Racial disparities in bladder cancer outcomes have been documented with poorer survival observed among blacks. Bladder cancer outcomes in other ethnic minority groups are less well described. We examined trends in bladder cancer survival among whites, blacks, Hispanics, and Asian/Pacific Islanders in the US over a 30-year period.
From the Surveillance, Epidemiology and End Results cancer registry data, we identified patients diagnosed with transitional cell carcinoma of the bladder between 1975 and 2005. This cohort included 163,973 white, 7,731 black, 7,364 Hispanic and 5,934 Asian/Pacific Islander patients. We assessed the relationship between ethnicity and patient characteristics. Disease-specific 5-year survival was estimated for each ethnic group and for subgroups of stage and grade.
Blacks presented with higher stage disease than whites, Hispanics and Asian/Pacific Islanders, although a trend toward earlier stage presentation was observed in all groups over time. Five-year disease-specific survival was consistently worse for blacks than for other ethnic groups, even when stratified by stage and grade. Five-year disease-specific survival was 82.8% in whites compared with 70.2% in blacks, 80.7% in Hispanics and 81.9% in Asian/Pacific Islanders. There was a persistent disease-specific survival disadvantage in black patients over time which was not seen in the other ethnic groups.
Ethnic disparities in bladder cancer survival persist between whites and blacks, while survival in other ethnic minority groups appears similar to that of whites. Further study of access to care, quality of care and treatment decision making among black patients is needed to better understand these disparities.
ethnic groups; healthcare disparities; survival; urinary bladder; urinary bladder neoplasms
The number of radical prostatectomies has increased. Many urologists have shifted from the open surgical approach to minimally invasive techniques. It is not clear whether the risk of post-prostatectomy incisional hernia varies by surgical approach.
Materials and Methods
In the linked Surveillance, Epidemiology, and End Results (SEER)-Medicare dataset we identified men age 66 and older who had minimally invasive (MIRP) or open radical prostatectomy (ORP) for prostate cancer diagnosed 2003–2007. The main outcome was incisional hernia repair identified in Medicare claims following prostatectomy. We also examined the frequency of umbilical, inguinal and other hernia repairs.
We identified 3,199 patients who had MIRP and 6,795 who had open radical prostatectomy ORP. The frequency of incisional hernia repair was 5.3% (median follow-up 3.1 years) in the MIRP group and 1.9% (median follow-up 4.4 years) in the ORP group, corresponding to incidence rates of 16.1 and 4.5 per 1000 person-years for MIRP and ORP, respectively. Compared with ORP, MIRP was associated with a more than 3-fold increased risk of incisional hernia repair, controlling for patient and disease characteristics (adjusted hazard ratio 3.39, 95% CI, 2.63–4.38, p <0.0001). MIRP was associated with an attenuated but increased risk of any hernia repair compared with ORP (adjusted hazard ratio 1.48, 95% CI 1.29–1.70, p <0.0001).
MIRP was associated with a significantly increased risk of incisional hernia compared with ORP. This is a potentially remediable complication of prostate cancer surgery that warrants increased vigilance with respect to surgical technique.
prostatic neoplasms; prostatectomy; inguinal hernia; incisional hernia; umbilical hernia
To compare characteristics and outcomes of breast cancer in women with and without a history of radiation therapy (RT) for Hodgkin's lymphoma (HL).
Patients and Methods
Women with breast cancer diagnosed from 1980 to 2006 after RT for HL were identified from eight North American hospitals and were matched three-to-one with patients with sporadic breast cancer by age, race, and year of breast cancer diagnosis. Information on patient, tumor and treatment characteristics, and clinical outcomes was abstracted from medical records.
