Prognostic impact of the IASLC grading system of lung adenocarcinoma: a systematic review and meta‐analysis

Tumour grading is an essential part of the pathologic assessment that promotes patient management. The International Association for the Study of Lung Cancer (IASLC) proposed a grading system for non‐mucinous lung adenocarcinoma in 2020. We aimed to validate the prognostic impact of this novel grading system on overall survival (OS) and recurrence‐free survival (RFS) based on literature data.


Introduction
Tumour grading is an important component of pathologic assessment that promotes patient management.Therefore, several proposals have been introduced to prognostically stratify lung adenocarcinomas according to differentiation.In 2011, a pattern-based classification system was proposed. 1Figure 1 demonstrates the most frequent growth patterns of invasive non-mucinous adenocarcinomas.The architectural classification based on the predominant growth pattern had a good correlation with disease-free survival and overall survival (OS). 2 This architectural grade was included in the World Health Organization (WHO) Classification of lung tumours 3 ; however, there are several uncovered issues in this recommended grading system.Although the recently described complex glandular or discohesive patterns (Figure 2) or non-predominant high-grade patterns are associated with a worse prognosis, [4][5][6][7][8][9] they are not addressed in this grading system.
Formerly, Sica et al.10 proposed a scoring system that takes into account the two predominant growth patterns, and this successfully separated highly homogeneous groups of patients based on their risk of recurrence or death from primary pulmonary adenocarcinoma.Subsequently, Kadota et al. introduced a combined grading system focusing on architectural patterns and mitotic activity and it was an independent predictor of recurrence-free probability. 11Liu et al. recommended a prognostic model integrating the TNM categories and the morphological features to predict early relapse for patients with stage I lung adenocarcinoma. 12In our previous publication, spread through air spaces (STAS) was demonstrated as a superior unfavourable prognostic factor in lung adenocarcinoma; therefore, we proposed a grade combining architectural pattern and STAS.This grade was a predictor for recurrence in patients diagnosed with stage I pulmonary adenocarcinomas. 13oreira et al.8 evaluated the impact of different morphological factors of pulmonary adenocarcinoma on survival.Based on their results, they introduced a grading system in 2020, which considers the predominant architectural pattern with particular emphasis on high-grade components. 8Tumours with more than 20% of high-grade patterns, namely, solid, micropapillary and complex glandular patterns, were determined as grade 3 tumours.Furthermore, the International Association for the Study of Lung Cancer (IASLC) recommended this grading system for routine application.
The present study aimed to validate the prognostic impact of the novel IASLC grading system on OS and recurrence-free survival (RFS) using meta-analytic calculations.

Methods
We report our systematic review and meta-analysis in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 Statement. 14The study protocol was registered in the PROSPERO International Prospective Register of Systematic Reviews (CRD42023396059).Besides the outcomes stated in the study protocol, we also Meta-analysis: prognostic value of IASLC grading 53 evaluated vascular invasion as a secondary outcome.This should be noted as protocol deviation.
The comprehensive and systematic literature search was conducted by two independent review authors in three scientific databases, including MEDLINE (via PubMed), Embase, and Cochrane Central Register of Controlled Trials (CENTRAL).The search was conducted on 25 January, 2023.We searched in all fields / all text in each database.Regarding the fact that the novel IASLC grading system was published in 2020, a publication date filter from January, 2020 to January, 2023 was applied.The following search key was used in all databases: ("adenocarcinoma") AND ("lung" OR "pulmonary") AND ("international association for the study of lung cancer" OR "IASLC" OR "grade" OR "grading").

S E L E C T I O N A N D E L I G I B I L I T Y C R I T E R I A
Based on the search strategy, randomized controlled trials (RCTs) and non-randomized controlled studies might have been eligible for inclusion, but only the latter study design was found and included in the meta-analysis.Studies that compared the OS and RFS estimates (Outcome) of IASLC grade 1 with grade 2, grade 1 with grade 3, grade 2 with grade 3, and grade 1 + 2 with grade 3 (Exposure À Comparison) in at least 18-year-old patients diagnosed with pulmonary invasive non-mucinous adenocarcinoma (Population) were eligible for inclusion.Studies that did not report the IASLC grading system in primary pulmonary non-mucinous lung adenocarcinoma or did not evaluate human patients or were not written in the English language were excluded.Furthermore, studies that did not contain hazard ratio (HR) estimation with the corresponding 95% confidence intervals (CIs) of OS or RFS and studies reporting only neoadjuvant cases were also excluded.OS and RFS were defined as the time from the date of resection to death from any cause and the date of recurrence, respectively.
Screening for duplicates was performed by two independent review authors (F.H., N.Z.T.) based on 'title', 'publication year', and 'author' labels using the EndNote X9 reference management software (Clarivate, Philadelphia, PA, USA).After the removal of duplicates, two independent review authors (F.H., N.Z.T.) assessed all records at title, abstract, and full-text levels.All disagreements were resolved by third-party arbitration (T.Z.).

