Introduction

Lung adenocarcinoma (ADC) is the most common histologic subtype of primary lung cancer, which is considered a heterogeneous tumor with respect to its molecular, clinical, radiological, surgical, and pathological aspects [1]. According to the 2015 WHO classification, lung ADC is subclassified into five major histological patterns: lepidic, acinar, papillary, micropapillary, and solid ADC. Furthermore, other growth patterns, such as cribriform pattern, signet-ring cell proliferation, inflammatory cell-rich pattern, and tumor budding (TB), and their impact on prognosis have also been reported [2,3,4,5].

Discohesive growth pattern (Disco-p), characterized by single cells or small clusters of tumor cells with fibrous stroma, is known as “diffuse carcinoma” or “poorly cohesive carcinoma” of the stomach [6,7,8], and is sometimes also observed in lung ADC; however, there is no consensus regarding its nomenclature or subclassification. In lung ADC, the similar invasive pattern is discussed as TB in some previous reports. TB is identified in various cancers, especially in colorectal carcinoma (CRC), and its clinical impact on prognosis has been most intensively studied [9,10,11]. The methodology is well established in CRC: counting the number of tumor cells composing the nest at a higher magnification (×200) in the invasive margin [12]. With regard to lung ADC, a few reports that discussed TB referred to the CRC methodology [5, 13]. Kadota et al. [5] demonstrated that TB was an independent prognostic factor of stage I lung ADC. Yamaguchi et al. [13] showed that TB was observed in 43.1% of 181 patients with small-sized lung ADC and that the overall 5-year survival rate of the budding-positive and budding-negative groups was 67.5% and 88.3%, respectively, with multivariate analysis showing that high-grade TB was an independent prognostic marker. However, these reports had certain limitations: not involving the thin trabecular pattern with ≥5 tumor cells, not highlighting fibrous stroma such as “diffuse carcinoma” of the stomach, and the complexity of assessing at both low and high magnifications. To address these matters, in this study, we sought to analyze the clinicopathological significance of Disco-p through a histological review of 1062 resected lung ADCs. Furthermore, we compare its prognostic value to that of TB.

Materials and methods

Patients

A retrospective analysis was conducted on patients with lung ADC who underwent complete resection with curative intent at Kyoto University Hospital between 2001 and 2015. The patients were excluded from the present evaluation if they had multiple primary lung cancers, were treated with chemotherapy or radiotherapy prior to surgery, underwent incomplete resection, or had incomplete follow-up data, based on the clinical data retrieved from the Thoracic Surgical Database. As a result, the present histological investigation included 1062 lung ADCs. This study was approved by the institute’s ethics committee (Approval No. R1814).

Histologic evaluation

All resected specimens were formalin-fixed, sectioned, and stained with hematoxylin and eosin (H&E) in the conventional manner. Small tumors were entirely sampled histologically. Elastic stains were performed to evaluate the presence of pleural or vessel invasions in all cases. All the histological slides were reviewed by two pathologists (MRK and AY), who were blinded to patient information. All histological parameters were determined through a discussion between the pathologists reviewing the slides. The average number of tumor slides reviewed for each case was 3.3 (range, 1–20). According to the 2015 WHO classification, each tumor was classified using comprehensive histologic subtyping (five major histological patterns), and the percentage of each histologic component was recorded in 5% increments [1]. Histological grading was determined according to the 2004 WHO classification [14].

We defined Disco-p as an invasive growth pattern composed of single tumor cells, or trabeculae or small nests of tumor cells associated with desmoplastic fibrous stroma (Fig. 1). Disco-p was often observed adjacent to glandular, solid or micropapillary growth patterns (Fig. 2a–d). To distinguish it from solid growth patterns, the trabeculae of Disco-p were categorized as thin, consisting of one or two tumor cell rows (Fig. 2b). The number of tumor cells composing the nest had no limitation. Furthermore, we carefully distinguished between Disco-p and stromal invasive-type micropapillary pattern (SMPP) as follows: if small nests of tumor cells were observed within a slit-like structure, we considered them as Disco-p (Fig. 2d). If, however, the small tufts were floating within distinct lacunar spaces, we decided that the area was rather an SMPP (Fig. 2e). Disco-p can be highlighted with CK7 (clone, OVTL12/30, Dako, Denmark) immunoexpression (Fig. 2f, g).

