EARLY ONLINE RELEASE

Transcription

1 EARLY ONLINE RELEASE Note: This article was posted on the Archives Web site as an Early Online Release. Early Online Release articles have been peer reviewed, copyedited, and reviewed by the authors. Additional changes or corrections may appear in these articles when they appear in a future print issue of the Archives. Early Online Release articles are citable by using the Digital Object Identifier (DOI), a unique number given to every article. The DOI will typically appear at the end of the abstract. The DOI for this manuscript is doi: /arpa RA The final published version of this manuscript will replace the Early Online Release version at the above DOI once it is available College of American Pathologists

2 Review Article The Association of Invasive Cribriform Lesions With Adverse Prostatic Adenocarcinoma Outcomes An Institutional Experience, Systematic Review, and Meta-analysis Xunda Luo, MD, PhD; Jasvir S. Khurana, MD; Nirag Jhala, MD; Huaqing Zhao, PhD, MS; He Wang, MD, PhD Context. Recent studies and a few reviews suggest that presence of invasive cribriform lesions (ICLs) in prostatic acinar adenocarcinoma correlates with adverse outcomes. However, a systematic review with meta-analysis on this correlation is currently lacking. Objective. To compare the likelihood of adverse outcomes by the status of ICLs in prostatic acinar adenocarcinoma with the meta-analysis of high-quality published data and institutional experience. Data Sources. PubMed, Scopus, manually searched references, and institutional data. Study Selection. Observational retrospective case-control studies or prospective cohort studies of adverse outcomes stratified by the status of ICLs were selected. Data Extraction. Study quality was analyzed. The prevalence of adverse outcomes stratified by the status of ICLs was extracted. Conclusions. Eighty-five cases were reviewed. Extraprostatic extension, seminal vesicle invasion, and regional lymph node metastasis were observed in 18 (45%), 14 (35%), and 7 (17.5%) of the 40 cases with cribriform lesions, respectively. These features were observed in 4 (8.9%), 1 (2.2%), and 0 (0%) of the 45 cases without ICLs. During the follow-up, biochemical prostate-specific antigen recurrence, local recurrence, and metastasis/diseasespecific death were documented in 7 (17.5%), 2 (5%), and 2 (5%) of the 40 cases with ICLs. These poor outcomes were found in 6 (13.3%), 1 (2.2%), and 1 (2.2%) of the 45 cases without ICLs. Meta-analysis revealed a significant increase in the risk of adverse outcomes in patients who had ICLs relative to those who did not (odds ratio, 3.95; 95% CI, ; I 2 ¼ 53%; Z ¼ 6.52; P,.01). These results suggest that presence of ICLs is associated with adverse outcomes. (Arch Pathol Lab Med. doi: /arpa RA) Prostate adenocarcinoma is the most common visceral malignancy and the second most common cause of cancer-associated deaths in men in the United States. 1,2 Currently, the histology pattern based Gleason score is used as a primary pathologic prognostic indicator, 3,4 along with tumor stage based indicators such as extraprostatic extension (EPE) and seminal vesicle invasion (SVI). The Gleason scoring system was initially proposed in 1966 and has been modified multiple times over the years to incorporate novel findings on the association of morphology and clinical outcomes in the disease of prostatic acinar adenocarcinoma. 5 8 Recently, the International Society of Urological Pathology (ISUP) proposed that all invasive cribriform lesions (ICLs) in prostatic acinar adenocarcinoma should be assigned a Gleason score of 4. 6 This was based on several studies suggesting a correlation between presence of ICLs and poor outcomes This correlation has also been described in several recently published review articles. 13,14 However, a systematic review with meta-analysis on this correlation is currently lacking. The present study aimed to elucidate this correlation, by performing a systematic review and meta-analysis on both the published data and the observations made in the present institution. MATERIALS AND METHODS Unpublished Data From the Present Institution All prostatic acinar adenocarcinoma cases with radical prostatectomy (RP) performed in our institution from January 2015 to January 2016 were reviewed. Cases with history of neoadjuvant/ Accepted for publication September 17, adjuvant therapies were excluded. The Grade Group 6, status of From the Department of Pathology and Laboratory Medicine, ICLs without central comedonecrosis, EPE, SVI, and regional Temple University Hospital, Philadelphia, Pennsylvania (Drs Luo, lymph node metastasis (LN) were extracted from the pathology Khurana, Jhala, and Wang); and Clinical Sciences, Lewis Katz School of Medicine, Temple University, Philadelphia, Pennsylvania (Dr reports of the RP specimens. Slides were reviewed to identify