Use of Monitoring Tests Among Patients with Localized Prostate Cancer Managed with Observation

Abstract

Introduction:

It is unknown whether compliance with recommended monitoring tests during observation of localized prostate cancer has changed over time.

Methods:

We performed a retrospective cohort study of Medicare beneficiaries diagnosed with low- or intermediate-risk prostate cancer in 2004–2016 who were initially managed with observation for a minimum of 12 months. The primary objective was to examine rates of PSA testing, prostate biopsy, and prostate MRI. We used multivariable mixed effects Poisson regression to determine whether rates of PSA testing and prostate biopsy increased over time. In addition, we identified clinical, sociodemographic, and provider factors associated with the frequency of monitoring tests during observation.

Results:

We identified 10,639 patients diagnosed at a median age of 73 (interquartile range [IQR] 69–77) years. The median follow-up time was 4.3 (IQR 2.7–6.6) years after diagnosis. Among patients managed without treatment for five years, 98% received at ≥1 PSA test, 48.0% ≥1 additional prostate biopsy and 31.0% ≥1 prostate MRI. Among patients managed with observation for ≥12 months, mixed effects Poisson regression revealed that rates of PSA testing and biopsy increased over time (per calendar year: relative risk (RR) 1.02, 95% CI: 1.02–1.03 and RR 1.10, 95% CI: 1.08–1.11, respectively). Clinical and sociodemographic factors including age, clinical risk, race/ethnicity, census tract poverty, and region were associated with rates of biopsy and PSA testing.

Conclusions:

Use of recommended monitoring tests including repeat prostate biopsy remains low among Medicare beneficiaries undergoing observation for low and intermediate-risk prostate cancer.

Introduction

Regular monitoring is a definitional component of active surveillance for prostate cancer.¹ Practice guidelines recommend serial prostate-specific antigen (PSA) measurement, physical exam and prostate biopsy; and support the use of imaging (ultrasound or magnetic resonance imaging, MRI) during active surveillance.^{2, 3} The need for monitoring is informed by the potential for initial misclassification or disease progression over time, events that occur in up to half of patients enrolled in active surveillance.^4–6 Earlier identification of disease reclassification, particularly through confirmatory biopsy, may improve the long-term safety of the approach. For example, higher risks of metastasis among patients who received active monitoring based on PSA only in the ProtecT trial (10-year incidence 5.6%) than in major institutional series (e.g. 0.6% 10-year probability in the Memorial Sloan Kettering cohort) may be related to less intensive monitoring and infrequent use of repeat biopsy.^{4, 7–9} As a result, close monitoring is a hallmark of high quality active surveillance.¹⁰

The majority of patients with low-risk prostate cancer in the United States are now initially managed with active surveillance.¹¹ However, little is known about the compliance with monitoring guidelines in contemporary practice.¹² Observational studies conducted in the early era of active surveillance have shown low compliance with recommended monitoring protocols.¹³ For example, Loeb et al. previously demonstrated low use of prostate biopsy and PSA testing among Medicare beneficiaries managed expectantly through 2009, with less than 13% undergoing repeat biopsies beyond the third year.¹⁴ With growing acceptance of active surveillance, it is unclear whether compliance with recommended monitoring tests has meaningfully changed. Assessing compliance in the contemporary period is needed as active surveillance is increasingly offered in a diverse range of practice settings and clinical contexts, including to patients with intermediate-risk features, a subset that faces higher risks of disease progression.¹⁵

To improve the quality and safety of active surveillance, we evaluated the compliance to monitoring tests among patients managed with observation for prostate cancer. Using a population-based sample of Medicare beneficiaries, we explored temporal changes in the use of PSA testing, prostate biopsy, and prostate MRI. There was a precipitous increase in the use of active surveillance in the United States beginning in 2010.¹¹ Therefore, we sought to compare practice in the periods before and after widespread uptake.¹⁶ As clinical management practices for low-risk prostate cancer vary substantially at the provider level and across sociodemographic factors, we further explored whether these factors contributed to the frequency of testing.¹⁷

Methods

Study Design and Cohort Identification

The objective of this study was to determine whether the use of PSA testing and prostate biopsy among patients initially managed with observation for prostate cancer has increased over time. The secondary objectives were to identify patient and provider factors associated with the use of monitoring tests among patients diagnosed with low and intermediate risk prostate cancer in 2004–2016 who received ≥12 months of observation.