Fig. 1: Representative image and schema of discohesive growth pattern (Disco-p).
figure 1

a Disco-p was defined as an invasive growth pattern composed of single tumor cells, or trabeculae or small nests of tumor cells with desmoplastic fibrous stroma (H&E staining). b Schematic image of Disco-p (blue and yellow indicate tumor cells and fibrous stroma, respectively).

Full size image
Fig. 2: Representative images of discohesive growth pattern (Disco-p) with major histological growth pattern.
figure 2

a Trabecular nest of tumor cells with desmoplastic stroma (right upper-half, Disco-p) associated with anastomosing glandular proliferation (left lower-half) which mimics complex glandular pattern (CGP). b Disco-p (right-half) associated with a solid growth pattern (left-half). c Disco-p (right-half) associated with a micropapillary pattern (aerogenous micropapillary pattern: AMPP). d Small nests of tumor cells with fibrous stroma were also considered to be Disco-p. e Stromal invasive-type micropapillary pattern (SMPP) showing small tufts surrounded by distinct lacunar spaces, unlike Disco-p. f, g Disco-p is highlighted with cytokeratin immunoexpression (scale bar: 100 μm). The concept of Disco-p is not the same as those of SMPP and CGP in terms of invasive pattern composed of single tumor cells, or trabeculae or small tumor nests with desmoplastic stroma.

Full size image

To assess Disco-p, first, we reviewed the slides at low- to intermediate-power fields (×40–×100). Then, we recorded the percentage of Disco-p in 5% increments independently of the major histologic pattern (Fig. 3). For the determination of the predominant pattern of lung ADC in this study, we opted to subclassify Disco-p as a solid pattern, because it is poorly differentiated and does not conform to any other growth pattern. In order to evaluate the utility of Disco-p, we also assessed TB according to the report of Kadota et al. [5]. Briefly, we reviewed the slide at intermediate-power fields (×100) and assessed the area with the maximum number of TB (Fig. 3). In their report, TB as small tumor nests composed of <5 tumor cells. In this study, we quantified it by counting ten high power fields (HPFs) at ×200 magnification and grading TB by the maximum number of tumor buds per HPF. TB was classified as grade 0 (no bud per HPF), grade 1 (1–4 buds), grade 2 (5–9 buds), or grade 3 (≥10 buds).

Fig. 3: Assessment of discohesive growth pattern (Disco-p) and tumor budding (TB).
figure 3

Disco-p was assessed by low- to intermediate-power fields (×40–×100) and recorded in 5% increments, while TB was assessed at intermediate-power fields (×100) and quantified by counting 10 HPFs (×200).

Full size image

Finally, we investigated the associations between Disco-p and the following clinicopathological factors: sex, age, smoking status, tumor grade, histological subtypes, tumor cell character (terminal respiratory unit (TRU) type or non-TRU type) [15], lymphatic invasion, vascular invasion, pleural invasion, spread through air spaces (STAS) according to the 2015 WHO classification [1], and TNM staging according to the 8th TNM classification [16].

Detection of genetic alterations in EGFR, KRAS, HER2, BRAF, ALK, ROS1, and RET

The association between tumors with Disco-p and gene alterations in EGFR, KRAS, HER2, BRAF, ALK, ROS1, and RET was evaluated. Mutation analysis was reported in our previous reports. Briefly, EGFR mutations were studied using the polymerase chain reaction (PCR)–single-strand conformation polymorphism method (PCR–SSCP) before 2009 [17], and the PNA-LNA PCR clamp method after 2010. HER2 mutations were also studied using PCR–SSCP [18]. KRAS and BRAF mutations were investigated using a modified mutagenic PCR-restriction enzyme fragment-length polymorphism technique [17, 19]. ALK fusion was detected using reverse-transcription PCR and fluorescence in situ hybridization (FISH) [20, 21]. ROS1 fusion was detected by FISH using Vysis ROS1 Dual Color Break Apart Probes (Vysis LSI; Abbott Laboratories, Chicago, IL, USA), according to the manufacturer’s instructions [22]. RET fusion was also detected by FISH using Kreatech RET (10q11) Break FISH probe (Leica Biosystems, Germany) and RET split Dual color FISH probe (GSP Lab., Inc, Japan), according to the manufacturer’s instructions.

Statistical analysis

Chi-square and Fisher’s exact tests were used to analyze categorical data. The means and standard deviations (SDs) of all variables were calculated. The survival rates were calculated using the Kaplan–Meier method, and the differences were analyzed using the log-rank test. Multivariate analysis was performed using the Cox’s proportional hazards model. Multivariate models were built to include factors that were significant in the univariate analysis. All statistical tests were two-sided at a significance level of 5%. Data analysis and summary graphs were generated using the JMP statistical software package, version 13 (SAS Institute, Cary, NC, USA).