ICLs. Zhao). Immunohistochemistry for basal cells was performed, whereas a The authors have no relevant financial interest in the products or differentiation between an invasive cribriform lesion and an companies described in this article. intraductal carcinoma was necessary. The adjusted odds ratios Corresponding author: Xunda Luo, MD, PhD, Department of (ORs) of EPE, SVI, and LN, based on the presence and absence of Pathology and Laboratory Medicine, Temple University Hospital, ICLs, were computed with multivariate regression controlled for 3401 N Broad St, Philadelphia, PA ( xundaoph@gmail. com). patient age, ethnicity, preoperative prostate-specific antigen (PSA) level, and tumor load in terms of percentage of submitted tissue Arch Pathol Lab Med Cribriform Prostatic Adenocarcinoma Outcomes Luo et al 1

3 occupied by tumor. Follow-ups on the enrolled cases included biochemical prostate-specific antigen recurrence (BPR, defined as a PSA concentration greater than 0.1 ng/ml on 2 consecutive occasions), local recurrence, metastasis (Met), and disease-specific death (DSD). This study was designated as the current study in the meta-analysis. Guidelines for the Systematic Review and Meta-analysis This systematic review adhered to the guidelines proposed by the Meta-Analysis of Observational Studies in Epidemiology (MOOSE) 15 and the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA). 16 Data Sources and Literature Search Strategy A literature search was conducted, using both PubMed and Scopus databases, with the following keywords: prostat* AND (adenocarcinoma OR cancer OR carcinoma) AND cribriform AND (progression OR recurrence OR mortalit* OR death OR adverse OR (extraprostatic extension*) OR (seminal vesicle invasion*) OR (surgical margin*) OR PSA OR metastas*). The search was with no language restriction and from the inception of the databases until June 3, Study Selection The inclusion criteria are as follows: (1) retrospective, observational case-control studies or prospective, observational cohort studies; (2) comparison of prognosis between prostatic acinar adenocarcinoma with and without cribriform pattern; and (3) availability of adverse outcome data in published form. The exclusion criteria are as follows: (1) the reference is a review, case report, a correspondence, an animal or in vitro study; (2) the reference is about a diagnosis other than prostatic acinar adenocarcinoma with cribriform pattern; (3) data were not stratified or were not able to be extracted by the presence or absence of cribriform pattern; and (4) lack of prognosis data. Data Extraction Data extraction from the studies included for systematic review and meta-analysis was performed. The information extracted included study design, number of patients stratified by the presence of cribriform pattern, and prevalence of adverse outcomes. None of the authors of the included published studies was contacted for additional data. Outcomes The adverse outcomes reported by the included studies included EPE, SVI, BPR/PSA failure (as defined in individual studies), Met, and DSD. The status of surgical margin (SM) was not included as an adverse outcome measure, and the rationale is included in the discussion. Assessment of Study Quality Study quality was assessed with the Newcastle-Ottawa Scale. 17 Case-control studies were evaluated by adequacy of case definition (score 0 1, same below), representativeness of the cases (0 1), selection of controls (0 1), definition of controls (0 1), comparability of cases and controls on the basis of study design or data analysis (0 2), ascertainment of exposure (0 1), consistency in method of ascertainment for cases and controls (0 1), and nonresponse rate (0 1). Cohort studies were evaluated by representativeness of the exposed cohort (0 1), selection of the nonexposed cohort (0 1), ascertainment of exposure (0 1), demonstration that outcome of interest was not present at start of study (0 1), comparability of cohorts on the bases of design or analysis (0 2), assessment of outcome (0 1), adequate follow-up for outcomes to occur (0 1), and adequacy of follow-up of cohorts (0 1). A maximum score of 9 can be achieved for both case-control and cohort studies, and a cutoff of 5 points for satisfactory quality was applied for quality assessment. 