- Open Access
The relationship between primary colorectal cancer histology and the histopathological growth patterns of corresponding liver metastases
BMC Cancer volume 22, Article number: 911 (2022)
The histopathological growth patterns (HGPs) are a prognostic and predictive biomarker in colorectal cancer liver metastasis (CRLM). This study evaluates the relationship between the HGP and primary colorectal cancer (CRC) histopathology.
A total of 183 treatment-naive patients with resected CRC and CRLM were included. Thirteen CRC histopathology markers were determined and compared between the desmoplastic and non-desmoplastic HGP; tumour sidedness, pT&pN stage, tumour grade, tumour deposits, perineural- (lympho-)vascular- and extramural venous invasion, peritumoural budding, stroma type, CRC growth pattern, Crohn’s-like lymphoid reaction, and tumour-infiltrating lymphocyte (TIL) density. Logistic regression analysis was performed using both CRC and CRLM characteristics.
Unfavourable CRC histopathology was more frequent in non-desmoplastic CRLM for all markers evaluated, and significantly so for a lower TIL density, absent Crohn’s-like lymphoid reaction, and a “non-mature” stroma (all p < 0.03). The cumulative prevalence of unfavourable CRC histopathology was significantly higher in patients with non-desmoplastic compared to desmoplastic CRLM, with a median (IQR) of 4 (3–6) vs 2 (1–3.5) unfavourable characteristics observed, respectively (p < 0.001). Multivariable regression with 9 CRC histopathology markers and 2 CRLM characteristics achieved good discriminatory performance (AUC = 0.83).
The results of this study associates primary CRC histopathology with the HGP of corresponding liver metastases.
The management of colorectal cancer liver metastasis (CRLM) is clinically challenging and requires a multidisciplinary approach. This multidisciplinary need stems from the amenability of CRLM to local therapies such as surgical resection, ablation, and radiotherapy, which is dependent on hepatic tumour load and anatomical location, and the ability of systemic chemotherapy to act upon this through tumour load reduction . Although up to half of all patients can be treated with curative intent, cancer recurrence after surgical treatment of CRLM still occurs in over two-thirds, with long-term cure achieved in approximately one fourth [2,3,4,5,6]. This illustrates a demand for reliable and discriminatory markers to guide clinical decision making, preferably within the pre-treatment setting.
Histopathology studies of CRLM have led to the discovery of distinct histopathological growth patterns (HGP) formed at the interface of liver metastases and the liver parenchyma . A desmoplastic type is recognised in approximately one fifth of resected patients, characterised by the full encapsulation of all liver metastases by desmoplastic stroma (Fig. 1A) . Opposing is the non-desmoplastic type, which is primarily characterised by the complete or partial absence of tumour encapsulation, and secondarily by either invasion (Fig. 1B) or, rarely, compression (Fig. 1C) of the liver parenchyma . The clinical importance of this histopathology marker has been established in multiple cohorts, which reported 5-year overall survival rates of up to 80% for desmoplastic and as low as 40% for non-desmoplastic [8, 9], and have also suggested a benefit for adjuvant systemic chemotherapy for the treatment-naive non-desmoplastic patients only . Since perioperative systemic chemotherapy is considered standard of care in most countries, and current HGP assessment requires a CRLM resection specimen, predicting the HGP beforehand could help identify patients with favourable prognosis that do not require perioperative systemic chemotherapy, and could prevent unnecessary chemotherapy-associated morbidity.
As approximately half of all CRLM are metachronous, resection specimens of primary colorectal cancer are often available [2,3,4]. A possible clue to assess the HGP preoperatively might therefore lie in the primary CRC histopathology, especially given the number of available and established markers. In addition, associations could reveal underlying biological mechanisms of the distinct HGPs. This study therefore performs an exploratory analysis on the relationship between primary CRC histopathology and the HGP of corresponding CRLM.
Materials and methods
A single centre retrospective cohort study was conducted in patients treated surgically with curative intent for CRLM at the Erasmus MC Cancer Institute (Rotterdam, the Netherlands) between January 2000 and February 2019. Eligible patients were those who had had resection of their primary CRC at either the Erasmus MC Cancer Institute or a referring centre affiliated with one of four regional Dutch pathology laboratories (Erasmus MC, Bravis hospital, Maasstad Hospital, or Pathan). Patients who received any preoperative radio- or systemic chemotherapy prior to CRC or CRLM surgery were excluded, as preoperative treatment may alter both CRC histopathology and the HGPs of CRLM [11, 12]. In addition, patients with metachronous CRLM treated with adjuvant chemotherapy had to have had no systemic chemotherapy six months prior to CRLM diagnosis. Data on patient, CRLM, treatment characteristics, and overall survival (OS) was extracted from a prospectively maintained database. Institutional ethical review was obtained from the medical ethics committee of the Erasmus University Medical Center, Rotterdam, the Netherlands (MEC-2018–1743).
Colorectal liver metastasis HGP
Determination of the liver metastasis HGP was performed previously within the context of retrospective cohort studies [8, 9]. Assessment was at the time performed by at least two trained observers simultaneously on haematoxylin and eosin (H&E) stained tissue sections of resected CRLM, in accordance with international consensus guidelines, and blinded for all patient characteristics (including primary CRC) and survival . In summary, assessment entails the systematic evaluation of the entire tumour liver interface using light microscopy to determine the relative proportion of each of three distinct HGPs (Fig. 1). In line with the upcoming updated consensus guidelines the Rotterdam cut-off was applied and patients were classified as desmoplastic if all metastases exclusively displayed a desmoplastic pattern (i.e. 100% desmoplastic, Fig. 1A), and as non-desmoplastic otherwise (i.e. < 100% desmoplastic, Fig. 1B and C).