Results

Clinicopathological characteristics

The distribution of clinicopathological variables is presented in Table 1. The mean age at diagnosis was 66.2 ± 9.8 years (range 23–88 years). There were 515 (48.5%) male and 547 (51.5%) female patients. Five hundred and sixty-one (52.8%) patients were smokers (221 current and 340 ex-smokers; smoking index: 45.7). The mean tumor size was 23.9 ± 14 mm (range 3–120 mm). Most patients underwent lobectomy or more extensive resection (n = 766, 72.1%), whereas the rest (n = 296, 27.9%) underwent limited (segmentectomy or wedge) resection. In total, 186 (17.5%) patients died during follow-up and 241 (22.7%) relapsed. The mean follow-up time at the end point of analysis was 61.5 ± 35.7 months. The numbers of patients at each pathologic stage were as follows: 0, 21 (2.0%) patients; I, 832 (78.3%) patients (IA1, 305 (28.7%) patients; IA2, 276 (26.0%) patients; IA3, 96 (9.0%) patients; IB, 155 (14.6%) patients); II, 111 (10.5%) patients (IIA, 13 (1.2%) patients; IIB, 98 (9.2%) patients); and III, 98 (9.2%) patients (IIIA, 82 (7.7%) patients, and IIIB, 16 (1.5%) patients).

Table 1 Clinicopathological features of lung adenocarcinoma with and without discohesive growth pattern (Disco-p).
Full size table

Correlation of Disco-p with clinicopathological characteristics

Of the 1062 cases, Disco-p was present in 203 tumors (19.1%) (Table 1). Its distribution varied from 5 to 40% (mean 10.75, SD 9.92). Disco-p was predominantly observed in male patients (p < 0.001) and smokers (p < 0.001). It was predominantly seen in tumors with larger tumors (<25 vs. over 25 mm), lymph node metastasis (no metastasis vs. N1 or N2), higher pathological stage, pleural invasion, lymphatic invasion, vascular invasion, or STAS (p < 0.001 for all characteristics); whereas Disco-p was not associated with age nor with tumor cell characteristics (TRU type vs. non-TRU type). In terms of its association with TB, high-grade TB (grades 2 and 3) was seen in 42 cases (3.9%), which accounts for 20.7% of Disco-p-positive tumors. According to the WHO classification, Disco-p was mostly seen in solid ADCs (94/154, 61.0%), followed by micropapillary ADCs (14/35, 40.0%), acinar ADCs (30/115, 26.1%), and papillary ADCs (61/459, 13.3%); whereas it was not observed in minimally invasive or lepidic ADCs. Of the 94 Disco-p cases in 154 solid ADCs, 84 cases predominantly had the “classical solid pattern,” which shows a major component of polygonal tumor cells forming sheets that lack lepidic, acinar, papillary, or micropapillary patterns [1]. Regarding micropapillary pattern, of a total of 1062 tumors, 103 presented both micropapillary component and Disco-p. These 103 tumors were further divided as follows: 87 showed an aerogenous type micropapillary pattern (AMPP), 2 displayed an SMPP, and 14 showed the combined AMPP and SMPP. Although eight tumors displayed a dubious pattern among tumors with SMPP and Disco-p in this evaluation, we carefully divided them according to our criteria. Furthermore, in 35 micropapillary predominant ADCs, 14 had Disco-p and 21 did not. Among the 21 Disco-p-negative cases, we found AMPP in 14 cases, SMPP in 2 cases, and a combined pattern in the remaining 5 cases. Among the 14 Disco-p-positive cases, AMPP was observed in 8 cases, SMPP in 1 case, and a combined pattern in the remaining 5 cases. Since the morphology of Disco-p is similar to SMPP, after removing SMPP and combined pattern cases from micropapillary predominant ADCs, we reanalyzed the frequency of Disco-p in micropapillary predominant ADCs. We found the frequency to be 36.4% (8/22), which is still higher than the frequency of other subtypes.