17 Statistical Analysis and Data Synthesis IBM SPSS Statistics V20 (IBM Corporation, Somers, New York) was used for statistical analysis on the unpublished data from the present institution. RevMan V5.3 (Cochrane Collaboration, Copenhagen, Denmark) was used for meta-analysis. The t test was used to compare means of the independent samples. Univariate analyses on the association of ICLs with ethnicity, EPE, SVI, LN, and SM status were performed with v 2 test. Multivariate binary logistic regression was performed to derive the adjusted ORs of EPE and SVI on the basis of the presence and absence of ICLs. Cox regression survival probability analysis, which was used in several of the included studies 10,18,19 to calculate the hazard ratio for BPR, was performed in the current study for the same purpose. Odds ratios and 95% CIs based on the published studies selected and the current study were calculated by using meta-analysis (Mantel- Haenszel method). RESULTS Data From the Present Institution Demonstrated That There Is a Statistically Significant Association Between the Presence of Invasive Cribriform Lesion and Both EPE and SVI One hundred nine cases with RP were identified for the specified period, among which 85 were negative for a history of neoadjuvant/adjuvant therapies. Age of the patients enrolled was (mean 6 standard deviation, or SD) years. Race/ethnicity of the patients was as follows: 37 whites (43.5%), 39 blacks (45.9%), 6 Hispanics (7.1%), and 3 Asians (3.5%). ICLs were identified in 40 of the 85 RP specimens. The clinical characteristics of cases with and without ICLs are summarized in Table 1. There was no significant difference in age between the 2 groups (t ¼.59, P ¼.56). Although none of the 3 Asian patients presented with ICLs in the RP specimen, the difference in ethnicity constitution between the 2 groups was not statistically significant (v 2 ¼ 3.86, P ¼.28). Preoperative PSA level was higher in cases positive for ICLs; the difference, however, was not statistically significant (t ¼ 1.81, P ¼.08). Prostates with no ICLs are on average heavier and larger than those with ICLs, and the differences were statistically significant (weight: t ¼ 2.11, P ¼.04; maximum size: t ¼ 1.97, P ¼.05). Tumors with ICLs presented with significantly higher Grade Groups (v 2 ¼ 21.40, P,.001). Analyses on the associations of ICLs with EPE, SVI, and LN revealed that the associations were significant (EPE: v 2 ¼ 11.39, P ¼.001; SVI: v 2 ¼ 15.66, P,.001; LN: v 2 ¼ 8.58, P ¼.003). However, when the ORs on prevalence of EPE and SVI (with cribriform lesion versus without) were compared after cases had been stratified on the basis of Grade Groups, it appeared that these ORs were mostly statistically insignificant except for the SVI in Grade Group 2 cases (actual ORs, v 2 and P values not shown) and did not vary significantly with Grade Groups (Breslow-Day homogeneous association test; EPE: v BD2 ¼ 1.33, P ¼.51; SVI: v BD2 ¼ 2.00, P ¼.37). Presence of ICLs, per univariate analyses, was significantly associated with a higher tumor load defined as the percentage of prostate tissue involved by tumor (t ¼ 3.36, P ¼.001) and a higher likelihood of positive SM (v 2 ¼ 11.20, P ¼.001). Adjusted ORs of EPE and SVI, based on presence and absence of ICLs, were calculated via multivariate binary logistic regression with the aforementioned factors. The Grade Group was not included in the models because of its obvious association with the presence of the ICLs. 6 African 2 Arch Pathol Lab Med Cribriform Prostatic Adenocarcinoma Outcomes Luo et al

4 Table 1. Parameters Patient Characteristics by Absence/Presence of Cribriform Lesions (Current Study) Cribriform Lesion (N) No (45) Yes (40) Statistics Age, mean 6 SD, y t ¼.59, P ¼.56 Ethnicity, N v 2 ¼ 3.86, P ¼.28 A 3 0 B H 4 2 W Pre-op PSA, mean 6 SD, ng/ml t ¼ 1.81, P ¼.08 Prostate weight, mean 6 SD, g t ¼ 2.11, P ¼.04 Maximum prostate size, mean 6 SD, cm t ¼ 1.97, P ¼.05 Case Nos. by Grade Group, N All 5 Grade Groups Combined 1 All: v 2 ¼ 21.40, P,.001 All 15 0 EPE: v 2 ¼ 11.39, P ¼.001; v BD2 ¼ 1.33, P ¼.51 EPEþ 0 0 SVI: v 2 ¼ 15.66, P,.001; v BD2 ¼ 2.00, P ¼.37 SVIþ 0 0 LN: v 2 ¼ 8.58, P ¼.003 LNþ All EPEþ 1 5 SVIþ 0 a 5 a LNþ All 3 10 EPEþ 2 8 SVIþ 1 6 LNþ All 0 1 EPEþ 0 0 SVIþ 0 0 LNþ All 1 4 EPEþ 1 3 SVIþ 0 3 LNþ 0 3 Tumor load, mean 6SD, % t ¼ 3.36, P ¼.001 þsm, N 9 22 v 2 ¼ 11.20, P ¼.001 Follow-up, mean 6 SD, mo t ¼.78, P ¼.44 Abbreviations: A, Asian; B, black; EPE: extraprostatic extension; H, Hispanic; LN, regional lymph node metastasis; Pre-op, preoperative; PSA, prostate-specific antigen; SD, standard deviation; SM, surgical margin status; SVI, seminal vesicle invasion; W, white; v BD2, chi-square from the Breslow-Day test. a Layer for which the odds ratio was of statistical significance. American ethnicity (Wald ¼ 4.85; P ¼.03; OR, 5.4; 95% CI, ) and tumor load (Wald ¼ 8.31; P ¼.004; OR, 1.07; 95% CI, ) were independent predictors of EPE; presence of ICLs, however, was not an independent predictor of EPE (Wald ¼ 2.60; P ¼.11; OR, 3.39; 95% CI, ). For SVI, independent predictors identified by regression were preoperative PSA level (Wald ¼ 4.52; P ¼.03; OR, 1.16; 95% CI, ) and presence of ICLs (Wald ¼ 6.30; P ¼.01; OR, 17.86; 95% CI, ). Meta-analysis Showed That the Association Between Presence of ICLs and Adverse Outcomes Was Statistically Significant By using the aforementioned databases and literature search strategy, a total of 248 references were identified. One hundred thirty-nine references remained after duplication elimination. Preliminary screening excluded 104 references, and the exclusion criteria reduced the number of remaining references further by 27. Eligible published studies that were included in the meta-analysis, after a fulltext review, were 8 studies (Figure 1) published over a span of 18 years ( ). 