ORIGINAL ARTICLE

Assessment of Tumor-infiltrating Lymphocytes Using

International TILs Working Group (ITWG) System
Is a Strong Predictor of Overall Survival
in Colorectal Carcinoma
A Study of 1034 Patients
Talia L. Fuchs, MD,*†‡ Loretta Sioson, BSc,* Amy Sheen, BSc,*
Kambin Jafari-Nejad, MD, FRCPA,§ Christopher J. Renaud, MBBS,*†
Juliana Andrici, MBBS, PhD,∥ Mahsa Ahadi, MD, FRCPA,*†‡
Angela Chou, PhD, FRCPA,*†‡ and Anthony J. Gill, MD, FRCPA*†‡

CRCs (P = 0.0001), and BRAF wild-type CRCs (P = 0.001). The

Abstract: The presence of increased tumor-infiltrating lymphocytes predictive value of TILs assessed using the ITWG system was supe-
(TILs) is established as a positive prognostic factor in many malig- rior to the assessment of intraepithelial lymphocyte performed pro-
nancies including colorectal carcinoma (CRC). However, multiple spectively using a standard system requiring ≥ 5 lymphocytes per
different approaches have been used to assess TILs. In 2014, the high-powered field in direct contact with tumor cells or between tu-
International TILs Working Group (ITWG) proposed a standardized mor clusters. We conclude that the ITWG system for assessing TILs is
methodology for evaluating TILs, initially in the context of breast a powerful predictor of all-cause survival in CRC independent of
cancer, but subsequently expanded to other malignancies. To date, the many prognostic factors and superior to the assessment of intra-
efficacy of the ITWG system has not been investigated in a large epithelial lymphocytes using a traditional system.
cohort of all-stage CRC. We, therefore, sought to validate this system
in CRC. We used the ITWG system to assess the density of stromal Key Words: tumor-infiltrating lymphocytes (TILs), colorectal
TILs in an unselected cohort of 1034 CRC patients undergoing pri- carcinoma (CRC), International TILs Working Group (ITWG)
mary tumor resection at our institution. The percentage TILs’ score (Am J Surg Pathol 2020;44:536–544)
was categorized into 3 groups: low (0% to 10%), intermediate (15% to
50%), and high (55% to 100%). The mean survival was 53, 67, and
75 months, respectively (P = 0.0001). This survival benefit remained
statistically significant in multivariate analyses (P = 0.0001) and sub-
group analyses of mismatch repair–proficient CRCs (P = 0.0001),
mismatch repair–deficient CRCs (P = 0.031), BRAFV600E-mutant
A ssessment of the tumor microenvironment has grown
in importance over the past decade, as strong evidence
has emerged showing a significant correlation with patient
survival in a number of different malignancies.1–3 In ad-
dition, the emergence of immunotherapeutic agents has
From the *Cancer Diagnosis and Pathology Group, Kolling Institute prompted a reassessment of the tumor microenvironment
of Medical Research, Royal North Shore Hospital, St Leonards;
†Department of Anatomical Pathology, NSW Health Pathology, as a potential biomarker of response to these agents.4 The
Royal North Shore Hospital; ‡Sydney Medical School, University of presence of increased tumor-infiltrating lymphocytes
Sydney, Sydney; §Douglas Hanly Moir Pathology, Macquarie Park, (TILs), a marker of immune activation by a tumor, has
NSW, Australia; and ∥Institute of Pathology, University of Bern, consistently been proven to be an independent predictor of
Bern, Switzerland.
Conflicts of Interest and Source of Funding: The authors have disclosed
better prognosis across a range of malignancies, including
that they have no significant relationships with, or financial interest colorectal carcinoma (CRC).1–3
in, any commercial companies pertaining to this article. Although the prognostic significance of TILs in CRC
Correspondence: Anthony J. Gill, MD, FRCPA, Department of has been recognized for several decades, early studies mostly
Anatomical Pathology, NSW Health Pathology, Royal North Shore concentrated on the association with microsatellite unstable
Hospital, Pacific Highway, St Leonards, NSW 2065, Australia
(e-mail: [email protected]). (MSI) tumors, often as a marker to identify which patients
Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, should undergo further screening for Lynch syndrome.5
Inc. This is an open-access article distributed under the terms of the Recently, attention has shifted to the prognostic value of
Creative Commons Attribution-Non Commercial-No Derivatives TILs in all CRCs, regardless of mismatch repair (MMR)
License 4.0 (CCBY-NC-ND), where it is permissible to download and
share the work provided it is properly cited. The work cannot be
status. However, the data is significantly limited by both
changed in any way or used commercially without permission from nonuniform scoring methodologies and inconsistent applica-
the journal. tion in clinical practice.2,5

