Low‐risk prostate lesions: An evidence review to inform discussion on losing the “cancer” label

Abstract Background Active surveillance (AS) mitigates harms from overtreatment of low‐risk prostate lesions. Recalibration of diagnostic thresholds to redefine which prostate lesions are considered “cancer” and/or adopting alternative diagnostic labels could increase AS uptake and continuation. Methods We searched PubMed and EMBASE to October 2021 for evidence on: (1) clinical outcomes of AS, (2) subclinical prostate cancer at autopsy, (3) reproducibility of histopathological diagnosis, and (4) diagnostic drift. Evidence is presented via narrative synthesis. Results AS: one systematic review (13 studies) of men undergoing AS found that prostate cancer‐specific mortality was 0%−6% at 15 years. There was eventual termination of AS and conversion to treatment in 45%−66% of men. Four additional cohort studies reported very low rates of metastasis (0%−2.1%) and prostate cancer‐specific mortality (0%−0.1%) over follow‐up to 15 years. Overall, AS was terminated without medical indication in 1%−9% of men. Subclinical reservoir: 1 systematic review (29 studies) estimated that the subclinical cancer prevalence was 5% at <30 years, and increased nonlinearly to 59% by >79 years. Four additional autopsy studies (mean age: 54−72 years) reported prevalences of 12%−43%. Reproducibility: 1 recent well‐conducted study found high reproducibility for low‐risk prostate cancer diagnosis, but this was more variable in 7 other studies. Diagnostic drift: 4 studies provided consistent evidence of diagnostic drift, with the most recent (published 2020) reporting that 66% of cases were upgraded and 3% were downgraded when using contemporary diagnostic criteria compared to original diagnoses (1985−1995). Conclusions Evidence collated may inform discussion of diagnostic changes for low‐risk prostate lesions.


| INTRODUCTION
In higher income countries including the USA, 1 Australia, 2

and
Canada, 3 a dramatic increase in the incidence of prostate cancer diagnoses occurred in the late 1980s and 1990s with the widespread uptake of prostate-specific antigen (PSA) testing. 4 Much of the increase was driven by the detection of low-risk prostate cancers with Gleason scores (GS) ≤6 on histopathology. 5 Subsequently, prostate cancer incidence has remained high. Although smaller in absolute terms, there have also been decreases in rates of first diagnosis at an advanced stage, and in prostate cancer mortality. 4,6 By identifying aggressive lesions at an early stage, PSA screening may have prevented some men from morbidity associated with advanced stage cancer (and its treatment), and some from dying of prostate cancer. 6 Against these potential benefits, harms from PSA screening include false-positive and false-negative results, biopsy complications (e.g., infection following transrectal biopsy, erectile dysfunction following transperineal biopsy), overdiagnosis, 7 and overtreatment 8-10 with risk of adverse outcomes (e.g., postoperative complications, erectile dysfunction, and urinary incontinence). 11,12 To prevent harms from overtreatment, active surveillance has emerged as a preferred management option for low-risk lesions. 13 The use of active surveillance for low-risk lesions has increased in recent years, but there remains substantial variation in uptake between countries [14][15][16][17] (ranging from 50% in United States of America 14 to >90% in England and Wales 17 ), and within countries, with lower uptake noted among minoritized racial groups 18 and those outside metropolitan areas. 15,18 Clinicians may not recommend active surveillance to patients because of a patient's clinical and personal characteristics (e.g., younger age), and perceptions of: patient disinterest, inadequacy of biopsy sampling, inconsistency in active surveillance guidelines, and inability of some patients to adhere to follow-up protocols. 19,20 While a clinician's recommendation is the most important factor in influencing a patient's decision to undergo active surveillance, 21,22 patients may also be unaware of conservative management options, or have difficulty understanding and weighing up treatment options. [22][23][24][25] The thought of not removing the cancer may also provoke fear of cancer growth and metastasis, which may contribute to decisions to opt out from active surveillance once started. 13 Given these significant patient and clinician barriers, effective strategies are needed to promote acceptance of active surveillance for men with low-risk lesions. Two possible strategies targeting the pathology diagnosis are (i) the recalibration of diagnostic thresholds, whereby the criteria by which prostate lesions are considered to be "cancer" is narrowed to only those at higher risk of progression 26,27 and/or (ii) adopting alternative diagnostic labels to describe low-risk prostate cancer lesions. 7,28,29 There is evidence that relabeling other low-risk lesions such as thyroid papillary microcarcinoma and ductal carcinoma in situ of the breast without using the word "cancer" may increase uptake of active surveillance and other conservative management options. [30][31][32] In the context of prostate cancer, the recent adoption of the International Society of Urological Pathology (ISUP) Grade Group system in pathology diagnosis was patient-centric and may reduce overtreatment of low-risk lesions. Relabeling "GS6" lesions as "GG1" lesions emphasizes the low-grade nature of the lesion to patients, having been assigned the most benign classification in the Grade Group system comprising grades 1−5, which may increase acceptability for conservative management options. 33,34 To inform consideration of potential changes to pathology diagnosis, we aimed to systematically search and synthesize published evidence to support or reject a change in the diagnostic criteria and/or terminology used to describe low-risk prostate lesions.
Specifically, we sought to answer four research questions about lowrisk prostatic lesions (GS6/GG1): (1) What are the clinical outcomes from active surveillance (compared to immediate treatment)? (2) What is the size of the reservoir of subclinical prostate cancer in men who died from other causes? (3)