9 12,18 21 Characteristics of the studies included in the metaanalysis are listed in Table 2. Seven of the 9 studies were case-control studies, and the other 2 were cohort studies. Sample size (N) ranges from 85 to 350. Presence of ICLs was confirmed by immunohistochemistry, or consensus, or both in all the studies. Adverse outcomes associated with ICLs reported by these studies included SM, EPE, SVI, Met, BPR, Gleason score upgrade for final RP from prior biopsy, and Arch Pathol Lab Med Cribriform Prostatic Adenocarcinoma Outcomes Luo et al 3

5 Figure 1. The Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRIS- MA) diagram for the literature search of the present meta-analysis. DSD. It is worthy to note that the definition of BPR varied significantly among and within the studies that monitored it. 9,10,18,19,21 The follow-up period for BPR, Met, and DSD also varied significantly. The significance of the associations between ICLs and most of the adverse outcomes was tested with multivariate regression analysis in 8 of 9 studies. The final numbers of the studies included that used EPE, SVI, regional lymph node metastasis, BPR, local recurrence, and Met/DSD as adverse prognostic outcomes were 4, 2, 1, 6, 1, and 3. The quality of these studies, assessed with the Newcastle- Ottawa Scale, 17 are summarized in Table 3. Eight studies scored full points (9 of 9), and one scored 8 of 9. Regression analysis was not used in the study that did not score full points to adjust the ORs; however, it did report OR of the prognostic parameter after stratifying the cases by Gleason scores, and hence got 1 point for comparability. 12 The qualities of these studies were all satisfactory with the previously specified criterion. Figure 2 is the Forest plot for each individual outcome measure included for the meta-analysis and for the combined effects. Pooled random-effects analysis of the 4 studies on the association between ICLs and EPE 12,20,21 revealed that this association was statistically significant (Z ¼ 4.08, P,.001), with a weighted mean OR of 4.94 and a 95% CI of 2.29 to The ratio of true heterogeneity to total observed variation, as quantified by I 2, was 44%. The between-studies variation, as quantified by Tau 2, was.26. Pooled random-effects analysis showed that the association between ICLs and SVI 21 was statistically insignificant (Z ¼ 1.77, P ¼.08), with an OR of 7.20 and a 95% CI of 0.81 to I 2 was 67%, and Tau 2 was The association between ICLs and regional lymph node metastasis, analyzed with the data from the current study, approached to statistical significance (Z ¼ 2.04, P ¼.04) with an OR of and a 95% CI of 1.12 to Pooled random-effects analysis of the 6 studies revealed a statistically significant association (Z ¼ 2.89, P ¼.004) between ICLs and BPR, 9,10,18,19,21 with a weighted mean OR of 2.72 and a 95% CI of 1.38 to I 2 was 71%, and Tau 2 was.45. Since BPR is associated with the length of followup, a hazard ratio is in theory better than an OR as a predictor for BPR. It was for this reason that a separate pooled random-effects analysis on the hazard ratios derived from Cox regression survival probability analyses of 4 4 Arch Pathol Lab Med Cribriform Prostatic Adenocarcinoma Outcomes Luo et al

6 Figure 2. Forest plots for odds ratios by subgroups of studies defined by the adverse outcomes, and for the odds ratio when all the subgroups were combined. A random-effect model was used for the meta-analysis. Abbreviations: BPR, biochemical prostate-specific antigen recurrence; df, degree of freedom; EPE, extraprostatic extension; Met/DSD, metastasis/disease-specific death; MH, Mantel-Haenszel method; SVI, seminal vesicle invasion. studies 10,18,19 was performed. This analysis also revealed a statistically significant association (Z ¼ 4.51, P,.001) between ICLs and BPR, with a weighted mean OR of 1.31, 95% CI of 1.16 to 1.47, I 2 of 67%, and Tau 2 of.01. The association between ICLs and local recurrence, analyzed with the data from the current study, was not statistical significance (Z ¼.67, P ¼.50) with an OR of 2.32 and a 95% CI of 0.20 to Arch Pathol Lab Med Cribriform Prostatic Adenocarcinoma Outcomes Luo et al 5