536 | www.ajsp.com Am J Surg Pathol  Volume 44, Number 4, April 2020

Am J Surg Pathol  Volume 44, Number 4, April 2020 TILs in CRC

Several different systems are currently in widespread MATERIALS AND METHODS

use to assess TILs. One widely used system is the Klintrup-
Makinen Score, wherein the immune infiltrate is scored Patients
from 0 to 3, where score 0 = no increase in inflammatory We searched the computerized database of the De-
cells; score 1 = a patchy increase of inflammatory cells at partment of Anatomical Pathology Royal North Shore
the invasive margin, but no destruction of invading cancer Hospital, Sydney Australia, for all patients undergoing
cell islets (tumor cell nests or glandular structures); score surgical resection for colorectal adenocarcinoma during the
2 = a band-like infiltrate at the invasive margin with some calendar years 2012 to 2015. Exclusion criteria included
destruction of cancer cell islets; and score 3 = a very extracolonic and appendiceal location, tumors undergoing
prominent inflammatory reaction, forming a cup-like zone biopsy alone or treated endoluminal only, and histologic
at the invasive margin, and frequent and invariable type other than adenocarcinoma, as defined by the World
destruction of cancer cell islets.6 In Australia, the Royal Health Organization 2019 system. When multiple tumors
College of Pathologists of Australasia (RCPA) defines were present, the highest stage tumor was assessed. For the
intraepithelial lymphocytes (IELs) as those that are in purpose of binary analysis in this study, the right colon was
direct contact with tumor cells or are located directly be- defined as cecum, ascending colon, hepatic flexure, and
tween tumor cell clusters.7 A high density of lymphocytes, transverse colon, whereas the left colon was defined as
defined as ≥ 5/HPF), is considered significant. However, splenic flexure, descending colon, sigmoid, and rectum. At
it is advised that the IEL counts are not necessary if the time of diagnosis, all patients had routine assessment of
mismatch repair deficiency (MMRd) or MSI is to be for- MMR status by immunohistochemistry with a 4-marker
mally assessed. Although previous versions of the College panel (PMS2, MLH1, MSH6, and MSH2), and the as-
of American Pathologists (CAP) protocol for reporting sessment of BRAFV600E mutation status using mutation-
CRC have recommended assessment of TIL/IELs using specific immunohistochemistry, which we have demon-
a semiquantitative system (none; mild to moderate [0 to strated to be consistently reliable in our hands.10,11 At the
2/HPF]; and marked [ ≥ 3/HPF]), this was primarily used time of primary reporting, the number of IELs was as-
to assist screening for Lynch syndrome, and, in the era sessed using the RCPA-endorsed system, as outlined
of universal MMR/MSI screening, the current CAP pro- above,7 as was the presence or absence of a Crohn’s-like
tocol does not recommend the routine assessment of TILs/ peritumoral lymphoid reaction. Further data on TNM
IELs.8 stage, histologic grade, and tumor location were obtained
In 2014, the International TILs Working Group from the pathology reports. All-cause survival data were
(ITWG) published a standardized approach for evalu- current as of August 1, 2019. Data on adjuvant chemo-
ating TILs on hematoxylin and eosin (H&E) sections in therapy were not available. The study was approved
the context of breast cancer.9 Briefly, stromal TILs are by the Northern Sydney Local Health District Human
scored as a percentage of the stromal areas between Research Ethics Committee.
nests of carcinoma cells, and areas occupied by the
malignant cells themselves are not included in the total Quantification of TILs Using the ITWG System
assessed surface area.9 Subsequently, the International A single H&E-stained slide showing deepest tumor
Immuno-Oncology Biomarkers Working Group has invasion was evaluated by 1 pathologist (T.L.F.) who was
recommended the use of the same methodology in the blinded to all clinical and pathologic data. After 1 hour of
evaluation of TILs in a range of other solid malig- intensive face-to-face training, the pathologist assessed the
nancies, including gastric, colorectal, and non–small cell number of TILs in strict accordance with the ITWG
lung carcinomas.2,3 guidelines.2,8,9
To the best of our knowledge, no studies have The ITWG methodology2,3 is summarized in Table 1
evaluated the prognostic value of TILs in CRCs using this and illustrated in Figure 1. Briefly, the density of TILs was
standardized ITWG scoring method in a sufficiently large assessed within the stromal compartment of the tumor mass,
cohort to justify implementation into clinical practice. and scored as a percentage of stromal area, rounded to the
We, therefore, sought to: nearest 5%. Scores were based on the average across the whole
(1) Validate the prognostic significance of the ITWG slide (not hotspots). Only TILs within the border of invasive
(also known as the International Immuno-Oncology tumors were assessed, so that dysplastic and in situ areas
Biomarkers Working Group) methodology for the (including growth confined to the lamina propria) and
assessment for TILs in a large unselected cohort inflammation outside the tumor borders were disregarded.
of CRCs. As per ITWG recommendations, all mononuclear cells (eg,
(2) Compare the prognostic significance of the ITWG lymphocytes and plasma cells) were included as TILs, whereas
scoring methodology for TILs with the RCPA- other inflammatory cells (ie, neutrophils/granulocytes) were
endorsed method of scoring IELs in clinical practice. excluded. Areas of necrosis, including the central “dirty
(3) Assess the interobserver concordance (κ score) of the necrosis” characteristic of CRC, were excluded from the
ITWG methodology across a range of observers. assessment, and only stromal TILs were assessed, so that TILs
(4) Determine whether the ITWG scoring methodology within nests of epithelial cells were excluded. Examples of
has prognostic significance independent of MMR/MSI TILs assessment using the ITWG system are provided in
status. Figure 2.

Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc. www.ajsp.com | 537
Fuchs et al Am J Surg Pathol  Volume 44, Number 4, April 2020

as a categorical variable. All statistical tests were 2 sided,

TABLE 1. Summary of the ITWG Methodology2,3 Adapted for
Assessing TILs in CRC and P-values ≤ 0.05 were considered significant.
1. TILs in CRCs should be assessed in the stromal compartment only
and reported as a percentage of the stromal area. For example, a score RESULTS
of 60% means that 60% of the stromal area is occupied by a
mononuclear infiltrate. Lymphocytes typically do not form solid Patient Characteristics
aggregates; therefore, a score of 100% would still allow some The cohort included 1034 patients with resected pri-
discernible stroma between individual lymphocytes
2. Stromal TILs evaluation should be confined to the borders of the
mary CRCs and is summarized in Table 2. TILs were scored
invasive tumor. TILs outside of the tumor borders should be excluded on all patients. MMR/MSI status was recorded for 1031
3. TILs in zones of necrosis, fibrosis, and abscess formation should be patients, of whom 1028 also had BRAFV600E mutation
excluded status recorded. The mean age at surgery was 73 years
4. One H&E-stained section can be considered to be sufficiently (range: 27 to 101 y). A total of 526 (50.9%) patients were
representative of the entire tumor for the assessment of TILs
5. TILs should be reported as a continuous variable, that is an average of female. During the follow-up period, 206 (19.9%) had death
the stromal TIL density over the entire section, rounded to the nearest recorded. The mean duration of follow-up was 5 years.
5%. Do not focus on hotspots After categorizing the continuous TILs data using the
6. All mononuclear inflammatory cells (lymphocytes and plasma cells) ITWG system, 395 cases (38.2%) were classified as low
should be scored, but other inflammatory cells (ie, neutrophils/
granulocytes) should be excluded
TILs, 584 (56.5%) as intermediate TILs, and 55 (5.3%) as
7. The assessment of TILs is confined to invasive carcinoma. Dysplastic high TILs. Using the RPCA-endorsed system, increased
and in situ areas (including growth confined to the lamina propria) are intratumoral lymphocytes were absent in 940 cases (92%)
excluded and present in 86 cases (8%). A peritumoral Crohn’s-like
lymphocytic response was categorized as absent in 755 cases
(74%) and present in 260 cases (26%).
Assessment of Interobserver Concordance of χ2 tests were performed to evaluate associations be-
TILs Using the ITWG System tween the TILs scores and age, sex, TNM stage, histologic
Two additional pathologists (C.J.R. and K.J.-N.) grade, tumor location, MMR status, and BRAFV600E
independently scored a subset of 100 and 181 randomly status. High TILs scores were significantly associated with
selected cases. To evaluate the importance of adequate advanced age (P = 0.002), low TNM stage (P =0.0001), and
training, 1 pathologist was given face-to-face interactive low histologic grade (P = 0.0001). No significant differences in
training on the ITWG methodology for 1 hour in addition TILs scores were observed for tumor location (P = 0.566), sex
to written information, whereas the other pathologist was (P = 0.07), MMR status (P = 0.641), or BRAFV600E status
provided written information only. These 2 pathologists (P = 0.583).
scored different subgroups of patients.
Survival Analyses
Statistical Analysis Overall survival was best in the high-TIL group
All statistical analyses were performed using SPSS, (mean survival: 75 mo), followed by the intermediate-TIL
version 25.0 (IBM Inc., Armonk, NY). Overall survival group (mean survival: 67 mo), and then the low-TIL group
was defined as the time from surgery until any-cause death (mean survival: 53 mo), P = 0.0001 (Fig. 3A).
(date of census: August 1, 2019). The continuous TILs This survival benefit remained statistically significant in
data were categorized into 3 groups: low (0% to 10%), subgroup analyses for mismatch repair proficient (MMRp)
intermediate (15% to 50%), and high (55% to 100%). The cases (P = 0.0001, Fig. 3B), MMRd cases (P = 0.031,
associations between categories of TILs and clin- Fig. 3C), BRAFV600E wild-type cases (P = 0.0001), and
icopathologic variables were compared using χ2 tests. BRAFV600E-mutant cases (P = 0.001). A survival analysis
Kaplan-Meier curves were constructed to compare sur- for combined stages II and III cases (n = 782) was performed
vival across groups of TILs. The effect of TILs on overall to investigate the potential value of TILs in aiding in
survival was also estimated in subgroup analyses stratified treatment decisions in this subgroup. TILs were significantly
by the TNM stage, MMR status, and BRAFV600E status. correlated with overall survival within this group (P < 0.0001).
The same survival analyses were performed using the When restricted to (n = 408) stage II cases only, TILs narrowly
presence of high or low/absent IEL scores and the pres- failed to show a statistically significant correlation with overall
ence or absence of peritumoral Crohn’s-like lymphocytic survival (P = 0.072).
response based on the RPCA-endorsed system obtained Age, tumor stage, histologic grade, tumor location,
from the original histopathologic reports. Univariate and BRAFV600E status, and TILs were all significantly asso-
multivariate Cox proportional hazard regression analyses ciated with overall survival in univariate analyses
were used to evaluate the association between TILs, age, (Table 3). Multivariate analyses showed significant
sex, TNM stage, histologic grade, tumor location, MMR associations between decreased overall survival and low
status, BRAFV600E status, and overall survival. Finally, TILs (0% to 10%) (hazard ratio [HR] = 1.891; 95%
Cohen κ analyses were performed to determine the level of confidence interval [CI]: 1.415-2.527; P = 0.0001), age
agreement between the pathologist who scored all 1034 > 73 years (HR = 2.711; 95% CI: 1.961-3.748; P = 0.0001),
patients (T.L.F., considered the gold standard for purpose tumor stage ≥ 3 (HR = 2.795; 95% CI: 2.059-3.795;
of analysis) and the 2 other pathologists who scored TILs P = 0.0001), right-sided location (HR = 1.405; 95% CI:

538 | www.ajsp.com Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc.
Am J Surg Pathol  Volume 44, Number 4, April 2020 TILs in CRC

FIGURE 1. Assessing TILs in CRC using the ITWG standardized methodology. A, Only TILs within the borders of the invasive tumor
(encircled in this image) are evaluated. B, Immune infiltrates outside the border of the tumor, including lymphoid aggregates
(arrows), are excluded. C, Only stromal lymphocytes are assessed. Lymphocytes within the epithelial component of the tumor mass
(shaded) including the areas of the so-called “dirty necrosis” characteristic of CRC are not included in the evaluation. D, Only
stromal mononuclear inflammatory cells (lymphocytes and plasma cells) are included. Granulocytic infiltrates both within areas of
necrosis (*) or other areas including peritumoral abscesses are not included (H&E-stained sections).

1.040-1.899; P = 0.027); MMRp status (HR = 1.852; 95% Interobserver Agreement

CI: 1.190-2.884, P = 0.006), and high histologic grade Cohen κ analysis showed good interobserver agree-
(HR = 1.635; 95% CI: 1.199-2.231; P = 0.002). ment between the gold-standard pathologist and the 1
The presence of a peritumoral Crohn’s-like response pathologist who received face-to-face training, κ = 0.753
assessed using standard reporting protocols was associated (95% CI: 0.623-0.883), P = 0.0001 (n = 100). Without in-
with a significantly better overall survival (mean survival: tensive training, the interobserver agreement was only
68 mo) compared with cases reported as having an absent moderate, κ = 0.436 (95% CI: 0.313-0.559), P = 0.0001
Crohn’s-like response (mean survival: 59 mo), P = 0.001 (n = 181) with the gold-standard pathologist.
(Fig. 3D). Although this association remained significant
in MMRp cases (mean survival: 70 vs. 59 mo, P = 0.001),
it was not seen in the subgroup analyses for MMRd DISCUSSION
cases (mean survival: 65 vs. 59 mo, P = 0.333) (Fig. 3E). Given that approaches to scoring TILs have varied
There was no statistically significant association between considerably both between and within tumor types, and
increased IELs assessed using the RCPA method across different institutions, a standardized methodology
and overall survival (mean survival: 59 mo for the such as the ITWG has clear benefits. Although initially
presence of IELs vs. 62 mo for absence of IELs; proposed and validated in breast cancer,9 this system has
P = 0.782) (Fig. 3F). subsequently been evaluated in relatively small patient

Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc. www.ajsp.com | 539
Fuchs et al Am J Surg Pathol  Volume 44, Number 4, April 2020