| Study selection
Titles and abstracts retrieved from the database searches were screened by one reviewer (C. R. S.). The full texts of potentially SEMSARIAN ET AL. | 499 relevant articles were retrieved and independently screened by two reviewers (C. R. S. and T. M.), with disagreements adjudicated by a third reviewer (K. J. L. B.). Data was extracted by C. R. S. and T. L.

| Inclusion criteria
We included original studies published in the English language that examined prostatic lesions that were diagnosed as GS6. For Question One (active surveillance), prospective studies that reported on relevant clinical outcomes (e.g., disease progression, mortality rates) at 5 years or beyond, who were diagnosed with low-risk prostate cancer lesion (GS ≤6/GG ≤1 and PSA < 10 ng/mL) and not actively Preexisting systematic reviews were identified for both the active surveillance 13 and autopsy 35 questions, and provide summaries of the relevant evidence published before October 2017 and July 2013, respectively which we included in our review. Therefore, only studies that were published after these dates were additionally considered for inclusion in those categories.

| Exclusion criteria
We excluded abstracts, protocols, review articles, and opinion articles if no new data were presented. We also excluded studies where Gleason grading was not used, and where study cohorts comprised patients with only high-grade prostatic lesions or metastatic disease.
Among the active surveillance studies, we excluded studies that did not report on clinical outcomes relevant to disease progression, retrospective studies, and studies with <100 participants. Among the autopsy studies, we excluded studies where a systematic histological examination of the prostate was not performed, and studies with <100 cases. Among the reproducibility studies, we excluded studies with <10 independent assessments of the same histopathological slides. Given the large quantity of reproducibility studies retrieved, we also excluded studies that were judged to be at high risk of bias for representativeness of the sample (extent to which included participants were representative of the target clinical population).

| Data collection and synthesis
The data extraction templates were developed and piloted by two reviewers (B. N. and K. J. L. B). Data were extracted by one reviewer (C. R. S.) into an electronic datasheet and then reviewed by a second reviewer (T. M.) with disagreement resolved through discussion. We did not attempt to pool data, but instead undertook a narrative synthesis of the evidence in each category.

| Quality assessment of primary studies
All studies were assessed for risk of bias by C. R. S. and K. J. L. B.
using a list of standardized items adapted from the ROBINS-I tool 36 (active surveillance studies), the tool by Hoy et al. 37 (Table 1). Prostate cancer-specific mortality and rates of termination of active surveillance without medical indication from these studies are summarized in Figure 1.
The Kinsella review 13 included two randomized controlled trials (RCTs), 45,46 9 prospective cohort studies, 6,47-54 and 2 retrospective cohort studies 55,56 of active surveillance, with data collected from 1990 to 2016 from 10,354 patients from cohort studies that were multinational, 49 and in the United States of America, 6,47,48,53,55 UK, 45,50 Canada, 51 Denmark, 52 Australia, 56 Sweden, 46 and Italy. 54 No risk of bias assessment was reported. Prostate cancer-specific mortality rates were low in all studies, and similar rates of conversion from active surveillance to treatment were reported. The RCTs, Prostate Testing for Cancer and Treatment (ProtecT) 45 and Goteborg, 46 reported 10-year prostate cancer-specific mortality rates of 1.2% and 0% respectively in the active surveillance arms. In the two trials 55% (ProtecT) and 66%  41 Initially only men with very-low risk prostate cancer (n = 1293) were included, but later patients with low-risk prostate cancer (n = 525) were also included. After 15 years, the combined rate of metastasis (n = 1) and prostate cancer-specific mortality (n = 4) was 0.1%. The proportion who converted to treatment (without GG ≥2 cancer on monitoring biopsy in brackets) increased over time: 36% (15%) at 5 years, 48% (18%) at 10 years, and 52% (20%) at 15-years. 41 Meunier described a prospective study in an African Caribbean cohort (n = 234, 2005−2016). Participants had very low-risk prostate cancer (n = 190), low-risk prostate cancer (n = 25), or favorable intermediate-risk prostate cancer (n = 19). After 10 years, the rate of metastasis was 0.4% (one local lymph node metastasis) and prostate cancer-specific mortality was 0%. After a median follow-up of 4 years, 40% (n = 93) had terminated active surveillance; this was not medically indicated in 3.9% (patient preference n = 7, noncompliance n = 2). 42 Merrick reported a prospective study in the USA cohort (n = 340, 2005−2018). The rate of metastasis or prostate cancer-specific mortality was 0% over a median follow-up of 5.2 years. By 10 years, treatment was initiated in 7% (n = 24), with median time to treatment of 4.9 years.
After a median follow-up of 7.7 years, 57% (187/329) had terminated active surveillance and were treated, including 9% (n = 31) due to patient wish or physician's advice (without GG ≥2 cancer on monitoring biopsy or PSA rise).

|
What is the size of the reservoir of subclinical prostate cancer in men not known to have prostate cancer during their lifetime? (n = 33, Table 2

|
What is the reproducibility of the histopathological diagnosis of Gleason score 6 prostate lesions? (n = 8, Table 3) We screened the titles and abstracts of 235 articles identified through our database searches, and 2 additional articles 68,70 identified in the reference list of an included article. 67 This resulted in 120 full texts to screen, of which 112 were excluded (flow diagram in Supporting Information: 2c, reasons for exclusion in Supporting Information: 3c). We included 8 studies in our narrative synthesis (Table 3). One study was rated at low risk of bias, and 7 studies were rated at moderate risk of bias (Supporting Information: 4c).
The cross-sectional reproducibility study by Bulten et al.
(published in 2020) 61 was rated at low risk of bias. This study circulated a set of 100 prostate biopsies to 13 pathologists and 2 pathologists-in-training from 14 independent laboratories across 10 countries (see Table 3  Studies by Harnden et al., 65 Griffiths et al., 66 De la Taille et al., 67 and Lessels et al. 68   Albertsen et al. 71

| DISCUSSION
This review found robust evidence to inform discussion of possible recalibration of diagnostic thresholds, or use of alternative labels, for lowrisk prostate cancer. RCTs and cohort studies demonstrate that active surveillance is a safe option for very low-risk and low-risk prostate cancers, with extremely low rates of metastasis and prostate cancer- There have been a number of calls to adopt non-"cancer" labels for low-risk prostate cancer to reduce potential harms from T A B L E 2 (Continued) Reproducibility studies refer to cross-sectional studies whereby there were two or more independent readings of the same histopathological slides by three or more pathologists for the purpose of diagnostic classification. b See Supporting Information: Figure S4c for complete risk of bias assessment.  | 511 overdiagnosis and overtreatment. 28,29,[76][77][78][79] However, some pathologists argue against such a change on biological grounds, saying that GG1 lesions are morphologically indistinguishable from GG2-5 lesions, share many molecular hallmarks of prostate cancer, and therefore these lesions should retain the "cancer" label. 80,81 They also cite a high prevalence (up to 30%) of sampling variation (e.g., where a GG1 lesion is detected on biopsy in an individual harboring a higherrisk cancer). 80,82 Others have countered by arguing that even if GG1 lesions appear morphologically similar to higher-grade lesions and/or invade the stroma, they consistently behave as precancerous lesions-with extraprostatic extension and metastasis being extremely rare. 29 The low rates of metastasis and prostate cancer deaths in men undergoing active surveillance in this review support this characterization of a more benign clinical behavior. Further, among patients who are candidates for active surveillance based on clinical features, pathological upgrading (i.e., discovery of higher-grade lesions after the initial biopsy) has been found to have limited longterm effects on clinical outcomes. 29,83 New methods of detection (e.g., combined MRI-targeted and systematic biopsies) may also help to minimize the incidence of sampling variation. 84 Another key component of this debate is the hypothesized impact of relabeling lesions as noncancerous on patients' attitudes and behaviors. A common fear raised by those against such a move is that without the "cancer" label, patients may be less engaged with follow-up screening practices, possibly leading to poorer clinical outcomes. They point to suboptimal adherence to active surveillance protocols in both community and clinical trial settings, with biopsy adherence rates decreasing over time on surveillance. [85][86][87] A change in label is also argued to be unnecessary, as the shift in terminology from GS6 to GG1 communicates the low-risk nature of the lesion, and uptake of active surveillance has been increasing. 80 Countering these concerns, others hypothesize that relabeling cancerous lesions may decrease patient and family anxiety, as well as other psychosocial impacts on relationships, employment, and insurance policies.
This may further facilitate uptake of active surveillance, as well as preventing termination of active surveillance without medical indication. 78

| CONCLUSION
The decision whether to relabel low-risk prostate cancer as a noncancerous lesion (by adopting a new label or recalibration of the diagnostic threshold) is a complex issue. Many aspects of this discussion are evolving, such as the integration of multiparametric MRI in diagnostic protocols, 88 and the use of artificial intelligence to increase reproducibility of histopathological diagnoses. Further robust debate is needed in both clinical and pathology communities, informed by evidence such as that collected here.