7 Table 2. Characteristics of the Studies Included for the Meta-analysis Radical Prostatectomy and Pathology Evaluation Source, y Study Design Prior Biopsy N Year Other Information Confirmation of Presence/Absence of Invasive Cribriform Lesions Rubin et al, Iczkowski et al, Dong et al, Kir et al, Sarbay et al, CCS Reviewed 150 (consecutive) Gleason score, PSA, stage, tumor volume, cribriform lesion, and high-grade cribriform PIN were documented for regression analysis CCS U 153 U Cases were selected to match N, follow-up duration, tumor stage, grade, margin status, and patient age CCS U 241 (from an original consecutive sample of 775) Viewers blinded to patient outcome; Gleason score, architectural pattern, PSA, age, pathologic stage, SM, and prostate weight were documented for regression analysis CCS U 235 (consecutive) EPE, bladder neck invasion, pathologic stage, SVI, and SM were documented for regression analysis CCS U EPE, bladder neck invasion, SVI, and SM documented but not used for regression analysis Yes (IHC) Yes (consensus) Yes (consensus) Yes (IHC) Yes (IHC) Keefe et al, CS Reviewed Biopsies grade. 3þ4, neoadjuvant therapy recipients were excluded; age, PSA, PSA density were documented for regression analysis Yes for the biopsy specimens (consensus) Kweldam et al, Choy et al, Current study CCS U Cases were selected to match age at surgery, PSA at diagnosis, and pt stage; only Gleason score 7 cases were included; reviewers blinded to patient outcomes; age, PSA level, Gleason score, pt stage, surgical margin status, lymph node status, Gleason grade 4 patterns, intraductal carcinoma, and tertiary Gleason grade 5 were documented for regression analysis CS U 585 (consecutive; exclusion criteria excluded 235, final enrollment 350) Age, preoperative PSA, surgical margin status, percentage of Gleason 4, architectural pattern documented for regression analysis CCS Reviewed 85 (consecutive) 2015 Age, ethnicity, cribriform lesion status, and tumor load documented for regression analysis Yes (IHC and consensus) Yes (morphology) Yes (IHC) Abbreviations: BPR, biochemical prostate-specific antigen recurrence; CCS, case-control study; CS, cohort study; DSD, disease-specific death; EPE, extraprostatic extension; IHC, immunohistochemistry; IQR, interquartile range; LN, lymph node metastasis; Met, metastasis; NA, not applicable; PIN, prostatic intraepithelial neoplasm; PSA, prostate-specific antigen; SM, surgical margin status; SVI, seminal vesicle invasion; U, unknown. The association between ICLs and Met/DSD, as analyzed by a pooled random-effects analysis of 3 studies, 11,19 was statistically significant (Z ¼ 5.31, P,.001), with a weighted mean OR of 6.15 and a 95% CI of 3.15 to I 2 was 0%, and Tau 2 was.00. When all the studies using all these aforementioned outcome measures were pooled for random-effects analysis, the results demonstrated that the association between ICLs and adverse outcomes was statistically significant (Z ¼ 6.52, P,.001), with a weighted mean OR of 3.95 and a 95% CI of 2.61 to The ratio of true heterogeneity to total observed variation was 53%, and the between-studies variation was.33. Although moderate to significant heterogeneity was identified with the pooled analysis, meta-regression was not performed to examine the impact of outcome measures on the study effect size (ie, association of ICLs with adverse outcome) because of small sample sizes. Figure 3 is the funnel plot for all the outcomes included in the random-effect meta-analysis. Egger test results indicated that the asymmetry of the scattered data points was not statistically significant, with an intercept of 1.08, and a 2- tailed P value of.17. DISCUSSION In the current study, data from the present institution were analyzed first with multivariate regression to identify statistically significant associations of ICLs with EPE, SVI, and LN, which are stage-defining features in RP. These data, together with BPR, local recurrence, and Met/DSD, were then combined with those from 8 previously published studies in a systematic review and meta-analysis on the association of ICLs with adverse outcomes in prostatic acinar adenocarcinoma. The OR of adverse outcomes, when ICLs were present in contrast to absent in biopsy or RP specimens, was 3.95 on average, with a 95% CI of 2.61 to This increased risk was of statistical significance (Z ¼ 6.52, P,.001), which strongly supported the vote of the 2014 ISUP consensus meeting to remove the entity of Gleason 3 ICLs and recognize all ICLs without central comedonecrosis as Gleason 4 lesions. 6 Outcome Measure or Prognostic Indicator Use of EPE, SVI, LN, and SM Status in Risk Assessment In our study, EPE, SVI, and LN in an RP specimen were used as measures for adverse outcomes to investigate whether there were statistically significant differences in 6 Arch Pathol Lab Med Cribriform Prostatic Adenocarcinoma Outcomes Luo et al