FIGURE 2. (A-F) Examples of tumor infiltrating lymphocytes (TILs) scored using the methodology of the International TILs Working
Group (H&E, original magnifications 100x, insets 400 ).

cohorts over a range of other tumor types.2,3,12,13 To the by the lack of subgroup analyses to control for MMR/
best of our knowledge, the only study to evaluate this BRAFV600E status, and the lack of κ scores to examine
methodology in CRC is that of Iseki et al,13 who reviewed the interobserver agreement. These limitations have been
a small cohort (n = 160) of stage II and III tumors. In overcome in this study, which provides evidence that the
addition to their small sample size, their study was limited assessment of TILs using the ITWG system is a very

540 | www.ajsp.com Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc.
Am J Surg Pathol  Volume 44, Number 4, April 2020 TILs in CRC

these data should be considered a significant limitation of

TABLE 2. Patient Characteristics and Association With TILs for
(N = 1034) CRC Patients the study. Nevertheless, if other studies indicate that a low
TIL score assessed using the ITWG system identifies “high-
TILs, n (%)
risk” patients with stage II CRC, a strong argument can be
Low Intermediate High made to include TIL assessment using the ITWG method in
N = 1034 n (%) (N = 395) (N = 584) (N = 55) P the decision-making process for assessing these patients for
Age at 0.002 adjuvant treatment. Indeed, it is promising that, for stage
diagnosis (y) III (n = 375, P < 0.001), combined stages I and II as a group
Mean 73 (n = 613, P = 0.02), and combined stages II and III as a
Range 27-101
≤ 73 454 (44) 166 (42) 275 (47) 13 (24) group (n = 782, P < 0.0001), TILs assessed using the ITWG
> 73 580 (56) 229 (58) 309 (53) 42 (76) methodology remained prognostic.
Sex 0.07 Although the presence of a tumor-induced immune re-
Male 508 (49) 202 (51) 287 (49) 19 (35) sponse has long been known to be associated with better sur-
Female 526 (51) 193 (49) 297 (51) 36 (65)
TNM stage 0.0001*
vival in CRCs,1–3 it is important to note in this study that the
I 205 (20) 31 (8) 147 (25) 27 (49) predictive benefit of TILs assessed using the ITWG system was
II 408 (40) 144 (36) 243 (42) 21 (38) superior to the traditional approach of identifying a peritu-
III 375 (36) 184 (47) 184 (31) 7 (13) moral Crohn’s-like inflammatory response or IELs using a
IV 46 (4) 36 (9) 10 (2) 0 cutoff of ≥ 5 lymphocytes per HPF. Although it was initially
Histologic 0.0001
grade hypothesized that lymphocytes directly interacting with carci-
(N = 1031) noma cells might be more relevant, most studies have found
Low 787 (76) 274 (70) 474 (81) 39 (71) stromal TILs to be a more reproducible parameter.9 This seems
High 244 (24) 119 (30) 109 (19) 16 (29) to be because intratumoral TILs are usually present in smaller
Tumor 0.566
location
numbers and are more heterogenous and difficult to observe on
Right 518 (50) 192 (49) 294 (50) 32 (58) H&E-stained sections. Moreover, limiting evaluation to the
Left 516 (50) 203 (51) 290 (50) 23 (42) stromal compartment means that the TIL counts will not be
MMR status 0.641 affected by the density and growth pattern of the carcinoma
(N = 1032) nests.9 In the present study, although the presence of a Crohn-
MMRp 803 (78) 312 (79) 453 (78) 10 (40)
MMRd 229 (22) 83 (21) 131 (22) 15 (60) like peritumoral lymphocytic response was found to be asso-
BRAFV600E 0.583 ciated with survival benefit, this association was lost in the
status subgroup analysis of MMRd cases, an interesting finding given
(N = 1032) that this parameter was initially developed to help identify
Positive 289 (28) 117 (30) 159 (27) 13 (24)
Negative 743 (72) 278 (70) 424 (73) 41 (76)
potential MMRd cases in the era before routine im-
munohistochemistry. Moreover, no survival benefit was found
Bold P values are statistically significant. for cases with increased IELs assessed prospectively using the
P-values are obtained using the χ2 test.
*Comparison between stages I to II vs. III to IV. standard RCPA method. These findings support the superior
prognostic value of the standardized ITWG methodology for
scoring TILs over these other methodologies.
The ITWG system is easy to learn and quick to apply
strong predictor of all-cause survival across an unselected in clinical practice, as it simply involves the evaluation of
cohort of 1034 CRCs undergoing surgical treatment. This stromal TILs, which are reported as a percentage of stromal
significant association between TILs, as assessed using the area. Indeed, in our study, the ITWG system showed good
ITWG system, and all-cause survival was independent of interobserver concordance, κ = 0.753 (95% CI: 0.623-0.883),
many patient and tumor characteristics, including MMR P = 0.0001, after just 1 hour of face-to-face training. How-
deficiency, and maintained significance in a multivariable ever, it is noted that the concordance slipped to a moderate
model that included sex, age, tumor site, histologic grade, level, κ = 0.436 (95% CI, 0.313-0.559), P = 0.0001 (n = 181),
MMR status, and BRAFV600E mutation status. when only written materials were provided—illustrating the
A particular area of clinical uncertainty is which stage importance of appropriate education.
II CRC patients would benefit from adjuvant therapy.14–19 Studies using immunohistochemistry with or without
There is no firm evidence that adjuvant chemotherapy im- digital quantitation of TILs have also demonstrated their
proves overall survival for all stage II CRC patients,17,18 prognostic value in CRC22–26 and have provided addi-
and, currently, the National Cooperative Cancer Network tional prognostic information compared with the semi-
(NCCN) guidelines recommend adjuvant therapy only for quantitative systems described above.27 At present, it is
stage II CRC with “high-risk features.”20 Despite ongoing unlikely that the costs associated with artificial in-
investigation of many molecular markers, “high risk” has telligence-based or immunohistochemical scoring are jus-
been difficult to define in the routine clinical setting.21 tified by the additional prognostic information provided
Unfortunately, in this study for (n = 408) stage II CRCs, by these techniques when TILs can be readily assessed
TILs just failed to reach statistical significance (P = 0.072). quickly and cheaply using the ITWG system on routine
It is possible that the use of adjuvant therapy in some cases H&E sections. Nevertheless, a standardized and reproducible
could have affected these results, and our lack of access to methodology for evaluating TILs on H&E sections in CRC

Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc. www.ajsp.com | 541
Fuchs et al Am J Surg Pathol  Volume 44, Number 4, April 2020

FIGURE 3. Overall 5-year Kaplan-Meier survival curves. A, ITWG assessment of TILs predicts overall survival for all CRC patients
(N = 1034). B, ITWG assessment of TILs predicts survival in MMRp CRC patients (N = 801). C, ITWG assessment of TILs predicts
overall survival for MMRd CRC patients (N = 228). D, Peritumoral Crohn’s-like lymphocytic response predicts survival in all CRC
patients (N = 1019). E, Peritumoral Crohn’s-like lymphocytic response is not statistically significantly associated with overall survival
in MMRd CRC patients (N = 226). F, IELs assessed using the RCPA method do not predict survival in all CRC patients (N = 1031).

may form the basis for the development of such artificial in- neoantigen creation,28 and this increased immune response
telligence software in the near future. is thought to account for both increased IELs and a better
MMRd CRCs are thought to be more immunogenic prognosis compared with MMRp CRC.29 It is, therefore,
secondary to a higher mutation burden, leading to more surprising that, in this study, there was no statistically

542 | www.ajsp.com Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc.
Am J Surg Pathol  Volume 44, Number 4, April 2020 TILs in CRC

TABLE 3. Univariate and Multivariate Analyses of Prognostic Factors for Overall Survival for (N = 1034) CRC Patients
Univariate Analysis† Multivariate Analysis†
Variables N = 1034 Mean Survival (mo)* HR 95% CI P HR 95% CI P
Age at diagnosis (y)
≤ 73 454 67 1 1
> 73 580 57 2.438 1.785-3.331 0.0001 2.711 1.961-3.748 0.0001
Sex (male)
Female (N = 524) 526 62 1 1
Male (N = 508) 508 61 0.935 0.712-1.227 0.628 0.997 0.754-1.319 0.985
TNM stage (eighth edition)
I-II 613 69 1 1
III-IV 421 51 3.121 2.342-4.159 0.0001 2.795 2.059-3.795 0.0001
Histologic grade
Low 787 64 1 1
High 244 51 1.989 1.495-2.646 0.0001 1.635 1.199-2.231 0.002
Location
Right 518 58 1.595 1.21-2.103 0.001 1.405 1.040-1.899 0.027
Left 516 65 1 1
MMR status
MMRp 803 61 1.129 0.807-1.579 0.479 1.852 1.190-2.884 0.006
MMRd 228 63 1 1
BRAFV600E mutation status
Positive 289 55 1.52 1.143-2.022 0.004 1.435 0.976-2.110 0.066
Negative 743 64 1 1
TILs
Low 394 52 2.496 1.887-3.301 0.0001 1.891 1.415-2.527 0.0001
Intermediate and high 638 68 1 1
Bold P values are statistically significant.
*Mean survival calculated using the Kaplan-Meier method.
†Cox regression model.