D A T A E X T R A C T I O N
Two independent review authors (F.H., N.Z.T.) extracted data from eligible studies into a standardized data collection form.The following data were extracted from the full text and corresponding supplementary information of eligible studies: article title, first author, publication year, Digital Object Identifier (DOI), TNM stages, RFS, OS, number and percentage of patients in different grades, sex distribution, age, vascular and lymphovascular propagation, pleural involvement, and the percentage of patients who actively smoked cigarettes at the time of surgery and who underwent surgery.In addition, HRs with the corresponding 95% CIs for OS estimates and RFS estimates of both univariate and multivariate analysis were also collected.Possible discrepancies were resolved by a third investigator (T.Z.).

R I S K O F B I A S A S S E S S M E N T
Based on the recommendations of the Cochrane Handbook for Systematic Reviews of Interventions, 15 the Quality In Prognosis Studies (QUIPS) tool was used for non-randomized controlled studies. 16The risk-of-bias tool was used by two independent review authors (F.H., N.Z.T.) to assess the quality of the studies included.A third review author (T.Z.) resolved any disagreement between the assessors.

S T A T I S T I C A L A N A L Y S I S
Both qualitative and quantitative synthesis of the data was performed.We provided summaries of intervention effects for each study by calculating pooled HRs for OS and RFS using R statistical software (v4.1.1)with the meta (v6.1-0) package.HRs were pooled into univariate and multivariate analyses using the random-effects model and inverse variance method with the restricted maximum likelihood (REML) estimation; they were displayed on forest plots.Pooled HR estimations and 95% CIs were calculated.Statistical heterogeneity was also analysed using the I 2 statistic and the v 2 test to acquire probability values; p < 0.1 was defined to indicate significant heterogeneity.

R E S U L T S O F S Y S T E M A T I C S E A R C H A N D S E L E C T I O N
The PRISMA flow diagram (Figure 3) displays the details of the selection process.A total of 3565 records were identified from the three scientific databases.8][19][20][21][22][23][24][25][26] In each of these full-text articles, the three categories of IASLC grade were listed separately, in addition to the OS and/or RFS data.

C H A R A C T E R I S T I C S O F T H E S T U D I E S I N C L U D E D
The characteristics of the included studies are presented in Table 1.A total of 4923 patients were studied in the 10 articles.Altogether 838, 2413, and 1672 patients belonged to grade 1, grade 2, and grade 3, respectively.Meta-analysis: prognostic value of IASLC grading 55

#
In these two publications investigating the differences of RFS estimates between grade 1 + 2 versus 3, three patients' cohorts were evaluated (stage I 22 , stage I 25 and stage I-III 25 ).
Due to duplicate of patients' cohorts, we were not able to do statistical analysis.However, we presented these RFS outcomes in OS was examined in nine articles, while RFS was evaluated in only six publications.In two publications 22,25 investigating the differences of RFS estimates between grade 1 + 2 versus 3, three patients' cohorts were evaluated.Due to duplicate of patients' cohorts, we were not able to do statistical analysis.Altogether, five, one, and four publications focused on stage I-III, stage III-IV, and stage I disease, respectively.