8 Table 2. Extended Clinical Course Outcome Measure(s) Reported Monitoring Measure(s) Definition (for BPR: Minimum Serum PSA Level, ng/ml) Follow-up Length History of Neoadjuvant/Adjuvant Therapy SM, EPE, SVI, LN, BPR BPR 0.2 on 2 consecutive occasions Not specified U BPR BPR 0.2 and persistent d No BPR, Met BPR, Met BPR: 0.5 before June 1999, then 0.2 till 2006, Median in years: BPR: 5.5; Met: 10.0 No then 0.1 since 2007; Met: cancer recurrence in a lymph node or at a distant site by radiography BPR BPR 0.1 Median 3.5 y (range, 1 7 y) No EPE, SM NA NA NA NA Gleason score upgrade, EPE, SVI NA NA NA No BPR, Met, DSD BPR, Met, DSD BPR: 0.2 on 2 consecutive occasions at least 3 mo apart; Met: cancer recurrence in a lymph node or at a distant site by radiography or pathology; DSD: determined by medical record review and death certificate Median, IQR (mo): Controls: 160, ; Cases: 100, All cases included were hormone naïve BPR BPR 0.2 Median 74 mo (range, mo), with death as end point No EPE, SVI, LN, BPR, local recurrence, Met or DSD BPR, local recurrence, Met or DSD 0.1 on 2 consecutive occasions Median 20 mo (range, 2 38 mo) No prevalence of these outcomes between case groups stratified by the status of ICLs. In 2 published studies included in the present meta-analysis, 12,20 EPE in an RP specimen was also used as an adverse outcome measure. The use of EPE and SVI as outcome measures in this scenario can be justified by the assumption that a carcinoma with higher Gleason score is more likely to be invasive, and thus less likely to be organ confined. More frequently, however, EPE and SVI in an RP specimen has been used as a prognostic indicator in multivariate regression to predict BPR, Met, and/or DSD. 9 11,19,21 This can be justified by the assumption that if the disease is already not organ confined by the time of RP, occult metastasis may have already occurred, and chances for BPR, clinically detectible Met, and DSD to occur during the follow-up period shall be increased. This assumption is also applicable to LN. However, since none of the RP cases free of ICLs were positive for regional lymph node metastasis, OR could not be computed and following analyses could not be performed (Table 1). In contrast to EPE, SVI, and LN, which can be used as both outcome measures and prognostic indicators, SM status instead may be suitable as only a prognostic indicator but not an outcome measure. A positive SM in an RP specimen means that the tumor is not fully excised, which projects to a likely worse prognosis compared to that following a complete excision. A positive SM in an RP specimen, unlike EPE, SVI, and LN, however, does not necessarily suggest that the tumor is more invasive. Many factors, which include but may not be limited to conservativeness of RP (surgeon-dependent) and the ways in which the RP specimen is processed in the gross room (pathologist-dependent), could contribute to a positive margin. It was for this reason that the SM status was not used in the unpublished study as an outcome measure. Risks of EPE, SVI, and LN with ICLs Stratified by the Grade Group in the Current Study Since the Grade Group was proposed in the 2014 ISUP consensus on the Gleason grading of prostatic adenocarcinoma and all ICLs in prostatic acinar adenocarcinoma were recognized as a Gleason score 4 pattern, 6 it is not surprising to see that no ICLs were identified in Grade Group 1 cases. Among cases with higher Grade Groups (2), prevalence of EPE and SVI stratified by Grade Groups was higher in those with ICLs; the ORs, however, were mostly not statistically significant. These results suggest that presence of ICLs is no longer a significant predictor for EPE and SVI when other pattern 4 and/or 5 lesions are present, and they support the categorization of ICLs as pattern 4. The fact that the results with Grade Group stratification were inconsistent with the results without stratification (EPE: v 2 ¼ 11.39, P ¼.001; SVI: v 2 ¼ 15.66, P,.001) indicates that the statistically identified significance may have originated from the 15 cases with a Arch Pathol Lab Med Cribriform Prostatic Adenocarcinoma Outcomes Luo et al 7

9 Table 2. Extended Selection of Outcome Measure(s) to Be Included in Meta-analysis Multivariate Statistical Analysis Cribriform Lesion Identified as an Independent Predictor by Multivariate Analysis Selected Excluded Reason(s) for Exclusion Cox hazards regression analysis No BPR, EPE, SVI, LN SM See Discussion Logistic regression analyses Yes BPR NA NA Cox hazards regression analysis Yes BPR, Met NA NA Cox hazards regression analysis Yes BPR NA NA Not used; however, the incidence of EPE and þsm was stratified by the Gleason scores of major components (3þ3 versus 3þ4/4þ3) Regression analysis NA EPE SM See Discussion Yes; this study is different from the rest in that it used biopsy cribriform lesions to predict adverse outcomes in radical prostatectomy specimens EPE Gleason score upgrade, SVI Not enough data provided in the original publication format to construct the cross tables Cox hazards regression analysis Yes Met, DSD BPR Not enough data provided in the original publication format to construct the cross tables Cox hazards regression analysis Yes BPR NA NA Logistic regression analyses Yes EPE, SVI, LN, BPR, local recurrence, Met or DSD NA NA Grade Group of 1, which increased the size of the cribriform-free case group and hence the overall OR. It is interesting to see that none of the ICL-free cases with a Grade Group of 2 or higher was positive for LN, whereas the prevalence of LN in ICL-positive patients increased from 4% (1 of 25) among Grade Group 2 cases, to 30% (3 of 10) among Grade Group 3 cases, and to 75% (3 of 4) among Grade Group 5 cases (Table 1). The association of LN with higher volume of ICLs has been reported previously. 