We conducted a retrospective cohort study using the Surveillance, Epidemiology and End-Results (SEER) records linked with individual level Medicare claims (SEER-Medicare).¹⁸ We identified patients with localized prostate adenocarcinoma diagnosed between Jan 1, 2004 and December 31, 2016. We selected patients with low (PSA<10 ng/mL, and Gleason grade group 1 and clinical stage ≤T2a) or intermediate-risk (PSA 10–20 ng/mL or Gleason grade group 2, or clinical stage ≤T2c) prostate cancer who were ≥66 years old at the time of diagnosis and who had no evidence of treatment in the 12 months following diagnosis (Figure 1). Participants were regarded as initiating observation at the time of their first diagnostic prostate biopsy demonstrating prostate cancer.

Figure 1.

Flow diagram showing study selection

Study Variables

We identified monitoring tests by individual claims for PSA, prostate biopsy, and prostate MRI using common procedural terminology codes and assessed treatment through claims in the follow-up period. We compiled patient sociodemographic and clinical characteristics and assigned each patient to a single urologist, based on the greatest number of outpatient visits in the year of diagnosis or as the clinician who performed the first prostate biopsy in the diagnosis period.^{19, 20} We then calculated the number of Medicare patients with prostate cancer that their urologist treated in the 12-month period prior to a given patient’s diagnosis.

The primary study outcome was the count of prostate biopsies and PSA tests during a patient’s observation period. The secondary outcomes were compliance with specific surveillance protocols. These included receipt of ≥10 PSA tests and ≥2 prostate biopsies, adherence to the Prostate Cancer Research International Active Surveillance (PRIAS) study protocol (≥14 PSA tests and ≥2 biopsies), as well as the Johns Hopkins protocol (≥10 PSA tests and≥4 biopsies) per Loeb et al.^{14 21} A confirmatory biopsy was defined as a biopsy occurring subsequent to a first diagnostic biopsy. The other secondary outcomes included the rate of biopsy and PSA testing and the proportion of patients receiving a prostate MRI study during observation.

Statistical Analysis

We described the baseline sociodemographic, clinical, and pathologic characteristics of eligible patients and the distribution of provider volume in the study sample. We calculated the proportion of patients receiving PSA tests, prostate biopsy, and prostate MRI after initial prostate cancer diagnosis. Among the subset of patients who completed ≥5 years of observation without treatment, we tabulated the count of prostate biopsy and PSA tests obtained over five years. We compared the distribution of patients receiving prostate biopsy, PSA tests, and prostate MRI in the early and late diagnosis periods using Wilcoxon tests for continuous variables and Pearson’s χ2 test for categorical variables. We compared the proportion of patients adherent to specific surveillance protocols. In addition, we calculated rates of PSA testing and prostate biopsy (per person-year) and rate ratios between early and late periods. Analyses were conducted in the overall sample and stratified by clinical risk status.

Among patients diagnosed in 2004–2016 with a minimum of 12 months of observation, we used mixed-effects Poisson regression models with robust standard errors and a logarithm link function to evaluate factors associated with use of PSA tests and prostate biopsy, expressed as relative risks (RR). The models included an exposure time offset and accounted for clustering of patients within individual physicians as a random effect. Patients were censored from analysis at the time of definitive treatment. Lastly, we used interrupted time series analyses to determine whether a significant period effect was identified among patients diagnosed in 2010–2016 versus 2004–2009. For these analyses we fit mixed-effects Poisson regression models including the year of diagnosis as a continuous variable, years after 2010 as a binary indicator of diagnosis period (i.e. 2004–2009 versus 2010–2016).

The Yale Human Investigations Committee determined that this study did not involve human subjects and did not require institutional review board approval. The statistical analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC). All tests were two-sided and a p-value threshold of <0.05 was used to define statistical significance.