Primary CRC histopathology
For eligible patients all available H&E slides of resected CRC were requested from the respective pathology laboratories through the nationwide network and registry of histo- and cytopathology in the Netherlands (PALGA) . A literature study was conducted to identify CRC histopathology markers of interest, being those assessable on H&E stained slides of resected CRC, with clinical evidence suggesting a prognostic impact on (overall) survival following CRC resection, and with standardised guidelines and/or detailed methods of assessment. The literature study identified thirteen histopathology markers of interest, which were grouped in four categories. The classical markers comprised tumour sidedness, histologic grade, pT-stage, pN-stage, and tumour deposits. Under invasion markers were grouped lymphovascular invasion, extramural venous invasion, and perineural invasion. Amongst the tumour interface markers were peritumoural budding, CRC growth pattern, and fibrotic stroma type. Lastly, the immunological markers consisted of Crohn’s-like lymphoid reaction, and tumour-infiltrating lymphocyte (TIL) density.
A scoring manual was drafted outlining the assessment, definitions, and classifications with corresponding H&E examples for each (novel) individual marker identified (histologic grade, pT&pN-stage, and tumour sidedness were not described). This scoring manual was reviewed by two expert pathologists (PBV and MD) to reach a final consensus (supplementary file 1). A practice session was conducted using 64 digitalised H&E slides of resected CRC from 10 patients to reach agreement on the interpretation and application of the scoring manual. Hereafter the histopathology markers of interest were determined on all available H&E stained slides of included patients. Assessment was performed on a multi-head microscope by a gastro-intestinal pathologist (MD) and several PhD candidates, using the scoring manual as a reference, and blinded for patient characteristics, survival, and liver metastasis HGP. Scoring was done over the course of multiple (> 20) brief morning sessions (1–2 h) to prevent deterioration in assessment quality due to fatigue.
A right-sided tumour was defined as an anatomical CRC localisation proximal to the splenic flexure. The determination of histologic grade, pT&pN-stage, and tumour deposits was done in accordance with the 8th edition of the American Joint Committee on Cancer staging manual for CRC . The 8th edition defines tumour deposits as discrete tumour nodules found within the lymph drainage area of CRC and containing no identifiable lymph node tissue or vascular/neural structures (Fig. 2A).
Lymphovascular invasion was defined as the presence of tumour cells within a definite endothelial-lined space (lymphatics or blood vessel) (Fig. 2D) , extramural venous invasion as tumour invasion into large veins located in the subserosal or pericolic fat tissue (Fig. 2E) , and perineural (or intraneural) invasion as the presence of tumour cells inside the nerve sheath, or when at least one-third of the nerve circumference was encompassed by tumour cells (Fig. 2F) .
Tumour interface markers
Peritumoural budding was assessed in accordance with the 2016 International Tumor Budding Consensus Conference recommendations . Peritumoural buds, defined as a single tumour cell or a cluster of up to four tumour cells without gland formation (Fig. 2J), were counted in a 20 × magnification field at the invasive margin “hotspot” (field with the greatest density of buds in all available slides) and classified using a three-tier system; Grade I (low) for 0–4 buds, Grade II (intermediate) for 5–9 buds, and Grade III (high) for ≥ 10 buds . The CRC growth pattern was assessed according to Jass et al. and classified as either expanding or infiltrative based on a 50% predominance cut-off . The expanding type is characterised by a pushing or well-circumscribed margin (Fig. 2B), whereas the infiltrative type invades diffusely with widespread penetration of normal tissue (Fig. 2C). The fibrotic stroma type according to Ueno et al. classifies the stroma beyond the muscularis propria (at least pT3 stage) into three distinct types based on morphology; immature in case randomly oriented collagen bundles are surrounded by myxoid stroma (Fig. 2G), intermediate when broad bands of brightly eosinophilic hyalinised collagen (ropy-like) are intermingled with stroma (Fig. 2H), and mature for a stroma composed of multiple fine, mature, and stratiform fibres (Fig. 2I) .
Crohn’s-Like lymphoid reaction is characterised by lymphoid aggregates of at least 300 µm in diameter observed at the advancing edge of the tumour (Fig. 2K) . Crohn’s-like lymphoid reaction was considered present in case at least one aggregate > 300 µm was observed in any slide. Tumour-infiltrating lymphocyte density was assessed by estimating the percentage of mononuclear inflammatory cells over the total intratumoural stromal area at the invasive front (Fig. 2L) .
Statistical comparisons between patients with a non-desmoplastic and desmoplastic phenotype were performed to compare baseline patient, CRLM, and treatment characteristics, and to test for associations between individual CRC histopathology markers and the HGP of corresponding CRLM. Nominal variables were compared using the χ2 test and are reported as absolute counts with corresponding percentages. Non-parametric ordinal and numerical variables were compared using the Kruskall Wallis test and are reported as medians with corresponding interquartile ranges (IQR). The cumulative prevalence of unfavourable CRC histopathology was compared defined as the number of unfavourable characteristics observed per patient. For markers with more than two classes, a dichotomous classification was adapted, and for TIL density a percentage equal to or below the median was considered unfavourable. Uni- and multivariable binary logistic regression models were fitted with the HGP as dependent variable, and all CRC histopathology and any preoperatively available CRLM characteristics as candidate predictors. The prognostic impact of the HGP on OS following resection of CRLM was estimated by Kaplan–Meier survival analysis. Uni- and multivariable Cox regression analyses were additionally performed on OS following CRLM resection with all CRC histopathology and CRLM characteristics as candidate prognosticators. Given the large number of candidate predictors only those with a univariable p-value below 0.2 were entered into the multivariable models. Regression results are reported as multivariable odds ratios (OR) or hazard ratios (HR) with corresponding 95% confidence intervals (CI). Discriminatory capability of the multivariable logistic regression model to predict the HGP was assessed using the Area Under the Curve (AUC) metric of the receiver operating characteristic curve. The statistical significance level was set at a two-sided α of 0.05. All statistical analyses and data visualisation was performed using the R project for statistical computing version 4.1.1 (www.r-project.org), with packages rms (6.0–1), tableone (0.12.0), pROC (1.16.2), circlize (0.4.11) , and ggplot2 (3.3.2).