Associations of Disco-p with gene alterations

Among the 1062 lung ADCs, 495, 221, 145, 228, 394, 236, and 262 tumors were assessed for EGFR, KRAS, HER2, BRAF mutations, and ALK, ROS1, and RET fusions, respectively. Of these, 240 (48.5%), 26 (11.8%), 6 (4.1%), and 3 (1.3%) tumors harbored EGFR, KRAS, HER2, or BRAF mutations, respectively. In addition, 17 (4.3%), 2 (0.8%), and 3 (1.1%) tumors harbored ALK, ROS1, or RET fusions, respectively. EGFR wild type (p < 0.001) and ALK fusions (p = 0.008, Fisher’s exact test) were frequently observed in tumors with Disco-p, whereas there were no correlation between Disco-p-tumors and KRAS mutations (p = 0.247), HER2 mutations (p = 0.688), BRAF mutations (p = 0.737), ROS1 fusions (p = 0.412), or RET fusions (p = 0.731) (Table 2).

Table 2 Correlation between discohesive growth pattern and gene alterations.
Full size table

Disco-p and patients’ outcome

Figure 4a, b shows the survival curves according to the presence or absence of Disco-p for all patients, excluding those with ADC in situ (AIS—stage 0 patients). The 5-year overall survival (OS) and disease-free survival (DFS) rates for patients with Disco-p (5-year OS rate, 61.1%; 5-year DFS rate, 44.3%) were lower than for those without Disco-p (p < 0.001). With subclass analysis of the patients with stage I lung ADCs, the 5-year OS and DFS rates for patients with Disco-p (5-year OS rate, 71.5%; 5-year DFS rate, 59.6%) were also lower than for those without Disco-p (p < 0.001) (Fig. 4c, d). Moreover, tumors with Disco-p had worse prognoses than tumors without Disco-p, when we subclassified tumors according to the predominant histological patterns: papillary (p < 0.001, both OS and DFS) and acinar (p < 0.001, both OS and DFS) subtypes (Table 3), whereas no significant difference was seen in micropapillary or lepidic subtypes. We showed that 84 “classical solid ADC” cases had Disco-p and 60 cases did not among 154 solid ADCs. In the subgroup analysis of tumors, including the 84 and 60 solid ADCs, tumors with Disco-p also had worse prognoses than tumors without Disco-p (p < 0.001, both OS and DFS) (Table 3). Regarding micropapillary predominant ADC, Disco-p was not associated with OS and DFS (data not shown). By assuming that Disco-p and SMPP share the same features, we attempted to analyze prognostic significance between tumors with AMPP alone and tumors with Disco-p and/or SMPP. We observed that tumors with only AMPP tended to have a better prognosis compared with tumors with Disco-p and/or SMPP, although no significant difference was seen both in OS and DFS (p = 0.117 and p = 0.123, respectively).

Fig. 4: Kaplan–Meier curves of overall survival and disease-free survival based on the presence of Disco-p.
figure 4

a, b Patients at all stages excluding stage 0 (AIS). c, d Stage I patients.

Full size image
Table 3 Subgroup analysis of patients’ outcome.
Full size table

We assessed the clinicopathological significance of Disco-p for prognosis based on univariate and multivariate analyses. These analyses were limited to stage I patients. On univariate analysis, age, smoking status, sex, stage, pleural invasion, lymphatic invasion, vascular invasion, solid pattern, micropapillary pattern, TB, and Disco-p were independently associated with worse prognoses (data not shown). Based on univariate analysis results, multivariate analyses were performed using the Cox’s proportional hazards model (Table 4). On multivariate analysis, age (hazard ratio (HR), 1.796; 95% confidence interval (CI), 1.307–2.498), pleural invasion (HR, 1.559; 95% CI, 1.112–2.182), lymphatic invasion (HR, 2.098; 95% CI, 1.420–3.044), and Disco-p (HR, 2.572; 95% CI, 1.789–3.680) were independent prognostic factors for OS. Moreover, age (HR, 1.585; 95% CI, 1.202–2.105), pleural invasion (HR, 1.545; 95% CI, 1.145–2.097), vascular invasion (HR, 1.879; 95% CI, 1.380–2.550), and Disco-p (HR, 3.413; 95% CI, 2.482–4.683) were independent prognostic factors for recurrence. Although TB was associated with a worse prognosis on univariate analysis, it was not shown to be an independent prognostic factor on multivariate analysis.

Table 4 Multivariate analysis of overall survival and disease-free survival.
Full size table

Discussion

In this study of 1062 resected lung ADCs, we found that Disco-p was present in 203 tumors (19.1%). Disco-p was predominantly observed in male patients and smokers, and associated with a larger tumor size, lymph node metastasis, higher pathological stage, lymphatic invasion, vascular invasion, pleural invasion, STAS, and wild-type EGFR and ALK fusions. We demonstrated that patients harboring tumors with Disco-p have a poorer OS and DFS, and Disco-p was also an independent prognostic factor on multivariate analysis, whereas TB was not. Our findings highlight the prognostic importance of Disco-p.