significant association between TILs, as assessed using the solid cancer: the next step in precision oncology. Virchows Arch.
ITWG system, and MMR status (P = 0.641). It is also in- 2019;474:463–474.
teresting to note that MMR status did not predict survival 2. Hendry S, Salgado R, Gevaert T, et al. Assessing tumor-infiltrating
lymphocytes in solid tumors: a practical review for pathologists and
in univariate analysis (P = 0.479) but did become very sig- proposal for a standardized method from the International Immuno-
nificant in a multivariate model (P = 0.006) because of an Oncology Biomarkers Working Group: part 2: TILs in melanoma,
association between the positive prognostic factor of gastrointestinal tract carcinomas, non-small cell lung carcinoma and
MMRd status and the adverse factors of advanced age and mesothelioma, endometrial and ovarian carcinomas, squamous cell
carcinoma of the head and neck, genitourinary carcinomas, and
right-sided location. The early literature suggested that the primary brain tumours. Adv Anat Pathol. 2017;24:311–335.
primary reason to identify TILs was to triage genetic testing 3. Hendry S, Salgado R, Gevaert T, et al. Assessing tumor-infiltrating
for Lynch syndrome by selecting patients for further MMR lymphocytes in solid tumors: a practical review for pathologists and
testing,30–35 and some protocols, including the CAP,8 no proposal for a standardized method from the International Immuno-
longer recommend TIL assessment in the era of universal Oncology Biomarkers Working Group: part 1: assessing the host
immune response, TILs in invasive breast carcinoma and ductal
screening for Lynch syndrome by MSI/MMR. However, carcinoma in situ, metastatic tumor deposits and areas for further
our data clearly support the benefit of assessing both TILs research. Adv Anat Pathol. 2017;24:235–251.
and MMR/MSI status and question whether the number of 4. Linette GP, Carreno BM. Tumor-infiltrating lymphocytes in the
TILs assessed using the ITWG system is a useful marker of checkpoint inhibitor era. Curr Hematol Malig Rep. 2019;14:286–291.
MMR/MSI status. 5. Kong JC, Guerra GR, Pham T, et al. Prognostic impact of tumor-
infiltrating lymphocytes in primary and metastatic colorectal cancer:
In conclusion, this study provides strong support for a systematic review and meta-analysis. Dis Colon Rectum. 2019;62:
the assessment of TILs in CRC using the ITWG system. 498–508.
This simple system shows good interobserver concordance 6. Klintrup K, Makinen JM, Kauppila S, et al. Inflammation and
after minimal training, has superior prognostic value when prognosis in colorectal cancer. Eur J Cancer. 2005;41:2645–2654.
7. Royal College of Pathologists of Australasia (RCPA). Colorectal cancer
compared with other methods of scoring tumor-associated structured reporting protocol (3rd edition); 2016. Available at: www.rcpa.
lymphocytic inflammation in clinical practice, and predicts edu.au/Library/Practising-Pathology/Structured-Pathology-Reporting-of-
all-cause survival independent of MMR/MSI status, TNM Cancer/Cancer-Protocols/Gastrointestinal/Protocol-colorectal-cancer.
stage, and a range of other patient and tumor factors. Accessed July 5, 2019.
8. College of American Pathologist (CAP). Cancer Protocol Templates.
Protocol for the examination of specimens from patients with
REFERENCES primary carcinoma of the colon and rectum; 2017. Available at:
1. Giraldo NA, Peske JD, Sautes-Fridman C, et al. Integrating www.cap.org/protocols-and-guidelines/cancer-reporting-tools/cancer-
histopathology, immune biomarkers, and molecular subgroups in protocol-templates. Accessed August 30, 2019.

Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc. www.ajsp.com | 543
Fuchs et al Am J Surg Pathol  Volume 44, Number 4, April 2020