Primary CRC slides were requested for a total of 196 eligible patients through PALGA and were available for 183 (93%). A desmoplastic HGP was observed in 31 (17%) out of the 183 patients included for analysis. Baseline patient, CRLM, and treatment characteristics stratified by HGP are reported in Table 1. Patients with a non-desmoplastic HGP had a significantly larger CRLM diameter (median [IQR]: 3.2 [2.2, 4.1] vs 2.0 [1.3, 3.0] cm, p < 0.001), a significantly higher preoperative serum carcinoembryonic antigen level (median [IQR]: 12.0 [5.0, 44.7] vs 5.7 [3.2, 10.6] µg/L, p = 0.002), and more often had positive surgical margins upon CRLM resection (n = 14 [10%] vs n = 0 [0%], p = 0.07) (Table 1). Within the 119 patients with metachronous CRLM non-desmoplastic patients more often received adjuvant systemic chemotherapy (n = 46 [45%] vs n = 3 [19%], p = 0.05, Table 1).
Primary CRC histopathology
A total of 913 H&E slides of resected CRC were reviewed. The median number of tumour containing slides assessed per patient was 4 (IQR: 3–6) and did not differ between patients with corresponding non-desmoplastic (4 IQR [3-7]) and desmoplastic (4 IQR [3–5.5]) CRLM (p = 0.27). The great majority were adenocarcinomas (n = 179, 98%), with only 4 (2%) mucinous adenocarcinomas, which were equally distributed between non-desmoplastic (n = 3, 2%) and desmoplastic (n = 1, 3%) patients (p = 0.66). Comparisons of all primary CRC histopathology markers stratified by corresponding liver metastasis HGP is reported in Table 2 and Fig. 3A-C.
Poorly differentiated (G3) tumours (5% vs 0%), right-sided tumours (30% vs 16%), pT4-stage (16% vs 6%), positive lymph nodes (62% vs 48%), and tumour deposits (21% vs 10%) were more common in patients with corresponding non-desmoplastic versus desmoplastic CRLM, but none of these differences reached statistical significance (p-values of 0.22, 0.11, 0.15, and 0.16, respectively, Table 2).
Invasion, either lymphovascular (41% vs 23%), extramural venous (41% vs 29%), or perineural (26% vs 13%), was more prevalent in patients with non-desmoplastic versus desmoplastic CRLM, but none of these differences reached formal statistical significance (p-values of 0.06, 0.22, and 0.13, respectively, Table 2).
Tumour interface markers
Peritumoural budding (grade II/III vs I: 20% vs 10%) and an infiltrative CRC growth pattern (53% vs 42%) were also more common for patients with corresponding non-desmoplastic versus desmoplastic metastases, but again these differences did not reach statistical significance (p = 0.18 and p = 0.28 respectively, Table 2). Of the patients with non-desmoplastic CRLM, 18 (13%) had an immature, 33 (24%) an intermediate, and 87 (63%) a mature fibrotic stroma type, whereas this was 2 (7%), 1 (3%), and 26 (90%) respectively for patients with desmoplastic CRLM, a difference that was statistically significant (p = 0.02, Table 2). Consequently, a non-mature (i.e., immature or intermediate) stroma was significantly more often observed in non-desmoplastic compared to desmoplastic patients (34% vs 10%, p = 0.008, Table 2).
Crohn’s-like lymphoid reaction was observed in 131 (86%) of the patients with non-desmoplastic CRLM versus in all 31 (100%) of the patients with desmoplastic CRLM (p = 0.03, Table 2). The TIL density (median [IQR]) was significantly lower for the non-desmoplastic (10% [5%-15%]) versus desmoplastic (15% [10%-20%]) patients (p = 0.02, Fig. 3B). Consequently, a median-to-low (≤ 10%) TIL density was significantly more common in the non-desmoplastic (n = 105 [69%]) compared to desmoplastic (n = 13 [42%]) group (p = 0.004, Table 2).
Cumulative prevalence and HGP prediction
Overall, unfavourable CRC histopathology was significantly more prevalent in non-desmoplastic compared to desmoplastic patients with a median (IQR) of 4 (3–6) versus 2 (1–3.5) unfavourable features observed, respectively (p < 0.001, Fig. 3C). The results of the uni- and multivariable binary logistic regression analyses to predict the HGP are reported in Table 3. Differentiation grade and Crohn’s-like lymphoid reaction could not be analysed using logistic regression given absent cases in the desmoplastic group. Upon univariable analysis the CRLM characteristics disease-free interval and diameter, and all primary CRC histopathology features except extramural vascular invasion and CRC growth pattern had a p-value below 0.2 and were considered for multivariable analysis (Table 3). Of all 11 predictors in the multivariable model, only the diameter of the largest CRLM and TIL density proved independent predictors for a desmoplastic HGP, with an OR (95%CI) of 0.55 (0.37–0.82) for each additional cm and 1.60 (1.01–2.54) per 10% increase in TIL density, respectively (Table 3). The multivariable logistic regression model achieved an AUC of 0.83 to predict the HGP.
Patients with a desmoplastic HGP had a significantly longer OS following resection of CRLM with an estimated 5-year (95%CI) survival of 77% (64–94%) compared to 39% (31–48%) for non-desmoplastic (p = 0.003, Fig. 4). Univariable OS regression analysis revealed four CRLM characteristics (age at resection, number of CRLM, extrahepatic disease, and the HGP) and nine CRC histopathology features (right-sided, differentiation grade, pT4-stage, positive lymph nodes, tumour deposits, [lympho]-vascular, extramural vascular and perineural invasion, and the CRC growth pattern) with a p-value below 0.2 and were considered for multivariable analysis (supplementary file 2). Of these, only age at resection, the HGP, and a right-sided tumour proved independent predictors for survival, with adjusted HRs (95%CI) of 1.33 [1.10–1.62] per 10-year age increase, 1.97 [1.10–3.53] for a non-desmoplastic HGP, and 1.86 [1.23–2.83] for right-sided tumours, respectively (supplementary file 2).
The present study evaluated the relationship between thirteen established CRC histopathology markers and the HGPs of corresponding CRLM in a cohort of 183 resected patients. For all markers, unfavourable CRC histopathology was more frequent for patients with corresponding non-desmoplastic CRLM. While many of these individual marker differences did not reach statistical significance, the cumulative prevalence of unfavourable CRC histopathology was significantly higher in the non-desmoplastic patients.
At least two other studies have previously evaluated CRC histopathology in relation to the CRLM growth pattern phenotype. The more recent study by Wu et al. evaluated 29 patients and compared primary CRC histopathology between 15 patients with a predominant (i.e. > 50%) desmoplastic versus 14 with a predominant replacement pattern . The study significantly associated the predominant replacement group (i.e. non-desmoplastic) with higher peritumoural budding grades, an infiltrative CRC growth pattern, and absent Crohn’s disease-like response. Rajaganeshan and colleagues evaluated in 55 patients the relationships between primary CRC growth pattern and CRLM encapsulation, the latter defined as > 50% fibrous capsule formation separating tumour from stroma (i.e. > 50% desmoplastic), and also significantly associated an infiltrative CRC growth pattern with corresponding non-encapsulated (i.e. non-desmoplastic) CRLM . An important distinction with these studies lies in the classification of the HGP, as both applied a 50% predominance cut-off as opposed to the newly recommended Rotterdam criteria of entirely desmoplastic versus otherwise. There is compelling evidence from both a prognostic [8, 9] and immunologic  standpoint that it is this distinction between desmoplastic and non-desmoplastic that delineates clinical relevance. Results of studies applying predominance cut-offs are therefore difficult to extrapolate in light of this new classification, as the predominant desmoplastic groups (i.e. > 50%) are by definition, and based on previous studies on HGP distribution [8, 9], actually for more than half composed of non-desmoplastic cases. Nevertheless, taking all current evidence as a whole, non-desmoplastic CRLM have repeatedly been associated with unfavourable CRC histopathology, something also evident from the relationship with lymph node positivity observed in multiple large cohort studies evaluating the HGPs [8, 9]. The current study confirms these associations and adds further compelling evidence of the relationship between CRC histopathology and the HGPs of corresponding CRLM.
While not all individual marker differences demonstrated a statistically significant association, both the immunology markers TIL-density and Crohn’s-like lymphoid reaction did, with higher TIL-densities and increased Crohn’s-like lymphoid reaction observed in the patients with desmoplastic CRLM. Both markers have been associated with a survival benefit after resection of primary CRC and are thought to reflect anti-tumour (host) immunity, with increased TIL-densities and Crohn’s-like reaction indicative of a more effective antitumour host-response . The cellular composition and structure of these Crohn’s-like lymphoid aggregates is similar to secondary lymphoid organs, and studies have linked these structures with increased TIL infiltration  and cytotoxic gene expression signatures , indicating that these lymphoid aggregates are functional components of the adaptive anticancer immune response in CRC . These results therefore suggest an increased adaptive immune response in the originating primary colorectal cancers of patients who develop corresponding desmoplastic liver metastasis. Evaluations of the immune microenvironment of CRLM have revealed similar results, that is an increased and distinctly cytotoxic immune response observed in the desmoplastic HGP [26, 30]. This now associates the desmoplastic phenotype with increased antitumor immunity in both the originating primary colorectal tumour, as well as the localised liver metastasis microenvironment, hinting at a degree of systemic anti-tumour response in these patients. Taken together with recent associations between microsatellite instability-high colorectal cancers, an actionable target for immunotherapy in stage I-III  and IV  CRC, and desmoplastic CRLM , there is growing evidence to suggest that (systemic) anticancer immunity plays an important role in the underlying biology of the HGPs. While this infers causality, the fact that patient-derived xenografts in SCID-beige mouse with defective T- B- and NK-cell activity have been successful in producing liver lesions with an identical HGP as the donor patient metastasis following intrahepatic transplantation however argues against the HGPs as a solely immunologically driven process . In addition, liver metastases appear to suppress systemic immunity in general, with immunotherapies appearing less effective in the presence of hepatic dissemination . Whether the HGP of a liver metastasis influences the degree of systemic immunosuppression remains to be explored.