A total of 253 patients with breast cancer with a history of RT for HL were matched with 741 patients with sporadic breast cancer. Median time from HL to breast cancer diagnosis was 18 years. Median age at breast cancer diagnosis was 42 years. Breast cancer after RT for HL was more likely to be detected by screening, was more likely to be diagnosed at an earlier stage, and was more likely to be bilateral at diagnosis. HL survivors had an increased risk of metachronous contralateral breast cancer (adjusted hazard ratio [HR], 4.3; 95% CI, 1.7 to 11.0) and death as a result of any cause (adjusted HR, 1.9; 95% CI, 1.1 to 3.3). Breast cancer–specific mortality was also elevated, but this difference was not statistically significant (adjusted HR, 1.6; 95% CI, 0.7 to 3.4).
In women with a history of RT for HL, breast cancer is diagnosed at an earlier stage, but these women are at greater risk for bilateral disease and are more likely to die as a result of causes other than breast cancer. Our findings support close follow-up for contralateral tumors in these patients and ongoing primary care to manage comorbid conditions.
Oropharyngeal cancers are increasingly associated with human papillomavirus (HPV). Little is known about the experience of patients receiving this diagnosis.
Semi-structured interviews were conducted with 10 survivors of HPV-related oropharynx cancer. The interviews were transcribed and recurring themes were identified.
Physicians were a trusted source of information regarding HPV. Framing the diagnosis in terms of prognosis resonated with patients. The uncertainty about transmission, latency, and communicability colored the dialogue about HPV. Despite some understanding of prevalence and transmission, patients worried about their partner’s risk. Patients sought information about HPV on the internet, but it was not easily navigable. Emotional reactions to the diagnosis remained mostly cancer-centric rather than HPV-centric. A patient education handout was developed in response to patient questions.
Additional educational resources explaining the facts about HPV in HNSCC in a consistent way including content of highest priority to patients may improve understanding of HPV.
Human papillomavirus (HPV); oropharynx; patient education; quality of life; patient communication
Scientific advances have improved our ability to target cancer interventions to individuals who will benefit most, and spare the risks and costs to those who will derive little benefit or even be harmed. Several approaches are currently used for targeting interventions for cancer risk reduction, screening and treatment, including risk prediction algorithms for identifying high-risk subgroups and diagnostic tests for tumor markers and germline genetic mutations. Economic evaluation can inform decisions about the use of targeted interventions, which may be more costly than traditional strategies. However, assessing the impact of a targeted intervention on costs and health outcomes requires explicit consideration of the method of targeting. Here we describe the importance of this principle by reviewing published cost-effectiveness analyses (CEAs) of targeted interventions in breast cancer. Few studies we identified explicitly evaluated the relationship between the method of targeting, the accuracy of the targeting test and outcomes of the targeted intervention. Those that did found that characteristics of targeting tests had a substantial impact on outcomes. We posit that the method of targeting and the outcomes of a targeted intervention are inextricably linked and recommend that CEAs of targeted interventions explicitly consider costs and outcomes of the method of targeting.
breast cancer; economic analysis; cost-effectiveness analysis; targeted therapy; personalized medicine; BRCA; trastuzumab; gene expression profiling
To assess the impact of mammography capacity on appointment wait times.
We surveyed by telephone all mammography facilities federally certified in 2008 in California, Connecticut, Georgia, Iowa, New Mexico and New York using a simulated-patient format. County-level mammography capacity, defined as the number of mammography machines per 10,000 women age 40 and older, was estimated from FDA facility certification records and US Census data.
1,614 (86%) of 1,882 mammography facilities completed the survey. Time until next available screening mammogram appointment was <1 week at 55% of facilities, 1-4 weeks at 34% of facilities, and >1 month at 11% of facilities. Facilities in counties with lower capacity had longer wait times, and a one-unit increase in county capacity was associated with 21% lower odds of a facility reporting a wait time >1 month (p<0.01). There was no association between wait time and the availability of evening or weekend appointments or digital mammography. Conclusion: Lower mammography capacity is associated with longer wait times for screening mammograms.
Enhancement of mammography resources in areas with limited capacity may reduce wait times for screening mammogram appointments, thereby increasing access to services and rates of breast cancer screening.
Androgen deprivation therapy (ADT) improves prostate cancer outcomes in specific clinical settings, but is associated with adverse effects, including cardiac complications and possibly thromboembolic complications. Our objective was to estimate the impact of ADT on thromboembolic events (TEs) in a population-based cohort.
In the linked Surveillance, Epidemiology and End Results (SEER)-Medicare database, we identified men aged over 65 diagnosed with non-metastatic prostate cancer between 1999 and 2005. Medical or surgical ADT was identified by Medicare claims for gonadotropin-releasing hormone agonists or bilateral orchiectomy at any time following diagnosis. TEs included deep venous thrombosis, pulmonary embolism, and arterial embolism. We estimated ADT’s impact on the risk of any TE and on total number of events, controlling for patient and tumor characteristics.
Of 154,611 prostate cancer patients, 58,466 (38%) received ADT. During a median follow-up of 52 months, 15,950 men had at least one TE, including 8,829 (55%) who had ADT and 7,121 (45%) with no ADT. ADT was associated with increased risk of a TE (adjusted hazard ratio 1.56, 95% CI: 1.50 to 1.61, P < 0.0001), and duration of ADT was associated with the total number of events (P < 0.0001).
In this population-based cohort, ADT was associated with increased risk of a TE, and longer durations of ADT were associated with more TEs. Men with intermediate- and low-risk prostate cancer should be assessed for TE risk factors before starting ADT and counseled regarding the risks and benefits of this therapy.
prostate cancer; deep vein thrombosis; pulmonary embolism; androgen deprivation; arterial embolism
Medicare managed care may offer enrollees lower out-of-pocket costs and provide benefits that are not available in the traditional fee-for-service Medicare program. However, managed care plans may also restrict provider choice in an effort to control costs. We compared rates of voluntary disenrollment from Medicare managed care to traditional fee-for-service Medicare among Medicare managed care enrollees with and without a cancer diagnosis.
We identified Medicare managed care enrollees aged 65 years or older who were diagnosed with a first primary breast (n = 28 331), colorectal (n = 26 494), prostate (n = 29 046), or lung (n = 31 243) cancer from January 1, 1995, through December 31, 2002, in Surveillance, Epidemiology, and End Results (SEER) cancer registry records linked with Medicare enrollment files. Cancer patients were pair-matched to cancer-free enrollees by age, sex, race, and geographic location. We estimated rates of voluntary disenrollment to fee-for-service Medicare in the 2 years after each cancer patient’s diagnosis, adjusted for plan characteristics and Medicare managed care penetration, by use of Cox proportional hazards regression.
In the 2 years after diagnosis, cancer patients were less likely to disenroll from Medicare managed care than their matched cancer-free peers (for breast cancer, adjusted hazard ratio [HR] for disenrollment = 0.78, 95% confidence interval [CI] = 0.74 to 0.82; for colorectal cancer, HR = 0.84, 95% CI = 0.80 to 0.88; for prostate cancer, HR = 0.86, 95% CI = 0.82 to 0.90; and for lung cancer, HR = 0.81, 95% CI = 0.76 to 0.86). Results were consistent across strata of age, sex, race, SEER registry, and cancer stage.
A new cancer diagnosis between 1995 and 2002 did not precipitate voluntary disenrollment from Medicare managed care to traditional fee-for-service Medicare.
To identify treatment patterns and predictors of receiving multimodality therapy in patients with locally advanced prostate cancer (LAPC).
PATIENTS AND METHODS
The cohort comprised patients ≥ 66 years with clinical stage T3 or T4 non-metastatic prostate cancer diagnosed between 1998 and 2005 identified from the Surveillance, Epidemiology and End Results (SEER) cancer registry records linked with Medicare claims.
Treatments were classified as radical prostatectomy (RP), radiation therapy (RT) and androgen deprivation therapy (ADT) received within 6 and 24 months of diagnosis.
We assessed trends over time and used multivariable logistic regression to identify predictors of multimodality treatment.
Within the first 6 months of diagnosis, 1060 of 3095 patients (34%) were treated with a combination of RT and ADT, 1486 (48%) received monotherapy (RT alone, ADT alone or RP alone), and 461 (15%) received no active treatment.
The proportion of patients who received RP increased, exceeding 10% in 2005 .
Use of combined RT and ADT and use of ADT alone fluctuated throughout the study period.
In all 6% of patients received RT alone in 2005.
Multimodality therapy was less common in patients who were older, African American, unmarried, who lived in the south, and who had co-morbidities or stage T4 disease.
Treatment of LAPC varies widely, and treatment patterns shifted during the study period.
The slightly increased use of multimodality therapy since 2003 is encouraging, but further work is needed to increase combination therapy in appropriate patients and to define the role of RP.
prostate cancer; locally advanced; treatment; SEER; practice patterns
As gene expression profile (GEP) testing for breast cancer may provide additional prognostic information to guide the use of adjuvant chemotherapy, we examined the association between GEP testing and use of chemotherapy, serious chemotherapy-related adverse effects, and total charges during the 12 months following diagnosis.
Medical record review was conducted for women age 30 to 64 years, with incident, non-metastatic, invasive breast cancer diagnosed 2006–2008 in a large, national health plan.
Of 534 patients, 25.8% received GEP testing, 68.2% received chemotherapy, and 10.5% experienced a serious chemotherapy-related adverse effect. GEP testing was most commonly used in women at moderate clinical risk of recurrence (52.0% vs. 25.0% of low-risk women and 5.5% of high-risk). Controlling for the propensity to receive GEP testing, women who had GEP were less likely to receive chemotherapy (propensity adjusted odds ratio, 95% confidence interval 0.62, 0.39 – 0.99). Use of GEP was associated with more chemotherapy use among women at low risk based on clinical characteristics (OR = 42.19; CI 2.50 – 711.82), but less use among women with a high risk based on clinical characteristics (OR = 0.12 CI 0.03 – 0.47). Use of GEP was not associated with chemotherapy for the moderate risk group. There was no significant relationship between GEP use and either serious chemotherapy-associated adverse effects or total charges.
While GEP testing was associated with an overall decrease in adjuvant chemotherapy, we did not find differences in serious chemotherapy-associated adverse events or charges during the 12 months following diagnosis.
breast cancer; utilization; genomics
Women with localized breast cancer face difficult decisions about adjuvant therapy. Several decision aids are available to help women choose between treatment options. Decision aids are known to affect treatment choices and may therefore affect patient survival. The authors aimed to model the effects of the Adjuvant! decision aid on expected survival in women with early stage breast cancer.
Patients and Methods
Data were obtained from a randomized trial of Adjuvant! (n =395). To calculate the effects of the decision aid on survival, the authors used the Adjuvant! survival predictions as a surrogate endpoint. Data from each arm were entered separately into statistical models to estimate change in survival associated with receiving the Adjuvant! decision aid.
Most women (~85%) chose a treatment option that maximized predicted survival. The effects of the decision aid on outcome could not be modeled because a small number of women (n =12, 3%) chose treatment options associated with a large (5%–14%) loss in survival. These women—most typically estrogen receptor positive but refusing hormonal therapy—were equally divided between Adjuvant! and control groups and were not distinguished by medical or demographic factors.
Expected benefit from treatment is a key variable in understanding patient behavior. A small number of women refuse adjuvant treatment associated with large increases in predicted survival, even when they are explicitly informed about the degree of benefit they would forgo. Investigation of the effects of decision aids on cancer survival is unlikely to be fruitful due to power considerations.
Adjuvant!; breast cancer; decision aids; women’s health; oncology; outcomes research
Evidence suggests that minimally-invasive radical prostatectomy (MRP) and open radical prostatectomy (ORP) have similar short-term clinical and functional outcomes. MRP with robotic assistance is generally more expensive than ORP, but it is not clear whether subsequent costs of care vary by approach.
In the linked SEER-Medicare database we identified men age 66 or older who received MRP or ORP in 2003-2006 for prostate cancer. Total cost of care was estimated as the sum of Medicare payments from all claims for hospital care, outpatient care, physician services, home health and hospice care, and durable medical equipment in the first year from date of surgical admission. We estimated the impact of surgical approach on costs controlling for patient and disease characteristics.
Of 5,445 surgically-treated prostate cancer patients, 4,454 (82%) had ORP and 991 (18%) had MRP. Mean total first-year costs were more than $1,200 greater for MRP compared with ORP ($16,919 vs. $15692, p=0.08). Controlling for patient and disease characteristics, MRP was associated with 2% greater mean total payments, but this difference was not statistically significant. First-year costs were greater for men who were older, black, lived in the Northeast, had lymph node involvement, more advanced tumor stage or greater comorbidity.
In this population-based cohort of older men, MRP and ORP had similar economic outcomes. From a payer’s perspective, any benefits associated with MRP may not translate to net savings compared with ORP in the first year after surgery.
prostate; prostatic neoplasms; prostatectomy; minimally invasive surgical procedures; laparoscopy
Pelvic lymph node dissection (PLND) is an important component of prostate cancer staging and treatment, especially for surgical patients with high-risk tumor features. It is not clear how the shift from open radical prostatectomy (ORP) to minimally invasive radical prostatectomy (MIRP) has affected use of PLND. Our objective was to identify predictors of PLND and assess the impact of surgical technique in a contemporary, population-based cohort.
In Surveillance, Epidemiology, and End Results (SEER) cancer registry data linked with Medicare claims, we identified men who had ORP or MIRP for prostate cancer in 2003–2007. We evaluated the impact of surgical approach on PLND and examined interactions between surgical procedure, PSA and Gleason score, controlling for patient and tumor characteristics.
Of 6,608 men who had ORP or MIRP, 70% (n=4,600) had PLND. Use of PLND declined over time, overall and within subgroups defined by procedure type. PLND was 5 times more likely in men receiving ORP than MIRP, controlling for patient and tumor characteristics. Elevated PSA and biopsy Gleason score, but not clinical stage, were associated with greater odds of PLND in both ORP and MIRP groups. However, the magnitude of the association between these factors and PLND was significantly greater for ORP patients.
PLND was less common in men who received MIRP, independent of tumor risk factors. A decline in PLND rates was not fully explained by an increase in MIRP. These trends may signal a surgical approach-dependent disparity in prostate cancer staging and therapy.
Prostate Cancer; Lymph Nodes; Lymphadenectomy; SEER-Medicare; Minimally Invasive Surgery; Radical Prostatectomy (Open Robotic, Laparoscopic)
Despite almost universal testing for HER2, many women with HER2-positive cancer may not receive trastuzumab. Fewer women received newer gene expression profile testing.
Given the likely proliferation of targeted testing and treatment strategies for cancer, a better understanding of the utilization patterns of human epidermal growth factor receptor 2 (HER2) testing and trastuzumab and newer gene expression profiling (GEP) for risk stratification and chemotherapy decision making are important.
We performed a medical record review of women age 35 to 65 years diagnosed between 2006 and 2007 with invasive localized breast cancer, identified using claims from a large national health plan (N = 775).
Almost all women received HER2 testing (96.9%), and 24.9% of women with an accepted indication received GEP. Unexplained socioeconomic differences in GEP use were apparent after adjusting for age and clinical characteristics; specifically, GEP use increased with income. For example, those in the lowest income category (< $40,000) were less likely than those with an income of $125,000 or more to receive GEP (odds ratio, 0.34; 95% CI, 0.16 to 0.73). A majority of women (57.7%) with HER2-positive disease received trastuzumab; among these women, differences in age and clinical characteristics were not apparent, although surprisingly, those in the lowest income category were more likely than those in the high-income category to receive trastuzumab (P = .02). Among women who did not have a positive HER2 test, 3.9% still received trastuzumab. Receipt of adjuvant chemotherapy increased as GEP score indicated greater risk of recurrence.