22 The increased prevalence of LN with higher Grade Groups observed in the ICL-positive cases could be due to higher volume of ICLs, coexistence of other high-grade lesions, or both. However, the relatively small numbers of ICL-free cases (and ICL-positive cases as well) of higher Grade Groups made it difficult to further investigate on this issue. Table 3. Newcastle-Ottawa Scale on Methodologic Quality of the Studies Included in the Meta-analysis Source, y Study Design Adequacy of Case Definition Representativeness of the Cases/Cohorts (Cribriformþ and Cribriform ) Selection of Controls/ Nonexposed Cohort Definition of Controls Comparability of Cases and Controls/ Cohorts Rubin et al, CCS Iczkowski et al, CCS Dong et al, CCS Kir et al, CCS Sarbay et al, CCS Keefe et al, CS NA 1 1 NA 2 Kweldam et al, CCS Choy et al, CS NA 1 1 NA 2 Current study CCS Abbreviations: CCS, case-control study; CS, cohort study; NA, not applicable. 8 Arch Pathol Lab Med Cribriform Prostatic Adenocarcinoma Outcomes Luo et al

10 Figure 3. Funnel plot for the meta-analysis. Solid vertical line represents the mean effect, and the 2 dashed lines define the area for 95% confidence interval (61.96 standard errors). Numbers by the symbols: numbers of corresponding references from which the data were extracted (some studies are listed more than once, since more than 1 adverse outcome is reported). Abbreviations: BPR, biochemical prostate-specific antigen recurrence; C, current study; EPE, extraprostatic extension; LN, regional lymph node metastasis; Met/DSD, metastasis/disease-specific death; MH, Mantel-Haenszel method; RE, random-effect model; SVI, seminal vesicle invasion. Heterogeneity of the Studies Included for Meta-Analysis Six studies were included in the meta-analysis for the analysis on the association between ICLs and BPR (Table 2). There were remarkable variations in the definition of BPR and the length of follow-up period among these studies. 9,10,18,19,21 Definition of BPR ranged from 0.1 to 0.5 ng/ml for a minimum serum PSA level. The number of occasions for serum PSA elevation was not specified in 2 studies, 10,18 on 2 occasions with no specific interval in 1 study, 21 on 2 occasions at least 3 months apart in 1 study, 11 and persistently above the cutoff in 1 study. 9 Follow-up time for BPR ranged from 35 days 9 (may likely be due to incomplete excision) to 190 months. 11 These differences translated into moderate to significant heterogeneity when these data were pooled for meta-analysis: the ratio of true heterogeneity to total observed variation, as quantified by I 2, was 71%; the interstudy variation, as quantified by Tau 2, was.45 (Figure 2). Variation in methodology, in addition to clinical diversity, was obviously a major contributor to this significant heterogeneity. When Cox regression derived hazard ratios were pooled for meta-analysis, since the effect of length of follow-up had been adjusted, the interstudy variation, as quantified by Tau 2, was.01 only. Three published studies 12,20,21 and the current study were pooled for the analysis on the association between ICLs and EPE (Table 2). The study conducted by Keefe et al 20 is different from the other 3 in that the status of ICLs in biopsy specimens before RP was used in this study to correlate with the prevalence of EPE in RP specimens. Moderate heterogeneity was observed when these data were pooled for meta-analysis: the ratio of true heterogeneity to total observed variation, as quantified by I 2, was 44%; the interstudy variation, as quantified by Tau 2, was.26 (Figure 2). As one would expect, the heterogeneity of the EPE data was not as significant as that of the BPR data, since the definition of EPE in RP was more consistent across the studies and the main contributor to the observed heterogeneity was clinical diversity. When all the outcome measures from all the studies were pooled, the ratio of true heterogeneity to total observed variation, as quantified by I 2, was 53%; the interstudy variation, as quantified by Tau 2, was.33. Both variation in methodology and clinical diversity were obviously contributors to this moderate to significant heterogeneity. Metaregression to elucidate the origin of heterogeneity, however, was not performed owing to limited number of studies included in the present meta-analysis. CONCLUSIONS The meta-analysis results on the currently available quality data provide high level of evidence that presence of ICLs is associated with adverse outcomes. Neoadjuvant therapy and a more radical procedure may be considered if ICLs are identified in the biopsy specimen, and a more frequent follow-up and adjuvant therapy may be considered in case ICLs are identified in the RP specimen. Ascertainment of Exposure Same Method for Ascertainment for Cases and Controls Nonresponse Rate Table 3. Extended Outcome of Interest Not Present at the Beginning of the Study Assessment of Outcome Follow-up Long Enough for Outcomes to Occur Adequacy of Follow-up NA NA NA NA 8 1 NA NA NA NA Total Quality Scores Arch Pathol Lab Med Cribriform Prostatic Adenocarcinoma Outcomes Luo et al 9

11 References 1. Brawley OW, Ankerst DP, Thompson IM. Screening for prostate cancer. CA Cancer J Clin. 2009;59(4): Siegel R, Ward E, Brawley O, Jemal A. Cancer statistics, 2011: the impact of eliminating socioeconomic and racial disparities on premature cancer deaths. CA Cancer J Clin. 2011;61(4): Srigley JR, Amin M, Boccon-Gibod L, et al. Prognostic and predictive factors in prostate cancer: historical perspectives and recent international consensus initiatives. Scand J Urol Nephrol Suppl. 2005;(216): Tabayoyong W, Abouassaly R. Prostate cancer screening and the associated controversy. Surg Clin North Am. 2015;95(5): Epstein JI, Allsbrook WC Jr, Amin MB, Egevad LL. The 2005 international society of urological pathology (ISUP) consensus conference on gleason grading of prostatic carcinoma. Am J Surg Pathol. 2005;29(9): Epstein JI, Egevad L, Amin MB, Delahunt B, Srigley JR, Humphrey PA. The 2014 international society of urological pathology (ISUP) consensus conference on Gleason grading of prostatic carcinoma: definition of grading patterns and proposal for a new grading system. Am J Surg Pathol. 2016;40(2): Gleason DF. Classification of prostatic carcinomas. Cancer Chemother Rep. 1966;50(3): Gleason DF, Group TVACUR. Histologic grading and clinical staging of prostatic carcinoma. In: Tannenbaum M, ed. Urologic Pathology: The Prostate. Philadelphia: Lea & Febiger; 1977: Iczkowski KA, Torkko KC, Kotnis GR, et al. Digital quantification of five high-grade prostate cancer patterns, including the cribriform pattern, and their association with adverse outcome. Am J Clin Pathol. 2011;136(1): Kir G, Sarbay BC, Gumus E, Topal CS. The association of the cribriform pattern with outcome for prostatic adenocarcinomas. Pathol Res Pract. 2014; 210(10): Kweldam CF, Wildhagen MF, Steyerberg EW, Bangma CH, van der Kwast TH, van Leenders GJ. Cribriform growth is highly predictive for postoperative metastasis and disease-specific death in gleason score 7 prostate cancer. Mod Pathol. 2015;28(3): Sarbay BC, Kir G, Topal CS, Gumus E. Significance of the cribriform pattern in prostatic adenocarcinomas. Pathol Res Pract. 2014;210(9): Branca G, Ieni A, Barresi V, et al. An updated review of cribriform carcinomas with emphasis on histopathological diagnosis and prognostic significance. Oncol Rev. 2017;11(1): Iczkowski KA, Paner GP, Van der Kwast T. The new realization about cribriform prostate cancer. Adv Anat Pathol. 2018;25(1): Stroup DF, Berlin JA, Morton SC, et al. Meta-analysis of observational studies in epidemiology: a proposal for reporting meta-analysis of observational studies in epidemiology (MOOSE) group. JAMA. 2000;283(15): Liberati A, Altman DG, Tetzlaff J, et al. The prisma statement for reporting systematic reviews and meta-analyses of studies that evaluate healthcare interventions: explanation and elaboration. BMJ. 2009;339:b Wells GA, Shea B, O Connell D, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in meta-analyses. ohri.ca/programs/clinical_epidemiology/oxford.asp. Accessed June 3, Choy B, Pearce SM, Anderson BB, et al. Prognostic significance of percentage and architectural types of contemporary Gleason pattern 4 prostate cancer in radical prostatectomy. Am J Surg Pathol. 2016;40(10): Dong F, Yang P, Wang C, et al. Architectural heterogeneity and cribriform pattern predict adverse clinical outcome for Gleason grade 4 prostatic adenocarcinoma. Am J Surg Pathol. 2013;37(12): Keefe DT, Schieda N, El Hallani S, et al. Cribriform morphology predicts upstaging after radical prostatectomy in patients with Gleason score 3 þ 4 ¼ 7 prostate cancer at transrectal ultrasound (trus)-guided needle biopsy. Virchows Arch. 2015;467(4): Rubin MA, De La Taille A, Bagiella E, Olsson CA, O Toole KM. Cribriform carcinoma of the prostate and cribriform prostatic intraepithelial neoplasia: incidence and clinical implications. Am J Surg Pathol. 1998;22(7): Kryvenko ON, Gupta NS, Virani N, et al. Gleason score 7 adenocarcinoma of the prostate with lymph node metastases: analysis of 184 radical prostatectomy specimens. Arch Pathol Lab Med. 2013;137(5): Arch Pathol Lab Med Cribriform Prostatic Adenocarcinoma Outcomes Luo et al