Of the other markers evaluated the extramural fibrotic stroma type was also significantly associated with the corresponding liver metastasis HGP. The prognostic impact of the extramural stroma type after resection of primary CRC has been demonstrated in multiple retrospective series, and more recently within a prospective phase III trial [20, 35, 36]. In addition, characterisation of the extramural stroma type of primary CRC also proved prognostic for survival following resection of corresponding CRLM . Although this could not be validated in the current study, which found a univariable HR (95%CI) of 1.06 (0.72–1.57) for a non-mature stroma type (supplementary file 2). Of the three types recognized, the immature type has the worst prognosis, followed by the intermediate type, and with the most favourable prognosis observed in the mature type. There are several arguments to indicate that the non-mature stroma types (i.e., immature and intermediate) reflect a state of activated epithelial-mesenchymal transition (EMT) promoting invasive and migratory cancer properties. Both have for instance been associated with higher degrees of tumour budding , a known phenotype of EMT-related gene expression , which was also true in this study cohort (data not shown). The non-mature stroma’s, and notably the characterising myxoid stroma of the immature type, also exhibit increased extracellular matrix component depositions amongst which fibronectin , a known activator of EMT . In addition, the defining eosinophilic collagen bundles of the intermediate type are similarly observed in keloids, a microenvironment characterised by overexpression of fibroblast associated growth factors including transforming growth factor β (TGF-β) . Taking the position that these are alike, TGF-β, a well-recognised EMT stimulating factor, is likely to be upregulated in the intermediate type . The association between non-desmoplastic CRLM and these non-mature stroma types suggests increased EMT activation in the primary tumours of these patients. Indeed, other histomorphological signs of invasive and migratory growth potential such as vessel invasion and peritumoural budding were also exclusively more frequent in the primary tumours of corresponding non-desmoplastic metastases, albeit not statistically significantly so.
Besides the association with individual histopathology markers, this study found that overall, unfavourable CRC histopathology was associated with non-desmoplastic CRLM given the significantly higher cumulative prevalence of unfavourable charactheristics observed in these patients. This suggests that the information contained in primary CRC histology may be exploited to predict the HGP. And indeed, a multivariable model containing 9 primary CRC characteristics and 2 CRLM characteristics achieved good performance (AUC = 0.83) to predict the HGP. Of all CRC markers included only TIL density however proved an independent predictor, with all other characteristics failing to reach statistical significance. When interpreting these results it is important to consider that almost all markers had an estimated odds-ratio around the 0.5 mark, but were insignificant as a result of a large uncertainty of this estimate, i.e. wide confidence intervals. The model therefore predominantly demonstrates that prediction of the HGP using both CRLM and CRC histopathology characteristics could be feasible, but that the current sample-size is insufficient to properly assess the individual predictive properties of all included markers. Something also highlighted by the fact that not all markers could be included in these regression analyses given absent cases in the desmoplastic group. As such, the results of this study should serve more as a stepping stone for a deep-learning digital-pathology approach in a larger cohort. Several deep-learning models already exist for the automated detection and classification of individual markers, for instance peritumoural budding , TIL density , and the fibrotic stroma type . Additionally, studies have shown deep-learning on histopathology capable of predicting relevant outcomes in a hypothesis-free manner, i.e., not training the model to predict specific markers but instead let the model identify relevant features itself for accurate prediction of the outcome of interest. Examples are the prediction of survival after resection of both primary CRC  and CRLM , and such a hypothesis-free approach could also be considered to predict the HGP of CRLM on digitalised slides of the corresponding primary CRC tumour. Such a study would require a large number of digitalised slides of resected CRC and corresponding CRLM from multiple independent cohorts. Collection of such datasets may therefore be worthwhile to pursue.
The survival analysis in light of the HGP, other CRLM and patient characteristics, and all included CRC histopathology markers demonstrated the HGP as one of three independent predictors for overall survival upon multivariable analysis. But again, the lack of statistical power was evident, as multiple markers – Including established prognosticators such as lymph-node positivity – demonstrated clinically relevant estimates but failed to reach statistical significance based on the estimate uncertainty (supplementary file 2). Nevertheless, these results support the clinical relevance of this biomarker, similarly to the two previous retrospective series on which this cohort is partly based [8, 9]. In addition, a recent study of over 4000 patients evaluating survival after CRLM surgery in light of new biomarkers including the HGP found it to be amongst the independent prognosticators with the largest impact on survival, only being equalled by KRAS and BRAF mutational status, respectively .
The results of this study have to be considered in light of its limitations. Most importantly the inadequate sample-size to detect small to moderate individual marker differences, increasing the likelihood of type 2 statistical errors and not allowing for sufficiently powered multivariable regression analysis increasing the risk for overfitted models. For example, as a general rule of thumb it is advised that the number of variables in a multivariable regression model should not exceed 10% of the total number of events. In our study this relates to the 31 patients with a desmoplastic HGP, and therefore the multivariable model should preferably be limited to less than 4 predictors instead of the 11 actually included. The results should therefore be interpreted with caution and serve more as a proof of concept for a future validation or follow-up study. In addition, this study only included patients who did not receive any chemo- or radiotherapy prior to both CRC and CRLM surgery. While this is from an analysis standpoint not a limitation per se, the applicability of the results is lessened as most patients who undergo surgical resection of CRLM are generally treated with preoperative systemic chemotherapy which can alter the HGP, with higher rates of the desmoplastic HGP found after chemotherapy . Interestingly, within the 119 patients with metachronous CRLM, those who received adjuvant systemic chemotherapy following CRC resection more often had a non-desmoplastic HGP. This is likely the result of confounding by indication, as both node-positivity and pT4 stage were more common in the patients with a non-desmoplastic HGP. Moreover, all patients in the current study did not receive any chemotherapy in the six months prior to CRLM diagnosis. Another limitation is the lack of data on molecular characteristics such as KRAS and BRAF mutational status. Unfortunately, many of the patients in this study were operated on before the implementation of these genetic markers into routine clinical practice, and data on these markers was consequently only available for less than one-fifth. Previous studies however did not find an association between these markers and the HGP of CRLM , and found the prognostic impact independent of these genetic alterations [6, 9], but in-depth genetic association studies remain lacking. Lastly, all histopathological assessment was observer-based, while an increasing number of automated assessments are available for more reproducible and precise estimations. These limitations underscore the need for external validation not limited to treatment-naive patients, and ideally with observer-independent methods of assessment.
In conclusion, our results associate primary colorectal cancer histopathology with the histopathological growth patterns of corresponding colorectal liver metastases, and may aid in their preoperative determination. In addition, it associates the desmoplastic phenotype with an increased host-immune response, and the non-desmoplastic type with histomorphological evidence of epithelial-mesenchymal transition at the primary tumour microenvironment level.
Availability of data and materials
All data and digital slides are available upon request and at the discretion of the corresponding author.
Primrose JN. Surgery for colorectal liver metastases. Br J Cancer. 2010;102(9):1313–8.
Elferink MA, de Jong KP, Klaase JM, Siemerink EJ, de Wilt JH. Metachronous metastases from colorectal cancer: a population-based study in North-East Netherlands. Int J Colorectal Dis. 2015;30(2):205–12.
van der Geest LG, Lam-Boer J, Koopman M, Verhoef C, Elferink MA, de Wilt JH. Nationwide trends in incidence, treatment and survival of colorectal cancer patients with synchronous metastases. Clin Exp Metastasis. 2015;32(5):457–65.
Manfredi S, Lepage C, Hatem C, Coatmeur O, Faivre J, Bouvier AM. Epidemiology and management of liver metastases from colorectal cancer. Ann Surg. 2006;244(2):254–9.
Tomlinson JS, Jarnagin WR, DeMatteo RP, Fong Y, Kornprat P, Gonen M, Kemeny N, Brennan MF, Blumgart LH, D’Angelica M. Actual 10-year survival after resection of colorectal liver metastases defines cure. J Clin Oncol. 2007;25(29):4575–80.
Buisman FE, Giardiello D, Kemeny NE, Steyerberg EW, Höppener DJ, Galjart B, Nierop PMH, Balachandran VP, Cercek A, Drebin JA, et al. Predicting 10-year survival after resection of colorectal liver metastases; an international study including biomarkers and perioperative treatment. Eur J Cancer. 2022;168:25–33.
van Dam PJ, van der Stok EP, Teuwen LA, Van den Eynden GG, Illemann M, Frentzas S, Majeed AW, Eefsen RL, Coebergh van den Braak RRJ, Lazaris A, et al. International consensus guidelines for scoring the histopathological growth patterns of liver metastasis. Br J Cancer. 2017;117(10):1427–41.
Galjart B, Nierop PMH, van der Stok EP, van den Braak R, Hoppener DJ, Daelemans S, Dirix LY, Verhoef C, Vermeulen PB, Grunhagen DJ. Angiogenic desmoplastic histopathological growth pattern as a prognostic marker of good outcome in patients with colorectal liver metastases. Angiogenesis. 2019;22(2):355–68.
Hoppener DJ, Galjart B, Nierop PMH, Buisman FE, van der Stok EP, Coebergh van den Braak RRJ, van Amerongen MJ, Balachandran VP, Jarnagin WR, Kingham TP, et al. Histopathological Growth Patterns and Survival After Resection of Colorectal Liver Metastasis: An External Validation Study. JNCI Cancer Spectr. 2021;5(3):pkab026.
Buisman FE, Van der Stok EP, Galjart B, Vermeulen PB, Balachandran VP, Coebergh van den Braak RRJ, Creasy JM, Hoppener DJ, Jarnagin WR, Kingham TP et al: Histopathological growth patterns as biomarker for adjuvant systemic chemotherapy in patients with resected colorectal liver metastases. Clin Exp Metastasis. 2020;37(5):593-605.
Nagtegaal ID, Glynne-Jones R. How to measure tumour response in rectal cancer? An explanation of discrepancies and suggestions for improvement. Cancer Treat Rev. 2020;84:101964.
Nierop PM, Hoppener DJ, Buisman FE, van der Stok EP, Galjart B, Balachandran VP, Jarnagin WR, Kingham TP, Shia J, Mauer M, et al. Preoperative systemic chemotherapy alters the histopathological growth patterns of colorectal liver metastases. J Pathol Clin Res. 2022;8(1):48–64.
Casparie M, Tiebosch ATMG, Burger G, Blauwgeers H, van de Pol A, van Krieken JHJM, Meijer GA. Pathology databanking and biobanking in The Netherlands, a central role for PALGA, the nationwide histopathology and cytopathology data network and archive. Anal Cell Pathol. 2007;29(1):19–24.
Weiser MR. AJCC 8th Edition: Colorectal Cancer. Ann Surg Oncol. 2018;25(6):1454–5.
Compton CC, Fielding LP, Burgart LJ, Conley B, Cooper HS, Hamilton SR, Hammond ME, Henson DE, Hutter RV, Nagle RB, et al. Prognostic factors in colorectal cancer. College of American Pathologists Consensus Statement 1999. Arch Pathol Lab Med. 2000;124(7):979–94.
Leijssen LGJ, Dinaux AM, Amri R, Taylor MS, Deshpande V, Bordeianou LG, Kunitake H, Berger DL. Impact of intramural and extramural vascular invasion on stage II-III colon cancer outcomes. J Surg Oncol. 2019;119(6):749–57.
Liebig C, Ayala G, Wilks JA, Berger DH, Albo D. Perineural invasion in cancer: a review of the literature. Cancer. 2009;115(15):3379–91.
Lugli A, Kirsch R, Ajioka Y, Bosman F, Cathomas G, Dawson H, El Zimaity H, Flejou JF, Hansen TP, Hartmann A, et al. Recommendations for reporting tumor budding in colorectal cancer based on the International Tumor Budding Consensus Conference (ITBCC) 2016. Mod Pathol. 2017;30(9):1299–311.
Jass JR, Love SB, Northover JM. A new prognostic classification of rectal cancer. Lancet. 1987;1(8545):1303–6.
Ueno H, Ishiguro M, Nakatani E, Ishikawa T, Uetake H, Murotani K, Matsui S, Teramukai S, Sugai T, Ajioka Y, et al. Prognostic value of desmoplastic reaction characterisation in stage II colon cancer: prospective validation in a Phase 3 study (SACURA Trial). Br J Cancer. 2021;124(6):1088–97.
Ueno H, Hashiguchi Y, Shimazaki H, Shinto E, Kajiwara Y, Nakanishi K, Kato K, Maekawa K, Miyai K, Nakamura T, et al. Objective criteria for crohn-like lymphoid reaction in colorectal cancer. Am J Clin Pathol. 2013;139(4):434–41.
Iseki Y, Shibutani M, Maeda K, Nagahara H, Fukuoka T, Matsutani S, Kashiwagi S, Tanaka H, Hirakawa K, Ohira M. A new method for evaluating tumor-infiltrating lymphocytes (TILs) in colorectal cancer using hematoxylin and eosin (H-E)-stained tumor sections. PLoS ONE. 2018;13(4):e0192744.
Gu Z, Gu L, Eils R, Schlesner M, Brors B. circlize implements and enhances circular visualization in R. Bioinformatics. 2014;30(19):2811–2.
Wu JB, Sarmiento AL, Fiset PO, Lazaris A, Metrakos P, Petrillo S, Gao ZH. Histologic features and genomic alterations of primary colorectal adenocarcinoma predict growth patterns of liver metastasis. World J Gastroenterol. 2019;25(26):3408–25.
Rajaganeshan R, Prasad R, Guillou PJ, Chalmers CR, Scott N, Sarkar R, Poston G, Jayne DG. The influence of invasive growth pattern and microvessel density on prognosis in colorectal cancer and colorectal liver metastases. Br J Cancer. 2007;96(7):1112–7.
Hoppener DJ, Nierop PMH, Hof J, Sideras K, Zhou G, Visser L, Gouw ASH, de Jong KP, Sprengers D, Kwekkeboom J, et al. Enrichment of the tumour immune microenvironment in patients with desmoplastic colorectal liver metastasis. Br J Cancer. 2020;123(2):196–206.
Rozek LS, Schmit SL, Greenson JK, Tomsho LP, Rennert HS, Rennert G, Gruber SB: Tumor-Infiltrating Lymphocytes, Crohn’s-Like Lymphoid Reaction, and Survival From Colorectal Cancer. JNCI. 2016;108(8):djw027.
Coppola D, Nebozhyn M, Khalil F, Dai H, Yeatman T, Loboda A, Mulé JJ. Unique ectopic lymph node-like structures present in human primary colorectal carcinoma are identified by immune gene array profiling. Am J Pathol. 2011;179(1):37–45.
Maoz A, Dennis M, Greenson JK. The Crohn’s-Like Lymphoid Reaction to Colorectal Cancer-Tertiary Lymphoid Structures With Immunologic and Potentially Therapeutic Relevance in Colorectal Cancer. Front Immunol. 1884;2019:10.
Brunner SM, Kesselring R, Rubner C, Martin M, Jeiter T, Boerner T, Ruemmele P, Schlitt HJ, Fichtner-Feigl S. Prognosis according to histochemical analysis of liver metastases removed at liver resection. Br J Surg. 2014;101(13):1681–91.
Chalabi M, Fanchi LF, Dijkstra KK, Van den Berg JG, Aalbers AG, Sikorska K, Lopez-Yurda M, Grootscholten C, Beets GL, Snaebjornsson P, et al. Neoadjuvant immunotherapy leads to pathological responses in MMR-proficient and MMR-deficient early-stage colon cancers. Nat Med. 2020;26(4):566–76.
André T, Shiu K-K, Kim TW, Jensen BV, Jensen LH, Punt C, Smith D, Garcia-Carbonero R, Benavides M, Gibbs P, et al. Pembrolizumab in Microsatellite-Instability–High Advanced Colorectal Cancer. N Engl J Med. 2020;383(23):2207–18.
Tabaries S, Annis MG, Lazaris A, Petrillo SK, Huxham J, Abdellatif A, Palmieri V, Chabot J, Johnson RM, Van Laere S, et al. Claudin-2 promotes colorectal cancer liver metastasis and is a biomarker of the replacement type growth pattern. Commun Biol. 2021;4(1):657.
Lee JC, Green MD, Huppert LA, Chow C, Pierce RH, Daud AI. The Liver-Immunity Nexus and Cancer Immunotherapy. Clin Cancer Res. 2022;28(1):5–12.
Ueno H, Shinto E, Shimazaki H, Kajiwara Y, Sueyama T, Yamamoto J, Hase K. Histologic Categorization of Desmoplastic Reaction: Its Relevance to the Colorectal Cancer Microenvironment and Prognosis. Ann Surg Oncol. 2015;22(5):1504–12.
Ueno H, Kanemitsu Y, Sekine S, Ishiguro M, Ito E, Hashiguchi Y, Kondo F, Shimazaki H, Kajiwara Y, Okamoto K, et al. A Multicenter Study of the Prognostic Value of Desmoplastic Reaction Categorization in Stage II Colorectal Cancer. Am J Surg Pathol. 2019;43(8):1015–22.
Ueno H, Konishi T, Ishikawa Y, Shimazaki H, Ueno M, Aosasa S, Saiura A, Hase K, Yamamoto J: Histologic categorization of fibrotic cancer stroma in the primary tumor is an independent prognostic index in resectable colorectal liver metastasis. Am J Surg Pathol 2014;38(10):1380-6.
Ueno H, Kanemitsu Y, Sekine S, Ishiguro M, Ito E, Hashiguchi Y, Kondo F, Shimazaki H, Mochizuki S, Kajiwara Y, et al. Desmoplastic Pattern at the Tumor Front Defines Poor-prognosis Subtypes of Colorectal Cancer. Am J Surg Pathol. 2017;41(11):1506–12.
De Smedt L, Palmans S, Andel D, Govaere O, Boeckx B, Smeets D, Galle E, Wouters J, Barras D, Suffiotti M. Expression profiling of budding cells in colorectal cancer reveals an EMT-like phenotype and molecular subtype switching. Br J Cancer. 2017;116(1):58–65.
Tzanakakis G, Kavasi RM, Voudouri K, Berdiaki A, Spyridaki I, Tsatsakis A, Nikitovic D. Role of the extracellular matrix in cancer-associated epithelial to mesenchymal transition phenomenon. Dev Dyn. 2018;247(3):368–81.
Bran GM, Goessler UR, Hormann K, Riedel F, Sadick H. Keloids: current concepts of pathogenesis (review). Int J Mol Med. 2009;24(3):283–93.
Lamouille S, Xu J, Derynck R. Molecular mechanisms of epithelial–mesenchymal transition. Nat Rev Mol Cell Biol. 2014;15(3):178–96.
Fisher NC, Loughrey MB, Coleman HG, Gelbard MD, Bankhead P, Dunne PD. Development of a semi-automated method for tumor budding assessment in colorectal cancer and comparison with manual methods. bioRxiv. 2021:2021.2006.2017.448482. https://doi.org/10.1101/2021.06.17.448482. PMID:34580909.
Saltz J, Gupta R, Hou L, Kurc T, Singh P, Nguyen V, Samaras D, Shroyer KR, Zhao T, Batiste R, et al. Spatial Organization and Molecular Correlation of Tumor-Infiltrating Lymphocytes Using Deep Learning on Pathology Images. Cell Rep. 2018;23(1):181-193.e187.
Nearchou IP, Ueno H, Kajiwara Y, Lillard K, Mochizuki S, Takeuchi K, Harrison DJ, Caie PD. Automated Detection and Classification of Desmoplastic Reaction at the Colorectal Tumour Front Using Deep Learning. Cancers. 2021;13(7):1615.
Skrede O-J, De Raedt S, Kleppe A, Hveem TS, Liestøl K, Maddison J, Askautrud HA, Pradhan M, Nesheim JA, Albregtsen F. Deep learning for prediction of colorectal cancer outcome: a discovery and validation study. The Lancet. 2020;395(10221):350–60.
Tellez D, Höppener D, Verhoef C, Grünhagen D, Nierop P, Drozdzal M, Laak J, Ciompi F: Extending unsupervised neural image compression with supervised multitask learning. In: Medical Imaging with Deep Learning: 2020. PMLR; 2020. p. 770–783. https://arxiv.org/abs/2004.07041.
The authors would like to express their gratitude to Stichting Coolsingel (Rotterdam, the Netherlands) for providing part of the funding of this study.
PALGA Group: The following departments contributed to the sample collection of this study:
-Department of Pathology, Erasmus MC, Rotterdam, the Netherlands
-Pathan BV, Rotterdam, the Netherlands
-Department of Pathology, Maasstad Hospital, Rotterdam, the Netherlands
-Department of Pathology, Bravis Hospital, Bergen op Zoom, the Netherlands
This study was partly funded by a grant from Stichting Coolsingel, Rotterdam, the Netherlands.
Ethics approval and consent to participate
Institutional ethical review was obtained from the medical ethics committee of the Erasmus University Medical Center, Rotterdam, the Netherlands (MEC-2018–1743), which granted a waiver for (renewed) informed consent given the retrospective nature of the study. The study was performed in accordance with the Declaration of Helsinki.
Consent for publication
None to declare.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Höppener, D.J., Stook, JL.P.L., Galjart, B. et al. The relationship between primary colorectal cancer histology and the histopathological growth patterns of corresponding liver metastases. BMC Cancer 22, 911 (2022). https://doi.org/10.1186/s12885-022-09994-3
- Histopathological growth patterns
- Colorectal cancer
- Colorectal liver metastasis
- TNM classification
- Tumour budding
- Desmoplastic reaction
- Tumour infiltrating lymphocytes
- Crohn’s-like lymphoid reaction