopsy and its physical consequences min = 1 (not at all worried), max = 5 (very worried) c Measures how overwhelmed patients feel by, and how much they care about, cancer-related information min = 1 (not at all overwhelmed), max = 5 (very overwhelmed)
PT
Table 3. Understanding of predictions and perceived usefulness of tool reported by men in the intervention group (n=31)
CE
Understanding of tool Grade group upgrade probability on same-day biopsy Grade group upgrade probability on same-day surgical removal Ten-year post-surgical cancer recurrence Ten-year post-surgical metastasis
AC
Usefulness of tool Understand information in tool Helped understand risk of currently having aggressive cancer Helped understand chance of upgrade on biopsy Helped decide whether or not to get a biopsy
Correct responses (%) 13 (42) 2 (6) 20 (65) 21 (70) Agree or strongly agree (%) 23 (82) 30 (97) 24 (77) 24 (77)
Clinical evaluation of an individualized risk prediction tool for men on active surveillance for prostate cancer Joseph H. Huntley BS , Rebecca Y. Coley PhD , H. Ballentine Carter MD , Archana Radhakrishnan MD, MHS , Melinda Krakow PhD, MPH, MA , Craig E. Pollack MD, MHS PII: DOI: Reference:
S0090-4295(18)30903-8 https://doi.org/10.1016/j.urology.2018.08.021 URL 21213
To appear in:
Urology
Received date: Revised date: Accepted date:
30 May 2018 15 August 2018 17 August 2018
Please cite this article as: Joseph H. Huntley BS , Rebecca Y. Coley PhD , H. Ballentine Carter MD , Archana Radhakrishnan MD, MHS , Melinda Krakow PhD, MPH, MA , Craig E. Pollack MD, MHS , Clinical evaluation of an individualized risk prediction tool for men on active surveillance for prostate cancer, Urology (2018), doi: https://doi.org/10.1016/j.urology.2018.08.021
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Title: Clinical evaluation of an individualized risk prediction tool for men on active surveillance for prostate cancer Authors and affiliations: Joseph H. Huntley, BS; Department of Medicine, Johns Hopkins School of Medicine, Baltimore, MD, USA; [email protected]
CR IP T
Rebecca Y. Coley, PhD; Kaiser Permanente Washington Health Research Institute, Seattle, WA, USA; [email protected] H. Ballentine Carter, MD; Department of Urology, The James Buchanan Brady Urological Institute, Johns Hopkins School of Medicine, Baltimore, MD, USA; [email protected]
AN US
Archana Radhakrishnan, MD, MHS; Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, MI, USA; [email protected]
Melinda Krakow, PhD, MPH, MA; National Cancer Institute, National Institutes of Health, Bethesda, MD, USA; [email protected]
M
Craig E. Pollack, MD, MHS; Department of Medicine, Johns Hopkins School of Medicine, Baltimore, MD, USA; [email protected]
CE
PT
ED
Corresponding author: Craig E. Pollack, MD, MHS Associate Professor, Division of General Internal Medicine Johns Hopkins School of Medicine 2024 E. Monument Street, Suite 2-519 Baltimore, MD 21287 Phone: 410-955-4201 Email: [email protected] Conflict of interest statement: All authors declare no conflict of interest.
AC
Acknowledgments: We would like to acknowledge funding from the Maryland Cigarette Restitution Fund. Keywords: risk assessment, prostatic neoplasms, active surveillance
ACCEPTED MANUSCRIPT
Abstract Objective: To determine whether providing individualized predictions of health outcomes to men on active surveillance (AS) alleviates cancer-related anxiety and
CR IP T
improves risk understanding.
Materials and Methods: We consecutively recruited men from our large, institutional AS program before (n=36) and after (n=31) implementation of a risk prediction tool. Men
AN US
in both groups were surveyed before and after their regular visits to assess their
perceived cancer control, biopsy-specific anxiety, and burden from cancer-related information. We compared pre-/post-visit differences between men who were and were not shown the tool using two-sample t-tests. Satisfaction with and understanding of the
ED
M
predictions were elicited from men in the intervention period.
Results: Men reported a relatively high level of cancer control at baseline. Men who
PT
were not shown the tool saw a 6.3 point increase (scaled from 0 to 100) in their perceived cancer control from before to after their visit whereas men who were shown
CE
the tool saw a 12.8 point increase, indicating a statistically significant difference between groups (p=0.04). Biopsy-specific anxiety and burden from cancer information
AC
was not significantly different between groups. Men were satisfied with the tool and demonstrated moderate understanding.
Conclusion: Providing individualized predictions to men on AS helps them better understand their cancer risk and should be considered at other clinical sites.
ACCEPTED MANUSCRIPT
Introduction Over 180,000 men are diagnosed with prostate cancer every year in the United States.1 An large proportion of all new prostate cancer diagnoses are indolent tumors that do not require immediate curative intervention.2 Yet, the majority of men with favorable risk
CR IP T
cancer will choose surgical removal or radiation of the prostate,3 which poses
considerable risk of permanent side effects including incontinence and impotence, and can be physically, emotionally, and financially burdensome.4
AN US
Despite evidence that active surveillance (AS) is a safe alternative to curative
intervention for favorable risk prostate cancer, uncertainty regarding the underlying prostate cancer precludes many men and their physicians from electing AS as a rational
M
strategy for managing their cancer.5,6 Further, previous literature has reported that up to 20% of men drop out of AS without evidence of disease progression,7–13 with the
ED
uncertainty involved in being on AS and fear of cancer spreading as primary drivers of
PT
their decision-making.7,14
CE
Data-driven tools that help men in AS understand their underlying risk of developing more aggressive prostate cancer may alleviate anxiety and improve adherence to AS.
AC
Despite the potential for these approaches, such tools have not, to our knowledge, been evaluated in clinical care. We sought to investigate whether a tool that provides individualized predictions for risk of developing more aggressive cancer15,16 is associated with lower self-reported worry about prostate cancer among men in AS. We hypothesized that men who received individualized risk predictions would report feeling
ACCEPTED MANUSCRIPT
more control (i.e., less worried) over their cancer relative to men who did not receive predictions, and would not feel more burdened by cancer-related information.
Patients and Methods
CR IP T
The Johns Hopkins Institutional Review Boards determined this project to be a quality improvement effort that did not constitute human subjects research.
Individualized risk prediction tool
AN US
The tool for individualized risk prediction for men in the Johns Hopkins AS program has been previously reported and its predictive validity assessed.15,16 Briefly, the underlying model combines data on men’s repeated prostate-specific antigen measurements
M
(PSA) and prostate biopsies within a Bayesian hierarchical model to predict the underlying grade of an individual’s prostate cancer, given clinical observations of similar
ED
patients. The model requires at least two PSA observations and one biopsy after diagnosis to generate an individualized prediction. When men are new to AS, their
PT
predicted pathological Gleason scores are closer to those of the population; however,
CE
as men receive more PSA tests and biopsies, the model generates predictions that
AC
allow providers to better determine how men vary from the population average.
The tool presents the probability that the individual’s next biopsy will be upgraded, as well as 5- and 10-year cure rates, biochemical recurrence rates, and metastasis rates after definitive treatment (Figure 1 and Appendix).
ACCEPTED MANUSCRIPT
Sample selection Recruitment was performed in two phases. Men in the Johns Hopkins AS program17 return to Johns Hopkins every six to twelve months for evaluation. Men who returned for a follow-up visit between 9 February 2017–4 April 2017 were eligible to be included in
CR IP T
the control phase and were consecutively recruited. These men received standard of care and were counseled about the risks associated with AS but were not shown
specific predictions from the tool during their visits. Men who returned for a follow-up visit between 20 September 2017–8 February 2018 were eligible to be included in the
AN US
intervention phase. These men were shown individualized predictions from the tool during their visit, discussed the predictions with their provider, and were given a print copy. Men in the study met with one of two AS providers who worked closely together.
M
The time between the control and intervention phases was spent integrating the tool into routine clinical care. Men who were enrolled in AS for a short time and had few PSA
ED
measurements or biopsies did not have enough data for accurate risk prediction and were excluded from the study sample. All eligible men agreed to participate. There were
CE
PT
no men who were included in both the control and intervention phases of the study.
Data collection
AC
Men were given paper surveys to complete before and after meeting with their provider. They were encouraged to complete the second survey immediately after their visit, but were given the option to return it by mail. A follow-up call was made to men who elected the mail-in option to ensure that they completed the survey within three days after their
ACCEPTED MANUSCRIPT
visit. Sociodemographic and clinical variables were extracted from the existing AS database.
Outcome measures
CR IP T
Our main outcome was change in men’s perceived control over their prostate cancer, measured by the validated Cancer Control scale18 administered before and after men’s visits. This is a 5-point Likert scale which uses five items to determine how worried
individuals are about their cancer spreading or becoming more aggressive, and how
AN US
confident they are that their treatment option works and their cancer is under control. Responses were scaled to fall between 0–100 and adjusted so that higher scores
M
reflected improved control.
We also aimed to determine whether men felt overwhelmed by cancer-related
ED
information. This was measured by a modified version of the Cancer Information Overload scale,19,20 which was composed of five items and scored according to a 5-
PT
point Likert scale. Lastly, we measured anxiety specific to the prostate biopsy through a
CE
modified, two question version of the PROMIS Emotional Distress – Anxiety Short Form 7a.21 Men responded on a 5-point Likert scale about their anxiety due to the prostate
AC
biopsy and its physical consequences (see Appendix). These two scales were asked on pre- and post-visit surveys.
For men in the intervention phase, we also measured understanding of the tool’s predictions by asking four multiple choice questions about predictions that were
ACCEPTED MANUSCRIPT
presented explicitly by the tool, as well as four questions (5-point Likert) about perceived usefulness and understandability of the tool’s predictions (see Appendix). Responses to questions about the explicit predictions were scored for accuracy. These
CR IP T
questions were asked on the post-visit survey only.
Covariates
On the pre-visit surveys, we measured medical decision autonomy and numeracy as covariates. Medical decision autonomy was determined by presenting men with a line
AN US
numbered from 0 (complete patient autonomy) to 10 (complete physician autonomy) and asking them to place a mark on the line that best represented their preference regarding the decision to receive a prostate biopsy. Numeracy was measured by asking
M
men generally how difficult it was for them to understand medical statistics, followed by
ED
one basic multiple-choice question (3 options) about medical statistics.22
From the AS database and electronic medical records system, we extracted
PT
sociodemographic characteristics including race (White or non-White), age (<75 yrs. or
CE
≥75 yrs.), relationship status (married or not married), education level (Master’s or less than Master’s), and employment status (full-time or not full-time) for all study
AC
participants. We also determined whether each participant met all pre-specified eligibility criteria (Gleason <7, positive cores <3, positive core involvement <50%, PSA density <0.15 ng/mL/cm3) before joining AS, how many years they had been in AS, and their change in PSA since diagnosis.
ACCEPTED MANUSCRIPT
Finally, for men in each recruitment phase, we used the tool’s predictions to calculate the average probability of a grade group increase from a same-day biopsy, the average probability of men’s current grade group matching their predicted pathologic grade group, and the average chance of being cured (i.e., no biochemical recurrence or
CR IP T
metastasis) of prostate cancer 5 and 10 years after a prostatectomy.
Statistical analysis
We used descriptive statistics to summarize our sample’s sociodemographic and clinical
AN US
characteristics, predictions from the tool, and other covariates. Descriptive p-values were calculated via χ2 tests and two-sample t-tests to assess differences in characteristics between patients enrolled in the control and intervention phases. To
M
estimate the impact of the tool’s predictions on the outcomes measured both pre- and post-visit, for each measure we subtracted the change in score (from pre- to post-) of
ED
the control group from the change in score of the intervention group. Two-sample t-tests were used to determine whether the changes from pre- to post-visit significantly differed
PT
between the two groups. Due to the small sample size and similarity in characteristics
CE
between the control and intervention groups, we did not perform multivariable analyses.
AC
Analyses were performed using Microsoft Excel (2016) and R version 3.3.2.23
Results
Sociodemographic and clinical characteristics Overall, 67 men were included in the study including 36 in the control group and 31 in the intervention group (Table 1). Nineteen men (28.4%) were seen by one provider, and
ACCEPTED MANUSCRIPT
48 (71.6%) were seen by the other provider. Men were mostly White (87%), less than 75 years old (58%), married (91%), and had a Master’s degree or higher (57%), without significant differences between groups. Most men in both groups met the pre-specified Johns Hopkins criteria to join AS (76%), were enrolled in AS for about seven years, and
CR IP T
had a slight but statistically insignificant PSA increase since diagnosis (1.1 ng/mL/cm3).
Though only men in the intervention phase were shown their risk predictions, the two groups of men were also similar in the various predictions generated from the tool, with
AN US
less than one in ten (8.6%) predicted to be placed in a higher risk grade group from a same-day biopsy, more than two-thirds (70%) predicted to have the same pathologic grade group as their current grade group, and nearly all predicted to be cured of
ED
after receiving a prostatectomy.
M
prostate cancer both five (96% average chance) and ten years (94% average chance)
Both groups of men showed a moderate preference for their physician to make
PT
decisions about receiving a prostate biopsy (7.1 out of 10), and answered the medical
CE
statistics question with roughly equal accuracy (71% overall) (Table 1).
AC
Primary and secondary outcomes Using the Cancer Control scale before and after AS follow-up visits, we found that perceived cancer control increased in both groups from pre- to post-visit (control ∆ = +6.3, intervention ∆ = +12.8), with a greater increase seen among the intervention group (intervention ∆ - control ∆ = +6.5, p=0.04). The intervention group had lower pre-visit
ACCEPTED MANUSCRIPT
cancer control scores than the control group (65.6 vs. 76.8) while still having roughly equal post-visit scores (78.4 vs. 83.1) (Table 2). Pre- to post-visit changes in anxiety specific to the biopsy and its physical consequences was not significantly different between the control and intervention groups (-0.5 vs. -0.3, p=0.54), nor were the
CR IP T
changes in feelings of being overwhelmed by prostate cancer-related information (-0.1 vs. +0.1, p=0.40).
Men in the intervention group appeared to have an adequate understanding of the tool’s
AN US
predictions, answering roughly half of the questions correctly overall (46%) (Table 3). More than three-quarters (82%) agreed or strongly agreed that they understood the tool’s predictions, almost all (97%) felt that the predictions helped them better
M
understand their cancer risk, and more than three-quarters reported that the predictions gave them a better comprehension of the chance of an upgraded biopsy (77%) and
PT
Discussion
ED
more insight into whether they should get a biopsy (77%).
CE
Through this evaluation of the impact of our individualized risk prediction tool on men in AS at our institution, we identified several key findings. Compared to men who were not
AC
shown the tool’s predictions during their visit, men who were shown their predictions reported a larger increase in perceived control over their cancer without feeling overwhelmed by the amount of cancer-related information. Men exposed to the tool had a satisfactory understanding of its predictions, and mostly felt that the predictions helped them understand their cancer risk and aided their decision about whether to get
ACCEPTED MANUSCRIPT
a biopsy. These findings highlight the potential for using individualized tools to better inform current and eligible AS patients about their disease risk while not exacerbating health-related stress and feelings of being overloaded by cancer information.
CR IP T
Despite evidence of disease progression being the primary reason that men discontinue AS to pursue curative treatment, a large proportion drop out, at least in part, to their perception that surveillance is riskier than curative treatment.24 Prior research with
patients from a multisite group suggests that men on observation felt significantly less
AN US
control over their cancer (53.1 out of 100 on Cancer Control scale) compared to men who chose surgery (73.2), radiation (67.1), or hormone therapy (63.9).25 In contrast, the level of control of the Johns Hopkins AS cohort was high even prior to the follow-up visit,
M
which may be due to several factors including the program’s long-standing nature and strict inclusion criteria. It is notable that despite these high levels of control at baseline,
ED
which may lead to ceiling effects,26 we observed a significant increase in cancer control
PT
for men who were shown their risk predictions.
CE
Patients face a flood of cancer recommendations that may become ambiguous over time and discourage them from seeking cancer-related information.27,28 However, we did
AC
not find that exposure to the tool’s predictions changed men’s feelings of being overwhelmed by information. This finding may have been influenced by our sample’s relatively high level of education.29 Considering that patients with low numeracy are less inclined to choose AS initially30 and correspondingly may be more likely to drop out soon after starting, a crucial next step will be to replicate this study with a less educated
ACCEPTED MANUSCRIPT
sample. If less educated patients are overwhelmed by additional risk information, it may be necessary to reevaluate how detailed the provided information should be and the best ways to communicate risk information in order to influence patients’ decision-
CR IP T
making.31
For individualized risk prediction tools to be effective, patients need to understand what the predictions mean. In AS specifically, patients who feel better informed about their prostate cancer are less conflicted over the efficacy of AS32 and less fearful of disease
AN US
progression.33 Men in our sample who learned about their predictions considered the predictions to be most useful in helping them understand their risk of currently having aggressive prostate cancer and least useful in helping them decide whether to get a
M
biopsy. One reason that men may find the predictions less useful for the biopsy decision is because they rely on their provider’s opinion. Men in the intervention group on
ED
average reported a preference for their provider to make treatment decisions, and
PT
accordingly may have felt that the biopsy decision was not theirs to make.
CE
Men in our study have been enrolled in AS for roughly seven years on average (range 1 to 16). In a large Swedish AS population followed for five years, men discontinued AS
AC
due to personal reasons mostly by two years after starting.11 Since most men we surveyed were past this critical two-year period, it is unlikely that many of them were at a high risk of dropping out of AS due to personal reasons such as anxiety. Future studies that specifically target and follow men who recently joined AS will be an important step to evaluate the generalizability and usefulness of the tool. It is plausible
ACCEPTED MANUSCRIPT
that even without many data points to individualize risk predictions, men who know that they are being followed with this data-driven approach may be more likely to remain in AS. Moreover, it is necessary to evaluate whether the provision of these types of tools
CR IP T
lead to increased uptake of AS in the first place.
The results from our study should be considered with several limitations in mind. First, we collected data from a relatively small sample at a single site, thereby limiting the generalizability of the study. It is possible that patients at our institution have different
AN US
levels of perceived control than patients at other institutions, leaving room for more or less improvement from the risk prediction tool. Calibrating the underlying statistical model to produce accurate risk predictions using different cohorts and deploying it in
M
additional clinical sites is an important next step. Because patients at our institution are unlikely to be representative of other AS patients, the model cannot be simply
ED
transported to other institutions. We are currently working on a model that combines data from data from several AS cohorts and adjusts for known differences while sharing
PT
information where appropriate, such as how prostate volume and age affect PSA. Our
CE
cohort may have also been unique in their high average level of education, which may have improved their understanding of the risk predictions. Second, our outcome
AC
measures were slightly modified based on existing, validated measures to more closely align with the cancer risk being presented. Third, there was a nearly six-month period between control and intervention phases and more than one provider who counseled men about their risk over the course of the study. Because our AS program and its providers are well-established without other program changes during the study period,
ACCEPTED MANUSCRIPT
we do not expect that secular trends or provider variation were significant. Fourth, we did not include patient comorbidity in the study. Patients’ health history may influence how they perceive risk of their cancer and of continuing in AS and will be important to investigate in future work. Finally, we surveyed men immediately after their visit in an
CR IP T
attempt to capture their feelings at their most salient. It is possible that doubt and
anxiety could peak later and that men’s understanding of the tool’s predictions could diminish.
AN US
Conclusion
This study underscores the potential value of using a statistical tool to help men in AS better understand their cancer risk, allowing men in AS and their providers to be more
M
confident and informed about the most suitable way to manage prostate cancer. It will be important to investigate the extent to which tools such as this help retain men in AS
ED
and, even earlier in the process, help men decide to elect AS as their initial prostate
CE
References
PT
cancer management strategy.
AC
1. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2016: Cancer Statistics, 2016. CA Cancer J Clin. 2016;66(1):7-30. doi:10.3322/caac.21332 2. Mohler J, Bahnson RR, Boston B, et al. NCCN clinical practice guidelines in oncology: prostate cancer. J Natl Compr Cancer Netw JNCCN. 2010;8(2):162-200. 3. Cooperberg MR, Broering JM, Carroll PR. Time Trends and Local Variation in Primary Treatment of Localized Prostate Cancer. J Clin Oncol. 2010;28(7):1117-1123. doi:10.1200/JCO.2009.26.0133 4. Resnick MJ, Koyama T, Fan K-H, et al. Long-term functional outcomes after treatment for localized prostate cancer. N Engl J Med. 2013;368(5):436-445. doi:10.1056/NEJMoa1209978
ACCEPTED MANUSCRIPT
AC
CE
PT
ED
M
AN US
CR IP T
5. Dall’Era MA, Albertsen PC, Bangma C, et al. Active surveillance for prostate cancer: a systematic review of the literature. Eur Urol. 2012;62(6):976-983. doi:10.1016/j.eururo.2012.05.072 6. Fenton JJ, Weyrich MS, Durbin S, Liu Y, Bang H, Melnikow J. Prostate-Specific Antigen-Based Screening for Prostate Cancer: Evidence Report and Systematic Review for the US Preventive Services Task Force. JAMA. 2018;319(18):1914-1931. doi:10.1001/jama.2018.3712 7. Berger ZD, Yeh JC, Carter HB, Pollack CE. Characteristics and experiences of patients with localized prostate cancer who left an active surveillance program. The Patient. 2014;7(4):427-436. doi:10.1007/s40271-014-0066-z 8. Lang MF, Tyson MD, Alvarez JR, et al. The Influence of Psychosocial Constructs on the Adherence to Active Surveillance for Localized Prostate Cancer in a Prospective, Population-based Cohort. Urology. 2017;103:173-178. doi:10.1016/j.urology.2016.12.063 9. Patel MI, DeConcini DT, Lopez-Corona E, Ohori M, Wheeler T, Scardino PT. An analysis of men with clinically localized prostate cancer who deferred definitive therapy. J Urol. 2004;171(4):1520-1524. doi:10.1097/01.ju.0000118224.54949.78 10. Venderbos LDF, van den Bergh RCN, Roobol MJ, et al. A longitudinal study on the impact of active surveillance for prostate cancer on anxiety and distress levels. Psychooncology. 2015;24(3):348-354. doi:10.1002/pon.3657 11. Loeb S, Folkvaljon Y, Makarov DV, Bratt O, Bill-Axelson A, Stattin P. Five-year Nationwide Follow-up Study of Active Surveillance for Prostate Cancer. Eur Urol. 2015;67(2):233-238. doi:10.1016/j.eururo.2014.06.010 12. Martin RM, Gunnell D, Hamdy F, Neal D, Lane A, Donovan J. Continuing controversy over monitoring men with localized prostate cancer: a systematic review of programs in the prostate specific antigen era. J Urol. 2006;176(2):439-449. doi:10.1016/j.juro.2006.03.030 13. Dall’Era MA, Cooperberg MR, Chan JM, et al. Active surveillance for early-stage prostate cancer: Review of the current literature. Cancer. 2008;112(8):1650-1659. doi:10.1002/cncr.23373 14. Xu J, Neale AV, Dailey RK, Eggly S, Schwartz KL. Patient perspective on watchful waiting/active surveillance for localized prostate cancer. J Am Board Fam Med JABFM. 2012;25(6):763-770. doi:10.3122/jabfm.2012.06.120128 15. Coley RY, Fisher AJ, Mamawala M, Carter HB, Pienta KJ, Zeger SL. A Bayesian hierarchical model for prediction of latent health states from multiple data sources with application to active surveillance of prostate cancer. Biometrics. 2017;73(2):625-634. doi:10.1111/biom.12577 16. Coley RY, Zeger SL, Mamawala M, Pienta KJ, Carter HB. Prediction of the Pathologic Gleason Score to Inform a Personalized Management Program for Prostate Cancer. Eur Urol. 2017;72(1):135-141. doi:10.1016/j.eururo.2016.08.005 17. Tosoian JJ, Trock BJ, Landis P, et al. Active surveillance program for prostate cancer: an update of the Johns Hopkins experience. J Clin Oncol Off J Am Soc Clin Oncol. 2011;29(16):2185-2190. doi:10.1200/JCO.2010.32.8112 18. Clark JA, Bokhour BG, Inui TS, Silliman RA, Talcott JA. Measuring patients’ perceptions of the outcomes of treatment for early prostate cancer. Med Care. 2003;41(8):923-936. doi:10.1097/01.MLR.0000078147.80071.78
ACCEPTED MANUSCRIPT
AC
CE
PT
ED
M
AN US
CR IP T
19. Jensen JD, Carcioppolo N, King AJ, Scherr CL, Jones CL, Niederdieppe J. The cancer information overload (CIO) scale: establishing predictive and discriminant validity. Patient Educ Couns. 2014;94(1):90-96. doi:10.1016/j.pec.2013.09.016 20. Costa DSJ, Smith A, Lim BT, Fardell JE. Simplifying the assessment of cancer information overload: A comment on Jensen et al. (2014). Patient Educ Couns. 2015;98(11):1450. doi:10.1016/j.pec.2015.04.020 21. Health Measures. List of Adult Measures. http://www.healthmeasures.net/explore-measurement-systems/promis/intro-topromis/list-of-adult-measures. Accessed February 14, 2018. 22. Lipkus IM, Samsa G, Rimer BK. General Performance on a Numeracy Scale among Highly Educated Samples. Med Decis Making. 2001;21(1):37-44. doi:10.1177/0272989X0102100105 23. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Core Team; 2016. https://www.R-project.org/. 24. Kim SP, Gross CP, Nguyen PL, et al. Perceptions of Active Surveillance and Treatment Recommendations for Low-risk Prostate Cancer: Results from a National Survey of Radiation Oncologists and Urologists. Med Care. 2014;52(7):579-585. doi:10.1097/MLR.0000000000000155 25. Clark JA, Inui TS, Silliman RA, et al. Patients’ perceptions of quality of life after treatment for early prostate cancer. J Clin Oncol. 2003;21(20):3777-3784. doi:10.1200/JCO.2003.02.115 26. Ceiling Effect. In: The SAGE Encyclopedia of Social Science Research Methods. 2455 Teller Road, Thousand Oaks California 91320 United States of America: Sage Publications, Inc.; 2004. doi:10.4135/9781412950589.n102 27. Tan ASL, Nagler RH, Hornik RC, DeMichele A. Evolving Information Needs among Colon, Breast, and Prostate Cancer Survivors: Results from a Longitudinal Mixed-Effects Analysis. Cancer Epidemiol Biomarkers Prev. 2015;24(7):1071-1078. doi:10.1158/1055-9965.EPI-15-0041 28. National Cancer Institute. HINTS Briefs Number 9: Confusion about Cancer Prevention. https://hints.cancer.gov/docs/Briefs/HINTS_Brief_9_010708.pdf. 29. Kim K, Lustria MLA, Burke D, Kwon N. Predictors of cancer information overload: findings from a national survey. Inf Res. 2007;12(4). http://students.lti.cs.cmu.edu/11899/files/cp3a_readingw1_cancerarticle.pdf. 30. López-Pérez B, Barnes A, Frosch DL, Hanoch Y. Predicting prostate cancer treatment choices: The role of numeracy, time discounting, and risk attitudes. J Health Psychol. 2017;22(6):788-797. doi:10.1177/1359105315615931 31. Hibbard JH, Peters E. Supporting Informed Consumer Health Care Decisions: Data Presentation Approaches that Facilitate the Use of Information in Choice. Annu Rev Public Health. 2003;24(1):413-433. doi:10.1146/annurev.publhealth.24.100901.141005 32. Goh AC, Kowalkowski MA, Bailey Jr DE, Kazer MW, Knight SJ, Latini DM. Perception of cancer and inconsistency in medical information are associated with decisional conflict: a pilot study of men with prostate cancer who undergo active surveillance. BJU Int. 2012;110(2b):E50-E56. doi:10.1111/j.1464-410X.2011.10791.x 33. Parker PA, Davis JW, Latini DM, et al. Relationship between illness uncertainty, anxiety, fear of progression and quality of life in men with favourable-risk prostate
ACCEPTED MANUSCRIPT
cancer undergoing active surveillance. BJU Int. 2016;117(3):469-477. doi:10.1111/bju.13099
AC
CE
PT
ED
M
AN US
CR IP T
TABLES & FIGURES
Figure 1. A set of sample predictions from the tool provided to men in the intervention group
Legend: In addition to the predicted biopsy upgrade, prognostic grade group, and 5year outcomes (above), men were shown a chart of their PSA values over time,
ACCEPTED MANUSCRIPT
predicted 10-year outcomes, an explanation of grade groups, and text explanations of
AC
CE
PT
ED
M
AN US
CR IP T
the predictions (see Appendix for full output).
ACCEPTED MANUSCRIPT
Table 1. Demographics and cancer-related characteristics of sample
TOTAL Demographics Race White Non-White Age <75 years ≥75 years Relationship status Married Other Education level Master’s degree or higher Less than Master’s degree Employment status Full-time Other
32 (89) 4 (11)
26 (84) 5 (16)
20 (56) 16 (44)
19 (61) 12 (39)
32 (91) 3 (9)
28 (90) 3 (10)
CR IP T
AN US
M
ED
CE
PT
Outcomes from statistical tool Upgrade probability on next biopsy, mean (SD) Estimated probability of current grade group matching grade group from same-day prostatectomy, mean (SD) Predicted cure chance 5 years post-prostatectomy, mean (SD) Predicted cure chance 10 years postprostatectomy, mean (SD)
AC
Intervention n (%) 31 (100)
p-value
0.55
Cancer-related characteristics Met all pre-specified AS eligibility criteria Years in active surveillance, mean (SD) Change in PSA since joining AS, mean (SD)
Other covariates Numeracy Finds medical statistics easy or very easy Correctly answered medical statistics question Decision autonomy, mean (SD)a a
Control n (%) 36 (100)
0.64 0.85 0.91
18 (60) 12 (40)
17 (59) 12 (41)
9 (27) 24 (73)
13 (42) 18 (58)
28 (78) 7.5 (3.6) 0.9 (3.5)
23 (74) 6.6 (4.4) 1.4 (4.0)
0.73 0.34 0.57
8.4% (3.3)
8.8% (3.7)
0.58
70.6 (18.9)
68.4 (18.0)
0.62
95.6% (2.4)
0.36
93.4% (2.3)
0.41
20 (67) 20 (67) 7.4 (2.4)
0.22 0.50 0.33
96.1% (2.1) 93.9% (2.0)
28 (80) 26 (74) 6.8 (2.4)
0.14
Measures whether patient prefers himself or his physician to make decisions regarding his prostate cancer min = 0 (complete patient autonomy), max = 10 (complete physician autonomy)
ACCEPTED MANUSCRIPT
Table 2. Comparisons of outcomes by study arm Control (n = 42)
Intervention change minus control change
p-value
+12.8
+6.5
0.04
-0.3
+0.2
0.54
+0.10
0.40
Postvisit
Change
Previsit
PostChange visit
76.8
83.1
+6.3
65.6
78.4
2.5
2.0
-0.5
3.0
2.7
1.8
1.7
-0.1
a
CR IP T
Previsit
AN US
Perceived cancer controla Biopsyrelated anxietyb Cancer information overloadc
Intervention (n = 31)
1.6
1.6
0.0
ED
M
Measures patients’ worry/uncertainty about cancer spreading or progressing, and views about efficacy of active surveillance min = 1 (not at all worried), max = 5 (very worried); scoring is flipped and scaled to fall between 0– 100, with higher scores indicating better perceived control b Measures how worried patients are about the biopsy and its physical consequences min = 1 (not at all worried), max = 5 (very worried) c Measures how overwhelmed patients feel by, and how much they care about, cancer-related information min = 1 (not at all overwhelmed), max = 5 (very overwhelmed)
PT
Table 3. Understanding of predictions and perceived usefulness of tool reported by men in the intervention group (n=31)
CE
Understanding of tool Grade group upgrade probability on same-day biopsy Grade group upgrade probability on same-day surgical removal Ten-year post-surgical cancer recurrence Ten-year post-surgical metastasis
AC
Usefulness of tool Understand information in tool Helped understand risk of currently having aggressive cancer Helped understand chance of upgrade on biopsy Helped decide whether or not to get a biopsy
Correct responses (%) 13 (42) 2 (6) 20 (65) 21 (70) Agree or strongly agree (%) 23 (82) 30 (97) 24 (77) 24 (77)