Identifying and eliminating unnecessary variation in the use of these expensive tests and treatments should be part of quality improvement and efficiency programs.
Testing technologies are increasingly used to target cancer therapies. Human epidermal growth factor receptor 2 (HER2) testing to target trastuzumab for patients with breast cancer provides insights into the evidence needed for emerging testing technologies.
We reviewed literature on HER2 test utilization and cost-effectiveness of HER2 testing for patients with breast cancer. We examined available evidence on: percentage of eligible patients tested for HER2; test methods used; concordance of test results between community and central/reference laboratories; use of trastuzumab by HER2 test result; and cost-effectiveness of testing strategies.
Little evidence is available to determine whether all eligible patients are tested; how many are retested to confirm results; and how many with negative HER2 test results still receive trastuzumab. Studies suggest that up to 66% of eligible patients had no documentation of testing in claims records; up to 20% of patients receiving trastuzumab were not tested or had no documentation of a positive test; and 20% of HER2 results may be incorrect. Few cost-effectiveness analyses of trastuzumab explicitly considered the economic implications of various testing strategies.
There is little information about the actual use of HER2 testing in clinical practice, but evidence suggests important variations in testing practices and key gaps in knowledge exist. Given the increasing use of targeted therapies, it is critical to build an evidence base that supports informed decision-making on emerging testing technologies in cancer care.
personalized medicine; targeted therapies; genomics; HER2; trastuzumab; breast cancer; utilization; cost-effectiveness; clinical practice patterns
Enthusiasm for laparoscopic surgical approaches to prostate cancer treatment has grown, despite limited evidence of improved outcomes compared with open radical prostatectomy. We compared laparoscopic (with or without robotic assistance) versus open radical prostatectomy in terms of postoperative outcomes and subsequent cancer-directed therapy.
Materials and Methods
Using a population-based cancer registry linked with Medicare claims, we identified men age 66 or older with localized prostate cancer who received a radical prostatectomy from 2003-2005. Outcome measures were general medical/surgical complications and mortality within 90 days following surgery; genitourinary/bowel complications within 365 days; receipt of radiation therapy, androgen deprivation therapy or both within 365 days; length of hospital stay.
Of the 5,923 men,18% received a laparoscopic radical prostatectomy. Adjusting for patient and tumor characteristics, there were no differences in rates of general medical/surgical complications (OR 0.93; 95% CI: 0.77-1.14) or genitourinary/bowel complications (OR 0.96; 95% CI: 0.76-1.22) or in the use of postoperative radiation, androgen deprivation or both (OR 0.80; 95% CI: 0.60-1.08). Laparoscopic prostatectomy was associated with a 35% shorter hospital stay (p<0.0001) and a lower rate of bladder neck/urethral obstruction (OR 0.74; 95% CI 0.58-0.94). In laparoscopic patients, surgeon volume was inversely associated with length of hospital stay and the odds of any genitourinary/bowel complication.
Laparoscopic and open radical prostatectomy have similar rates of postoperative morbidity and use of additional treatment. Men considering prostate cancer surgery should understand the expected benefits and risks of each technique to facilitate decision-making and to set realistic expectations.
prostate; prostatic neoplasms; prostatectomy; minimally invasive surgical procedures; laparoscopy
Radical nephrectomy (RN), compared with partial nephrectomy (PN), increases the risk of chronic kidney disease, a significant risk factor for cardiovascular (CV) events and death. Given equivalent oncologic efficacy in patients with small renal tumors (RTs), RN may result in overtreatment. We analyzed a population-based cohort of patients to determine if RN is associated with an increase in CV events and mortality compared with PN.
Materials and Methods
Using Surveillance, Epidemiology, and End Results (SEER) cancer registry data linked with Medicare claims, we identified 2991 patients older than 65 years of age treated with RN or PN for RTs 4 cm or smaller between 1995 and 2002. Primary end points of CV events and overall survival were assessed using Kaplan-Meier survival estimation, Cox proportional hazards regression, and negative binomial regression.
A total of 2547 (81%) patients underwent RN and 556 (19%) underwent PN. During a median follow-up of 4 years, 609 patients had a CV event and 892 patients died. Adjusting for preoperative demographic and comorbid variables, RN was associated with an increased risk of overall mortality (hazard ratio [HR] 1.38, P<0.01) and a 1.4 times greater number of CV events after surgery (P<0.05). RN, however, was not associated with an increased risk of time to first CV event (HR 1.21, P=0.10) or CV death (HR 0.95, P=0.84).
RN, currently the most common treatment for small RTs, may be associated with significant, adverse treatment effects compared with PN. PN should be considered for most patients with small RTs.
Radical Nephrectomy; Partial Nephrectomy; Mortality; Cardiovascular Events
Decision curve analysis is a novel method for evaluating diagnostic tests, prediction models and molecular markers. It combines the mathematical simplicity of accuracy measures, such as sensitivity and specificity, with the clinical applicability of decision analytic approaches. Most critically, decision curve analysis can be applied directly to a data set, and does not require the sort of external data on costs, benefits and preferences typically required by traditional decision analytic techniques.
In this paper we present several extensions to decision curve analysis including correction for overfit, confidence intervals, application to censored data (including competing risk) and calculation of decision curves directly from predicted probabilities. All of these extensions are based on straightforward methods that have previously been described in the literature for application to analogous statistical techniques.
Simulation studies showed that repeated 10-fold crossvalidation provided the best method for correcting a decision curve for overfit. The method for applying decision curves to censored data had little bias and coverage was excellent; for competing risk, decision curves were appropriately affected by the incidence of the competing risk and the association between the competing risk and the predictor of interest. Calculation of decision curves directly from predicted probabilities led to a smoothing of the decision curve.
Decision curve analysis can be easily extended to many of the applications common to performance measures for prediction models. Software to implement decision curve analysis is provided.
Diagnostic and prognostic models are typically evaluated with measures of accuracy that do not address clinical consequences. Decision-analytic techniques allow assessment of clinical outcomes, but often require collection of additional information, and may be cumbersome to apply to models that yield a continuous result. We sought a method for evaluating and comparing prediction models that incorporates clinical consequences, requires only the dataset on which the models are tested, and can be applied to models that have either continuous or dichotomous results.
We describe decision curve analysis, a simple, novel method of evaluating predictive models. We start by assuming that the threshold probability of a disease or event at which a patient would opt for treatment is informative of how the patient weighs the relative harms of a false-positive and a false-negative prediction. This theoretical relationship is then used to derive the net benefit of the model across different threshold probabilities. Plotting net benefit against threshold probability yields the “decision curve”. We apply the method to models for the prediction of seminal vesicle invasion in prostate cancer patients. Decision curve analysis identified the range of threshold probabilities in which a model was of value, the magnitude of benefit, and which of several models was optimal.
Decision curve analysis is a suitable method for evaluating alternative diagnostic and prognostic strategies that has advantages over other commonly used measures and techniques.
prediction models; multivariate analysis; decision analysis
Robotic surgery to remove a cancerous prostate has become a popular treatment. Internet marketing of this surgery provides an intriguing case study of direct-to-consumer promotions of medical devices, which are more loosely regulated than pharmaceutical promotions. We investigated whether the claims made in online promotions of robotic prostatectomy were consistent with evidence from comparative effectiveness studies. After performing a search and crosssectional analysis of websites that mentioned the procedure, we found that many sites claimed benefits that were unsupported by evidence and that 42 percent of the sites failed to mention risks. Most sites were published by hospitals and physicians, which the public may regard as more objective than pages published by manufacturers. Unbalanced information may inappropriately raise patients’ expectations. Increasing enforcement and regulation of online promotions may be beyond the capabilities of federal authorities. Thus, the most feasible solution may be for the government and medical societies to promote the production of balanced educational material.