To date, the invasive pattern composed of single tumor cells, or small tumor nests in lung ADC, has still been discussed as TB, which mimics the histological feature of Disco-p, in a few reports [5, 13]. TB has been well studied in the past two decades in many types of cancers [12]. Numerous methods of assessing TB were proposed: where to assess (“peritumoral” or “intratumoral”), where to cut off, and whether to use immunohistochemical staining or H&E staining only [12]. In CRC, the methodology is well established; TB is assessed using H&E staining, defined as single cells or clusters of <5 cells, counted in the hot spot area at ×200 magnification. To estimate the hot spot area, screening at least 10 HPFs at the invasive margin was suggested; findings of 0–4 buds are classified as low grade, 5–9 buds as intermediate grade, and ≥10 buds as high grade [9, 12]. As for TB in lung ADC, the way of assessing TB was in accordance with CRC criteria mentioned in a few previous studies [5, 13]. The methods of assessment in lung ADC are associated with a few issues. First, the definition of TB focuses on the number of tumor cells composing the tumor nest (<5 cells), although the definition of “diffuse carcinoma” or “poorly cohesive carcinoma” of the stomach does not limit the number of tumor cells of the nest. Thin trabecular pattern with ≥5 cells along with fibrosis is often seen in lung ADC as well as in “diffuse carcinoma” of the stomach, thus, we considered that this pattern should be evaluated. Second, assessing TB involves only the invasive margin of the tumor, that is, the peritumoral area. Certainly, TB of CRCs was mostly observed at the invasive deep margin, but in lung ADCs, the thin trabecular pattern or small tumor nests are scattered through the intratumoral area with desmoplastic fibrous stroma. Thus, in the lung, the most suitable area to evaluate must be selected at low to intermediate magnification. This process is somewhat arduous. To address these issues, we proposed Disco-p as an invasive growth pattern composed of single tumor cells, or trabeculae or small nests of tumor cells with desmoplastic fibrous stroma, irrespective of the number of tumor cells in the tumor nest. We also predicted Disco-p as an unfavorable prognostic factor. Disco-p can be identified at intermediate magnification anywhere in the tumor area; thus, it is easier to detect the invasive pattern than TB. Like cancers of the other sites, TB was reported to be related to a worse prognosis in lung ADC [5, 13]. For example, Kadota et al. demonstrated that TB was an independent prognostic factor of stage I lung ADC [5], and Yamaguchi et al. reported that high-grade TB was an independent prognostic marker [13]. In the current study, high-grade TB was associated with larger tumor size, lymph node metastasis, lymphovascular invasion, pleural invasion, STAS, and Disco-p (data not shown); however, on multivariate analysis, TB was not an independent prognostic factor. This suggests that identifying the distinct growth pattern, rather than focusing on the number of tumor cells, is more relevant for the prognosis. Taken together, our data suggest that Disco-p is superior to TB in terms of simplicity of methodology and prognostic value.