Q U A N T I T A T I V E S Y N T H E S I S O F O S A N D R F S
Regarding OS, altogether three cohorts (n = 1513) were included for univariate analysis: 199 patients had grade 1, 798 participants had grade 2, and 516 patients had grade 3 pulmonary adenocarcinoma.As Figure 4A demonstrates, there was a significant difference in OS between grade 1 and grade 3 (HR = 0.19, 95% CI: 0.05-0.66,p = 0.009).In addition, based on five studies (n = 2065), the OS estimate of patients with either grade 1 or 2 adenocarcinoma was better than in those with grade 3 adenocarcinomas, as shown in Figure 4B (HR = 0.43, 95% CI: 0.32-0.59,p < 0.001).In this multivariate analysis, grade 1 was defined in 433 patients, grade 2 in 912 participants, and grade 3 in 720 patients.
As shown in Figure 4C, concerning OS, three studies (n = 1772) were included in further multivariate analysis.Altogether, 258 patients had grade 1, 940 patients had grade 2, and 574 participants had grade 3 pulmonary adenocarcinoma.As shown in Figure 4C, the OS was better in patients with grade 1 than in those with grade 3 pulmonary neoplasm (HR = 0.21, 95% CI: 0.12-0.38,p < 0.001).Considering RFS, four cohorts (n = 2722) were included for multivariate analysis.Collectively, 398 patients were diagnosed with grade 1, 1427 with grade 2, and 897 with grade 3 pulmonary adenocarcinoma.As reflected in Figure 4D, grade 1 adenocarcinoma was more likely to have better RFS compared to grade 3 adenocarcinoma (HR = 0.22, 95% CI: 0.14-0.35,p < 0.001).
Despite our original protocol to perform subgroup analyses in all stages, we were able to implement subgroup analysis only in stage I.These analyses were multivariate.As Figure 5A demonstrates, three studies (n = 1644) proved that OS is better in grade 1 than in grade 3 pulmonary adenocarcinomas, with 254 patients in grade 1, 881 participants in grade 2, and 509 patients in grade 3 (HR = 0.17, 95% CI: 0.09-0.33,p < 0.001).Furthermore, as Figure 5B emphasizes, based on the same three studies, there was also Meta-analysis: prognostic value of IASLC grading 57 a significant difference in RFS between grade 1 and grade 3 tumours (HR = 0.20, 95% CI: 0.11-0.36,p < 0.001).Finally, four studies (n = 1680) were included in a model concerning OS with 421, 826, and 433 patients with grade 1, 2, and 3 lung adenocarcinomas, respectively.Figure 5C displays that OS was worse in patients with grade 3 than in patients diagnosed with either grade 1 or 2 pulmonary adenocarcinoma (HR = 0.46, 95% CI: 0.32-0.66,p < 0.001).The clinicopathologic covariates used in the multivariate model are displayed in Table S1.
In all but one model, the heterogeneity might not be important (I 2 = 0%-20%) based on the recommendations of the Cochrane Collaboration.In that model (Figure 4A) the heterogeneity was substantial (I 2 = 75%). 15

R I S K O F B I A S A S S E S S M E N T
In the univariate analysis, the overall risk of bias for OS was evaluated as low in two articles, while in one study, the overall risk of bias was presented as moderate.In the multivariate analysis, the overall risk of bias for OS was perceived as low and moderate in four and four studies, respectively.Finally, in the multivariate analysis for the RFS estimate, one article was found to have a low overall risk of bias, while three studies were evaluated as having a moderate risk.The most common reason for the moderate risk of bias was the lack of clinicopathological data for each grade.In these studies, this information was specified for the entire population.This fact may have caused confounding bias.Detailed results of the quality assessment for the risk of bias are presented in Figures S1-S3.

Discussion
Moreira et al. investigated the prognostic effect of different morphological parameters, namely, growth patterns, nuclear-, mitotic-, and cytological grades, STAS, and necrosis.Furthermore, they aimed to compose a prognostic grading system that more accurately stratifies the patients with lung adenocarcinoma.The growth pattern was the strongest predictor for recurrence and death.In addition, adenocarcinomas with ≥20% of high-grade patterns had a more aggressive behaviour, similar to those with a predominant high-grade pattern.During the model construction, they investigated different combinations of morphologic parameters to find the most optimal prognostic stratification.In contrast to others, 11,13 the other morphological parameters did not add further value to the model.The best model focused only on growth patterns with particular regard to high-grade patterns, and this was proposed by the IASLC for the grading of invasive pulmonary adenocarcinoma in 2020 (Table 2). 8n our meta-analysis, both in univariate and multivariate analyses, we demonstrated that OS is significantly better in grade 1 adenocarcinomas, than in  There are some studies in which no significant differences were detected among different grade categories.In the investigation of Weng et al., overall, 136 patients were included and only seven patients belonged to grade 1 adenocarcinoma (5.1%). 21This size discrepancy of grade categories could lead to statistical bias resulting in nonsignificant findings.In the study by Jeon et al., there were several non-cancerous deaths in the stage IA subgroup. 19Of the 32 patients, 17 (53.1%)participants died from other causes during the follow-up period.A further source of statistical bias could be the difference in adjuvant treatment modalities applied.Further differences between the findings of multivariate and univariate analyses may stem from the covariates used in the multivariate analysis, namely, age, gender, body mass index, smoking, diabetes, hypertension, surgical procedure type, SUVmax, carcinoembryonic antigen, adjuvant therapy, tumour size, TNM Stage, perineural invasion, lymphovascular invasion, STAS, visceral pleura invasion, poor differentiation, filigree micropapillary and discohesive growth pattern, CD8+ tumour infiltrating lymphocytes, Grading-Immunoscore type, architectural grade, Sica's grade, and EGFR mutation (Table S1).
However, there were more well-written publications 27,28 focusing on the validation of the prognostic role and reproducibility of the IASLC grading system; we were not able to include them in our meta-analysis for statistical reasons.
The limitations of our study are mainly attributed to deficiencies in the literature data mentioned above.Clinicopathological data were not available for each grade category, which may pose a risk of confounding bias.A significant proportion of the included studies did not investigate the prognosis of stage II and III pulmonary adenocarcinomas separately; therefore, we were unable to implement a subgroup analysis in these patients, only in stage I.In addition, we were not able to analyse the difference between grade 2 and grade 3, due to lack of data.In the future, multicentre and prospective studies are necessary, which better unify the clinicopathological data for each grade and involve more patients with stage II-III pulmonary adenocarcinoma.Future investigations should display more comparisons of grade 2 versus grade 3 tumours.Contrary to the protocol laid down, due to the lack of enough publications, we were not able to perform univariate analyses regarding RFS.The low number of publications included in the random effects model (limited data available) also represents a limitation in this meta-analysis; however, in all but one model the consistency (lack of heterogeneity) was excellent.
There are also some strengths to our study.To our knowledge, this is the first meta-analysis aimed at the prognostic validation of the IASLC grading system.Furthermore, we performed our statistical analysis on a patient cohort with more than 4900 participants.Our subgroup analysis included only stage I patients, in whom the grade may have the greatest prognostic role.The IASLC grading system is based on pattern recognition, which is currently part of the routine histopathology reporting for pulmonary adenocarcinoma.Therefore, this system is practical, easy to apply and it is not necessary to learn new methods for the assessment of grade.8][29] Table S2 displays the results of these investigations.According to their findings, the reproducibility of the IASLC grading system is between almost perfect and substantial.In the literature, seven studies compared different grading systems of pulmonary adenocarcinoma.Most investigations demonstrated that the IASLC grading system offers a superior prognostic grouping compared to the predominant pattern only. 8,22,24,25,27,28,30On the contrary, Park et al. could not prove a significant difference between the grading systems. 20Moreover, Rokutan-Kurata et al. found that the Sica's system reflects precisely the prognosis of patients, likewise the IASLC grading system. 28he aim of our study was not to assess the superiority of the IASLC classification system over others, but to investigate whether the expanded grade 3 category identifies patients with an unfavourable prognosis.Continuously collected published data and the result of our meta-analysis show the prognostic relevance of the IASLC grading system.This allows the IASLC grading system to become an internationally accepted and applied prognostic parameter in daily routine, and we encourage its utilization worldwide.

Figure 2 .
Figure 2. Some examples for grade 3 growth patterns of non-mucinous pulmonary adenocarcinoma.Poorly formed glands in a continuous spectrum between solid and acinar patterns (A), solid nests with artificial cleft formation (B), tumour cell trabecules in desmoplastic stroma (C), filigree micropapillary growth pattern (D), complex glandular growth patterns (E and F).

Figure 3 .
Figure 3. PRISMA flow diagram (2020) showing details of our systematic search and selection progress.

Figure 4 .
Figure 4. Albert Szent-Gy€ orgyi Medical School, University of Szeged.(A) Forest plot of univariate analysis of the included studies that compared grade 1 and grade 3 adenocarcinomas' overall survival estimates.(B) Forest plot of multivariate analysis of the included studies that compared grade 1 + 2 and grade 3 adenocarcinomas' overall survival estimates.(C) Forest plot of multivariate analysis of the included studies that compared grade 1 and grade 3 adenocarcinomas' overall survival estimates.(D) Forest plot of multivariate analysis of the included studies that compared grade 1 and grade 3 adenocarcinomas' recurrence-free survival estimates (CI, confidence interval; HR, hazard ratio; RE, random effects model).Ó 2024 The Authors.Histopathology published by John Wiley & Sons Ltd., Histopathology, 85, 51-61.

Figure 5 .
Figure 5. (A) Forest plot of multivariate analysis of the included studies that compared grade 1 and grade 3 adenocarcinomas' overall survival estimates.(B) Forest plot of multivariate analysis of the included studies that compared grade 1 and grade 3 adenocarcinomas' recurrence-free survival estimates.(C) Forest plot of multivariate analysis of the included studies that compared grade 1 + 2 and grade 3 adenocarcinomas' overall survival estimates (CI, confidence interval; HR, hazard ratio; RE, random effects model).

Table 1 .
Characteristics of included studies L, Lymphovascular invasion; n.d., No data; OS, Overall survival; PL, Pleural invasion; RFS, Recurrence-free survival; V, Vascular invasion.a Median range.b Median IQR.c Mean range.d Mean SD. e Mean.

Table 1 ;
these are not part of our statistical analysis.

Table 2 .
The categories of IASLC grade