Results

The final study cohort included 10,639 patients meeting eligibility criteria (Figure 1). The characteristics of the study participants are presented in Table 1. Patients with intermediate risk disease were older (p<0.001) and had a greater burden of comorbidity (p<0.001) than those with low-risk prostate cancer. The median follow-up time was 4.3 (interquartile range [IQR] 2.7–6.6) years after prostate cancer diagnosis. The median age at diagnosis was 73 (IQR 69–77) years. The cumulative incidence of definitive treatment was 27% and 37% at 3 years and 5 years post-diagnosis, respectively (Figure 2). Three-year (29% versus 25%, p <0.001) and 5-year (39% versus 34%, p <0.001) incidences of definitive treatment were higher for patients with intermediate-risk disease compared to those with low-risk disease, respectively. The eventual treatments included radiation therapy in 23%, radical prostatectomy in 7.9%, and androgen-deprivation therapy alone in 11% of patients.

Table 1.

Demographic, clinical, and provider characteristics of patients with low- and intermediate-risk prostate cancer initially managed with observation.

	Diagnosis Period 2004–2009			Diagnosis Period 2010–2014			P for χ2
	n	%		n	%
Total	2156			8483
Risk group
Low	334	15		4505	53		<0.001
Intermediate	1822	85		3978	47
Age at diagnosis (years)
66–69	359	17		2592	31		<0.001
70–74	556	26		2884	34
75–79	558	26		1912	23
80–84	490	23		797	9.4
≥85	193	9.0		298	3.5
Race
Non-Hispanic white	1708	79		6895	81		<0.001
Black	236	11		759	8.9
Hispanic	115	5.3		377	4.4
Asian	71	3.3		225	2.7
Other	26	1.2		227	2.7
Marital status
Married	1209	56		5203	61		<0.001
Unmarried	449	21		1436	17
Unknown	498	23		1844	22
Elixhauser Index
0	1188	55		4607	54		0.80
1–2	767	36		3075	36
≥3	201	9.3		801	9.4
Frail
No	1927	89		7724	91		0.017
Yes	229	11		759	8.9
Flu shot
No	1183	55		4088	48		<0.001
Yes	973	45		4395	52
Medicaid enroll
No	1993	92		7932	94		0.08
Yes	163	7.6		551	6.5
Census tract poverty (%)
0 -<5	630	29		2238	26		0.023
5 -<10	588	27		2382	28
10 -<20	552	26		2382	28
20 −100	386	18		>1470	>17
Unknown	0	0.0		<11	<0.1
SEER region
Northeast	451	21		1686	20		<0.001
Midwest	264	12		954	11
South	455	21		2163	25
West	986	46		3680	43
Metro
Metro	1839	85		7263	86		0.71
Non-metro	317	15		1220	14
Annual provider volume of patients with prostate cancer
0–5	463	21		2491	29		<0.001
6–10	631	29		2846	34
11+	1062	49		3146	37

Note: Percentages may not add up to 100 due to rounding.

Figure 2.

Time to definitive treatment among patients with low- and intermediate-risk prostate cancer initially managed with observation.

Nearly all patients had at least one PSA test (98%) during the follow-up period, including 99% of those with low-risk and 97% with intermediate-risk prostate cancer. The median number of PSA tests received was 7 (IQR 4–12). Repeat biopsy was undertaken in 48% of patients, including 63% of those with low-risk disease and 36% with intermediate-risk disease (Figure 3). Trajectories of patients, including use of PSA, prostate biopsy and definitive treatment are shown in the supplemental figure. Overall, 31% of patients received a prostate MRI scan, including 40% of those with low-risk and 24% among those with intermediate-risk disease.

Figure 3.

Proportion of patients receiving repeat prostate biopsy, PSA tests, and prostate MRI during observation for low- and intermediate-risk prostate cancer.

A total of 4,258 of 10,639 patients (40%) were observed without treatment through five years, including 1,072 (10%) in the early and 3,186 (30%) in late periods (Table 2). The proportion of patients receiving diagnostic monitoring aligned with published protocols was low but increased over time. Among those diagnosed in 2010–2014, 837 (26%) received ≥10 PSA tests and ≥1 additional biopsy following diagnosis, compared with 15.9% for patients diagnosed in 2004–2009 (p<0.001). Compliance with the Johns Hopkins protocol (≥10 PSA tests and ≥4 biopsies) was 2.0% among patients diagnosed in 2004–2009 and 4.3% for those diagnosed in 2010–2014 (p<0.001).

Table 2.

Total number of tests received during 5 years of follow-up for prostate cancer patients managed with observation for a minimum of 5 years, based on diagnosis period.

	2004–2014			Diagnosis Period
				2004–2009			2010–2014
	n	%		n	%		n	%		p
Number of patients	4,258			1,072			3,186
Number of PSA tests
Median (IQR)	9 (6–12)			9 (5–12)			9 (6–12)			0.002
0	135	3.2		48	4.5		87	2.7
1–4	572	13		164	15		408	13
5–9	1507	35		387	36		1120	35
10–19	1923	45		437	41		1486	47
≥20	121	2.8		36	3.4		85	2.7
Number of prostate biopsies
Median (IQR)	1 (1–2)			1 (1–2)			1 (1–2)			<0.001
1 (no repeat biopsy)	2577	61		792	74		1785	56
2	975	23		181	17		794	25
3	476	11		67	6.3		409	13
≥4	230	5.4		32	3.0		198	6.2
Prostate MRI
Median (IQR)	0 (0–1)			0 (0–1)			0 (0–1)			<0.001
0	3,425	80		1,007	94		2,418	76
1	506	12		54	5.0		452	14
2	187	4.4		†	†		>176	>5.5
≥3	140	3.3		†	†		>129	>4.0
Protocol Adherence
≥10 PSA and ≥2 biopsies	1007	24		170	16		837	26		<0.001
PRIAS 8 (≥14 PSA and ≥2 biopsies)	377	8.9		60	5.6		317	9.9		<0.001
Johns Hopkins14 (≥10 PSA tests and ≥4 biopsies)	159	3.7		21	2.0		138	4.3		<0.001

Abbreviations: PSA=prostate specific antigen; IQR=interquartile range; PRIAS=Prostate Cancer Research International Active Surveillance Study;

^†Value suppressed due to cell size <11; CI=confidence interval

Percentages may not add up to 100 due to rounding.

Mixed-effects Poisson regression models identified factors associated with biopsy and PSA testing during observation in the overall cohort (Table 3). Rates of PSA and biopsy increased over time (per calendar year: RR 1.02, 95% CI: 1.02–1.03 and RR 1.10, 95% CI: 1.08–1.11, respectively). However, we did not identify distinct period effects for 2004–2009 versus 2010–2016 for rates of prostate biopsy (p=0.72) or PSA testing (p=0.38).

Table 3.

Results of multivariable mixed effects Poisson regression models demonstrating factors associated with rate of prostate biopsy and PSA testing during prostate cancer active surveillance.

	PSA Tests				Prostate Biopsy
	RR	95% CI	p		RR	95% CI	p
Risk group
Low	1.00				1.00
Intermediate	0.93	0.91 – 0.95	<.001		0.76	0.72 – 0.80	<.001
Age at diagnosis (years)
66–69	1.00				1.00
70–74	0.99	0.96 – 1.02	0.38		0.82	0.78 – 0.87	<.001
75–79	0.99	0.96 – 1.02	0.49		0.50	0.46 – 0.54	<.001
80–84	1.02	0.97 – 1.07	0.40		0.25	0.22 – 0.29	<.001
≥85	0.93	0.87 – 1.00	0.05		0.15	0.11 – 0.19	<.001
Race/Ethnicity
Non-Hispanic white	1.00				1.00
Black	0.86	0.82 – 0.91	<.001		1.00	0.90 – 1.11	0.96
Hispanic	0.87	0.81 – 0.92	<.001		0.85	0.74 – 0.98	0.021
Asian	0.89	0.81 – 0.97	0.01		0.85	0.71 – 1.03	0.10
Other	0.92	0.85 – 1.00	0.06		0.94	0.82 – 1.07	0.35
Year of diagnosis (Versus 2004)	1.02	1.02 – 1.03	<.001		1.10	1.08 – 1.11	<.001
Marital status
Married	1.00				1.00
Unmarried	0.94	0.91 – 0.97	<.001		0.90	0.84 – 0.97	0.008
Unknown	0.98	0.95 – 1.01	0.24		0.92	0.86 – 0.99	0.019
Elixhauser Index
0	1.00				1.00
1–2	1.04	1.01 – 1.06	0.003		0.95	0.90 – 1.01	0.08
≥3	1.00	0.96 – 1.05	0.93		0.89	0.79 – 0.99	0.036
Frail	1.00	0.96 – 1.05	0.98		0.85	0.76 – 0.95	0.004
Flu shot	1.09	1.07 – 1.12	<.001		1.11	1.06 – 1.16	<.001
Medicaid enrollment	0.87	0.82 – 0.93	<.001		0.80	0.70 – 0.91	<.001
Census tract poverty (%)
0 -<5	1.00				1.00
5 -<10	0.99	0.96 – 1.02	0.43		0.95	0.90 – 1.01	0.12
10 -<20	0.95	0.92 – 0.99	0.005		0.94	0.87 – 1.01	0.08
20 −100	0.92	0.88 – 0.96	<.001		0.86	0.78 – 0.94	0.001
Unknown	1.99	0.75 – 5.3	0.17		0.66	0.58 – 0.75	<.001
Non-metro area	0.91	0.88 – 0.94	<.001		0.78	0.70 – 0.86	<.001
Provider Volume
0–5	1.00				1.00
6–10	1.00	0.97 – 1.03	0.95		0.90	0.84 – 0.96	0.001
11+	0.99	0.96 – 1.03	0.67		0.97	0.90 – 1.1	0.48
SEER region
Northeast	1.00				1.00
Midwest	1.01	0.96 – 1.07	0.57		0.79	0.69 – 0.90	<.001
South	1.01	0.97 – 1.05	0.64		0.75	0.68 – 0.83	<.001
West	1.06	1.02 – 1.10	0.004		0.93	0.85 – 1.0	0.07

Abbreviations: PSA=prostate specific antigen; RR=relative risk; CI=confidence interval; SEER=Surveillance, Epidemiology and End Results

At the patient level, Hispanic/Latino versus non-Hispanic/Latino white ethnicity (RR 0.85, 95% confidence interval [CI]: 0.74–0.98) and older age (compared with ages 66–69 years: 70–74 years RR 0.82, 95% CI: 0.78–0.87; 75–79 years RR 0.50, 95% CI: 0.46–0.54; 80–84 years RR 0.25, 95% CI: 0.22–0.29; and ≥85 years, RR 0.15, 95% CI 0.11–0.19) were associated with less use of prostate biopsy. In addition, rates of biopsy were lower among patients who were unmarried (RR 0.90, 95% CI: 0.84–0.97), had greater burden of comorbidity (Elixhauser ≥3 versus 0: RR 0.89, 95% CI: 0.79–0.99), were frail (RR 0.85, 95% CI: 0.76–0.95), Medicaid-eligible (RR 0.80, 95% CI: 0.70–0.91), resided in a census tract with higher level of poverty (20–100% impoverished versus <5%: RR 0.86, 95% CI: 0.78–0.94), lived in a non-metropolitan area (RR 0.78, 95% CI: 0.70–0.86), and lived in the Midwest (RR 0.79, 95% CI: 0.69–0.90) or Southern (RR 0.75, 95% CI: 0.68–0.83) SEER registry regions compared with the Northeast. Provider volume was not associated with biopsy frequency.

Discussion

Among a population-based sample of United States Medicare beneficiaries with low and intermediate-risk prostate cancer managed with observation through 2019, we found persistently low use of recommended monitoring tests. Although rates of prostate biopsy and PSA testing increased over time, they remained less frequent than recommended by most institutional protocols. Notably, patients with intermediate risk disease had lower rates of testing, a subgroup for whom active surveillance is associated with greater risks of distant progression.¹⁵ We also identified sociodemographic disparities that persisted over the study period, underscoring the need to better evaluate and address sources of variation in the quality of observational management. As observational management is now recognized as the standard of care for patients with low-risk prostate cancer and has enjoyed worldwide uptake, these results can inform strategies to enhance the quality of its delivery.

Compliance with PSA testing and repeat biopsy in the contemporary era remained below frequencies recommended by major protocols. Although almost all patients received additional PSA testing, less than half of patients undergoing observation for localized prostate cancer received a confirmatory prostate biopsy during follow-up. Lower rates of biopsy among patients with intermediate-risk disease in this SEER-Medicare sample may be related to intended expectant management, as this group was older and had a higher degree of comorbidity than low-risk patients. Nonetheless, findings of less use of prostate biopsy among patients with intermediate-risk disease were also observed in multivariable models adjusting for age and comorbidity. Prior studies of higher-risk patients enrolled on active surveillance have shown that guideline discordant follow-up was associated with a five-fold greater hazard of metastatic progression.²² As a result, these findings should garner greater attention to the distinct risks in the intermediate-risk population and underscore the need to improve monitoring in this group.

Low rates of PSA testing and prostate biopsy may also be related to the absence of clear guidance about the timing and necessity of testing during observation. Although major institutional protocols examined provide recommended frequencies of PSA testing and biopsy, and have been recognized as stringent compared with empiric clinical practice.²³ In contrast, clinical practice guidelines are less prescriptive.²⁴ For example, the National Comprehensive Cancer Network recommends a confirmatory prostate biopsy within 1–2 years, but otherwise only provides guidance on the maximum frequency of subsequent PSA testing and biopsy. To promote greater consistency, clinical practice guidelines should be more explicit about the preferred frequencies of monitoring tests during active surveillance.

Bolstered by increasing clinical evidence and guideline support, 2009–2010 has been identified as an inflection point in the use of active surveillance versus definitive treatment for low-risk prostate cancer,^{11, 25} motivating the dichotomization used in this study. Although rates of testing increased over time, we did not find a significant period effect in 2010–2016, suggesting that adherence to monitoring tests exhibited increasing secular trends rather than a discrete change in 2010. These findings could reflect the effects of accumulating data, physicians’ familiarity, and patients’ acceptance.

There were notable sociodemographic disparities in the frequency of active surveillance monitoring. Rates of biopsy and PSA testing were lower among patients whose ethnicity was identified as Latino/Hispanic, and there was less frequent PSA testing among patients identified as Black and Asian. In addition, there was less testing among patients residing in higher poverty and non-metropolitan areas. These findings mirror known disparities in the quality of prostate cancer care, such as the use of prostate MRI and initial selection of active surveillance.^{26 27} Our findings add to a large body of evidence on sociodemographic disparities in prostate cancer care by demonstrating their extension to the quality of monitoring itself.²⁸ Lastly, we did not find that provider volume was associated with the frequency of monitoring. These findings may be reassuring and suggest a broad capacity to adopt evidence-based active surveillance, regardless of physician volume.

Awareness of persistently low adherence to recommended monitoring can catalyze advocacy efforts, education and incentivization to enhance guideline-concordant care.² In addition, this study highlights the ongoing need to improve risk stratification, with the aim of personalizing monitoring strategies aligned with cancer risk, preferences, and competing comorbidities. Low compliance with prostate biopsy likely reflects reluctance to undergo invasive testing, which is a barrier to the selection and adherence to active surveillance.²⁹ Therefore, efforts to improve the safety and experience of biopsy, as well as the integration of non-invasive biomarkers to decrease reliance on biopsy are also needed. Lastly, there is a timely opportunity to leverage established implementation science strategies to enhance the uptake and quality of active surveillance, including financial incentivization to better align reimbursement with high value care.³⁰

There are several limitations that frame these results. Although population-based, the study sample is drawn from Medicare beneficiaries and is skewed towards older patients. Therefore, these findings may not reflect surveillance in younger patients that may be more intensive.³¹ Lower compliance among those with intermediate risk disease might suggest intended use of expectant management rather than active surveillance, which could not be distinguished using Medicare claims. Indeed, there were substantial increases in the proportions of patients with low versus intermediate risk disease who were observed between the two time periods which may confound estimates of changes in testing rates over time. To improve future studies, clearer distinctions between “active surveillance” and “watchful waiting” in clinical care, administrative coding, and research are needed. In addition, we did not evaluate the clinical impact of compliance with respect to distant progression or cancer mortality. In addition, this claims-based analysis is subject to potential under-ascertainment of healthcare services not reimbursed by Medicare. Lastly, we did not examine the role of other tools that may modify surveillance, such as genomic testing, or how the results of testing may impact the intensity of surveillance.^{32, 33}

Conclusions

In the era of increased use of active surveillance for localized prostate cancer, use of recommended monitoring strategies among older Americans managed with observation remains low. Despite increases in the rates of PSA testing and prostate biopsy, the majority of patients undergoing observation did not receive a confirmatory prostate biopsy. These findings suggest a need for interventions to improve the tolerability and safety of active surveillance.