In the 2015 WHO classification, it was recommended that lung ADC should be subclassified into five subtypes based on the five major histological patterns: lepidic, acinar, papillary, micropapillary, and solid [1]. We found that Disco-p was present in 19.1% of patients in this study, and that the range of area occupied by Disco-p in the tumor was 5–40% with a mean of 10.75% (SD 9.92). This suggests that Disco-p is not a rare component. Nevertheless, there is no consensus as to which subtype Disco-p should be classified. Based on our data, we do not propose a new subtype, hypothetically determined “Disco-p predominant ADC” because there were only four cases (0.4%) in which Disco-p was the predominantly observed growth pattern. In this study, we opted to subclassify Disco-p as a solid pattern, because it was considered to have a poorly differentiated growth pattern; we demonstrated that Disco-p was strongly correlated with poor prognosis similar to the solid growth pattern. However, it could be argued that Disco-p pattern should be categorized with the other pattern subtypes. Moreira et al. proposed a new growth pattern named “complex glandular pattern (CGP)” and concluded that this pattern should be considered high-grade ADC [23]. In addition, Kuang et al. investigated the clinical relevance of CGP according to the definition of Moreira’s paper and concluded that it would be reasonable to classify CGPs as a new type of lung ADC [24]. In their concept, CGP contained two patterns: cribriform pattern (defined as nests of tumor cells with sieve-like perforation) and fused gland pattern (defined as fused glands with irregular borders, back-to-back glands without intervening stroma, or ribbon-like formations). The latter pattern, “fused gland pattern,” may resemble Disco-p; however, we focused on the invasive pattern without glandular formation in desmoplastic fibrous stroma. Thus, we consider that the concept of Disco-p is not the same as that of CGP, although they may have partially overlapping features. From another perspective, it could be considered that Disco-p should be categorized as a micropapillary pattern. Currently, two patterns are recognized as micropapillary pattern [25]: small papillary tufts without a central fibrovascular core, floating within alveolar spaces (AMPP), and small tufts lacking a central fibrovascular core, surrounded by lacunar spaces invading the fibrotic stroma (SMPP) [26]. We consider that the first pattern is distinguishable from Disco-p; however, the latter pattern, which is often observed in organs such as the breast, can be harder to distinguish from Disco-p. We found that 103 tumors out of 1062 tumors simultaneously exhibited micropapillary component and Disco-p. Of those, only eight cases showed a dubious pattern between SMPP and Disco-p. Based on this, we argue that Disco-p can usually be distinguished from SMPP, and does not necessarily need to be categorized as a micropapillary pattern; nevertheless, a careful evaluation is necessary to distinguish Disco-p from SMPP. In the literature, Sica et al. have proposed a histological grading system of lung ADC, in which a Disco-p-like pattern was demonstrated in high-grade tumors, although the paper did not describe this pattern in detail [27]. Our findings suggest that it may be suitable and acceptable to consider Disco-p as a high-grade pattern, even if Disco-p is classified as solid pattern or SMPP. Although we estimated Disco-p as a solid component, it is necessary to further confirm this classification in future studies from different institutes.

Our data also demonstrated that tumors with Disco-p, pertaining to the papillary, acinar, and even the solid ADC subgroup, had worse prognoses than tumors without this pattern. In contrast, Disco-p was not associated with OS and DFS in micropapillary predominant ADC. This could be caused by Disco-p not being clearly separated from SMPP. Thus, by assuming that Disco-p and SMPP shared the same features, we attempted to analyze prognostic significance between tumors with AMPP alone and tumors with Disco-p and/or SMPP. We discovered that tumors with AMPP alone tended to have a better prognosis compared with tumors with Disco-p and/or SMPP, although no significant difference was seen both in OS and DFS. The results demonstrated that Disco-p can be associated (although not statistically significant) with prognosis even in micropapillary predominant ADC. We believe this occurred because AMPP is also a strong prognostic factor for recurrence and death. Based on these results, we consider that Disco-p should be recorded independently of the five major subtypes, although further validation studies are needed before such a recommendation is made. In this study, we found eight Disco-p tumors harboring ALK fusion, and there was an association between ALK fusion and Disco-p. ALK rearrangement ADC usually shows a unique histological pattern: a solid growth pattern with signet-ring cells or mucinous cribriform formation [28]. Moreover, it exhibits an aggressive behavior and an epithelial–mesenchymal transition (EMT) phenotype, characterized by the loss of E-cadherin and increased vimentin expression compared with another genotype [29]. Voena et al. showed the mechanism of modulating EMT by ALK rearrangement [30]. They showed that EML4-ALK regulated E-cadherin through repression of the epithelial splicing regulatory protein 1, a key regulator of the splicing switch during EMT, using ALK-rearranged cell lines. Since the morphology of EMT mimics Disco-p, we consider that our results support those of the previous papers. However, in this study, we did not perform immunohistochemical analyses for the EMT phenotype. In the future, more cases need to be collected and evaluated by a combination of morphological, immunohistochemical, and genetic analyses.

This study has some limitations. First, this is a single-center study without a validation cohort. Further studies are required to evaluate the impact of Disco-p. Second, there might be interobserver variability when estimating Disco-p. Some reports highlighted the tumor cells with antibodies for cytokeratin for detection of TB [13, 31]. However, they just demonstrated what TB is and did not utilize them to assess it. This might be more accurate than using H&E staining only, but it is costlier and more burdensome. We consider that assessment with H&E staining is sufficient to distinguish Disco-p.

In conclusion, we have demonstrated that Disco-p is an independent factor for poor prognosis with respect to recurrence and survival in patients with resected lung ADCs. In addition, tumors with papillary, acinar, and even solid subtypes with Disco-p had a worse prognosis than those without it. Disco-p has superiority over TB in terms of simplicity of methodology and applicability for prognosis. These findings show that Disco-p should be recognized as a new invasive pattern and accurately recorded, as it may be useful for the management of patients with resected lung ADCs.