9. Salgado R, Denkert C, Demaria S, et al. The evaluation of tumor- worldwide consortium-based analysis of 1,336 patients. J Clin Oncol.
infiltrating lymphocytes (TILs) in breast cancer: recommendations by an 2016;34 (suppl):3500.
International TILs Working Group 2014. Ann Oncol. 2014;26:259–271. 24. Matsutani S, Shibutani M, Maeda K, et al. Verification of the
10. Toon CW, Walsh MJ, Chou A, et al. BRAFV600E immunohis- methodology for evaluating tumor-infiltrating lymphocytes in color-
tochemistry facilitates universal screening of colorectal cancers for ectal cancer. Oncotarget. 2018;9:15180–15197.
Lynch syndrome. Am J Surg Pathol. 2013;37:1592–1602. 25. Eriksen AC, Sorensen FB, Lindebjerg J, et al. The prognostic value
11. Reagh J, Clarkson A, Bullock M, et al. Real world experience of of tumor-infiltrating lymphocytes in stage II colon cancer.
BRAFV600E mutation specific immunohistochemistry in colorectal A nationwide population-based study. Transl Oncol. 2018;11:979–987.
carcinoma. Pathology. 2018;50:342–344. 26. Ko Y, Pyo J. Clinicopathological significance and prognostic role of
12. Heikkinen I, Bello I, Wahab A, et al. Assessment of tumor- tumor-infiltrating lymphocytes in colorectal cancer. Int J Biol
infiltrating lymphocytes predicts the behavior of early-stage oral Markers. 2019;34:132–138.
tongue cancer. Am J Surg Pathol. 2019;43:1392–1396. 27. Park JH, McMillan DC, Edwards J, et al. Comparison of the
13. Iseki Y, Shibutani M, Maeda K, et al. A new method for evaluating prognostic value of measures of the tumor inflammatory cell
tumor-infiltrating lymphocytes (TILs) in colorectal cancer using infiltrate and tumor-associated stroma in patients with primary
hematoxylin and eosin (H&E)-stained tumor sections. PLoS One. operable colorectal cancer. Oncoimmunology. 2016;5:e1098801.
2018;13:e0192744. 28. Phillips SM, Banerjea A, Feakins R, et al. Tumour-infiltrating
14. Dotan E, Cohen SJ. Challenges in the management of stage II colon lymphocytes in colorectal cancer with microsatellite instability are
cancer. Semin Oncol. 2011;38:511–520. activated and cytotoxic. Br J Surg. 2004;91:469–475.
15. Meropol NJ. Ongoing challenge of stage II colon cancer. J Clin 29. Rozek LS, Schmit SL, Greenson JK, et al. Tumor-infiltrating
Oncol. 2011;29:3346–3348. lymphocytes, Crohn’s-like lymphoid reaction, and survival from
16. Vachani C, Giantonio B. Stage II colon cancer: to treat or not to colorectal cancer. J Natl Cancer Inst. 2016;108:djw027.
treat? 2013. Available at: www.oncolink.org/types/article.cfm?c= 30. Alexander J, Watanabe T, Wu T, et al. Histopathological
124&id=9621. Accessed July 8, 2019. identification of colon cancer with microsatellite instability. Am J
17. Gray R, Barnwell J, McConkey C, et al. Quasar Collaborative Pathol. 2001;158:527–535.
Group. Adjuvant chemotherapy versus observation in patients with 31. Greenson JK, Bonner JD, Ben-Yzhak O, et al. Phenotype of
colorectal cancer: a randomised study. Lancet. 2007;370:2020–2029. microsatellite unstable colorectal carcinomas: well-differentiated and
18. O’Connor ES, Greenblatt DY, LoConte NK, et al. Adjuvant focally mucinous tumors and the absence of dirty necrosis correlate
chemotherapy for stage II colon cancer with poor prognostic with microsatellite instability. Am J Surg Pathol. 2003;27:563–570.
features. J Clin Oncol. 2011;29:3381–3382. 32. Smyrk TC, Watson P, Kaul K, et al. Tumor-infiltrating lymphocytes
19. Kannarkatt J, Joseph J, Kurniali PC, et al. Adjuvant chemotherapy for are a marker for microsatellite instability in colorectal carcinoma.
stage II colon cancer: a clinical dilemma. J Oncol Pract. 2017;13:233–241. Cancer. 2001;91:2417–2422.
20. Benson AB, Venook AP, Al-Hawary MM, et al. NCCN Guidelines 33. Jenkins MA, Hayashi S, O’Shea AM, et al. Pathology features in
insights: colon cancer, version 2.2018. J Natl Compr Canc Netw. Bethesda guidelines predict colorectal cancer microsatellite insta-
2018;16:359–369. bility: a population-based study. Gastroenterology. 2007;133:48–56.
21. Gao S, Tibiche C, Zou J, et al. Identification and Construction of 34. Asad Umar C, Boland R, Terdiman JP, et al. Revised Bethesda
Combinatory Cancer Hallmark–based gene signature sets to predict Guidelines for hereditary nonpolyposis colorectal cancer (Lynch
recurrence and chemotherapy benefit in stage II colorectal cancer. syndrome) and microsatellite instability. J Natl Cancer Inst. 2004;96:
JAMA Oncol. 2016;2:37–45. 261–268.
22. Galon J, Costes A, Sanchez-Cabo F, et al. Type, density and location 35. Shia J, Tang LH, Vakiani E, et al. Immunohistochemistry as first-
of immune cells within human colorectal tumors predict clinical line screening for detecting colorectal cancer patients at risk for
outcome. Science. 2006;313:1960–1964. hereditary nonpolyposis colorectal cancer syndrome: a 2-antibody
23. Galon J, Mlecnik B, Marliot F, et al. Validation of the immunoscore panel may be as predictive as a 4-antibody panel. Am J Surg Pathol.
(IM) as a prognostic marker in stage I/II/III colon cancer: results of a 2009;33:1639–1645.

544 | www.ajsp.com Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc.