Ann Surg Oncol (2022) 29:4568–4578

https://doi.org/10.1245/s10434-022-11544-x

ORIGINAL ARTICLE – TRANSLATIONAL RESEARCH

LKB-1 Expression and High-Risk Histopathology are

Independent Prognostic Factors for Patients with Oral Cavity
Carcinoma
José F. Carrillo, MD1, Christian Cruz-Romero, MD1, Alejandro Avilés-Salas, MD2, Liliana C. Carrillo, MD3,
Margarita C. Ramı́rez-Ortega, MD, PhD4, Roberto Herrera-Goepfert, MD2, Rafael Vázquez-Romo, MD1,
Gabriela Figueroa-González, MD, PhD5, Javier I. Altamirano-Garcı́a, MD6, and Luis F. Oñate-Ocaña, MD,
MSc3
1
Departamento de Cabeza y Cuello, Instituto Nacional de Cancerologı́a, Mexico City, Mexico; 2Departamento de
Patologı́a, Instituto Nacional de Cancerologı́a, Mexico City, Mexico; 3Subdirección de Investigación Clı́nica, Instituto
Nacional de Cancerologı́a, Mexico City, Mexico; 4Subdirección de Investigación Básica, Instituto Nacional de Cardiologı́a
Ignacio Chávez, Mexico City, Mexico; 5Unidad de Investigación Multidisciplinaria (UMIEZ), Facultad de Estudios
Superiores Zaragoza, UNAM, Mexico City, Mexico; 6Subdirección de Radioterapia, Instituto Nacional de Cancerologı́a
(INCan), Mexico City, Mexico

ABSTRACT expression is considered the reference category, the factors

Background. The expression of liver kinase B1 (LKB-1) independently associated with recurrence were low (odds
has been associated with prognosis in squamous cell car- ratio [OR], 0.157; 95% confidence interval [CI],
cinoma of the oral cavity (SCCOC). This study aimed to 0.044–0.557), intermediate (OR, 0.073; 95% CI,
define the prognostic role of LKB-1 expression for patients 0.017–0.319), and intense (OR, 0.047; 95% CI,
with SCCOC and the suitability of its integration into a 0.007–0.304) expression of LKB-1. This model permitted
multivariate prognostic model. construction of a computer software program capable of
Methods. A retrospective cohort study of patients with prediction with receiver operating characteristic analysis
SCCOC was conducted in a cancer center. Expression of (area under the curve, 0.925) and led to the definition of
LKB-1 was evaluated by immunohistochemistry, and five prognostic groups with a biologic gradient.
multivariate analysis defined prognostic factors associated Conclusion. These results suggest that LKB-1 expression
with recurrence, recurrence-free survival (RFS), and in patients with SCCOC is of robust prognostic value and
overall survival (OS). The logistic regression model was complements the TNM staging system. The proposed
used to construct a predictive computer software program. model requires external validation in prospective observa-
Results. Of the 201 patients in this study, 104 (51.7%) tional studies.
experienced recurrence of their disease. Lower expression
of LKB-1, high-risk histopathology, and advanced tumor-
node-metastasis (TNM) stages were independent factors Squamous cell carcinoma (SCC), the most common
via multivariate analysis associated with the increased neoplasm of the head and neck area,1 usually is attributed
recurrence risk, poor RFS, and poor OS. If lack of LKB-1 to alcohol or tobacco use.2 Although SCC of the orophar-
ynx typically has been associated with human
papillomavirus (HPV) infection,3,4 the prevalence of HPV
has not been associated significantly with SCC oral cavity
Ó Society of Surgical Oncology 2022 carcinoma (SCCOC). Currently, SCCOC is staged with the
First Received: 19 August 2021 tumor-node-metastasis (TNM) classification, which usually
Accepted: 14 February 2022; provides risk stratification and adequate treatment selec-
Published Online: 23 March 2022
tion, thus avoiding over- or undertreatment.5,6 However,
L. F. Oñate-Ocaña, MD, MSc some patients with early-stage SCCOC experience a
e-mail: [email protected]
LKB-1 in oral cavity carcinoma 4569

recurrence with poor survival, even if surgical margins are and magnetic resonance imaging (MRI) of the head and
clear and lymph node metastases are absent. On the other neck region (panorex was used in the case of tumors
hand, for some patients with intermediate-advanced located in or near the gingivae), nasopharyngolaryn-
SCCOC, locoregional therapies, including surgery and goscopy, and chest x-ray. The patients also were required
radiotherapy, lead to satisfactory results.7,8 to have histopathology confirmation of their SCCOC
The presence of p53 mutations and HPV infection are diagnosis, and the specific site/subsite was determined via
not adequate prognostic markers in SCCOC, although they consensus by a surgical oncologist, an imaging specialist,
may be clinically significant factors in oropharynx SCC and a medical oncologist.
and probably in supraglottic larynx SCC according to The Comité de Investigación and the Comité de Bioética
series published in Western countries.4,9 en Investigación of the INCan reviewed and approved this
In recent years, a tendency has developed to investigate study protocol (registries 014/003/CCI and CEI/892).
the intrinsic pathologic and molecular characteristics of The TNM staging system of the eighth edition of the
SCCOC in order to predict in an efficient and rational way American Joint Commission on Cancer (AJCC) was
the prognosis of patients and the need for administration of used.18 Histopathologic diagnoses were confirmed with
adjuvant therapy and to consider the development of tar- review of the pertinent slides. Surgical margins were
geted agents to improve the outcomes. defined as negative when the neoplasm was found 10 mm
Liver kinase B1 (LKB-1, coded by the STK11 gene) is or farther from the border of the surgical specimen and
the expression of a tumor suppressor gene (STK11), which marginal when found farther than 5 mm but not farther
frequently is mutated and inactivated in non-small cell lung than 10 mm from the border of the surgical specimen.
cancer (NSCLC) and other tumors.10,11 Our group inves- Positive surgical margins were those detected not farther
tigated its role in the initiation and progression of head and than 5 mm from the surgical specimen or on the border of
neck SCC and reported that decreased expression of this the surgical specimen. Surgical bed margins were deter-
gene is associated with worse oncologic outcomes, mostly mined in the same manner as the circumferential margins
in SCCOC. Findings have shown this gene be under-ex- described earlier.
pressed by interaction with miR-100-3p.12,13 Histopathologic risk factors according to Brandwein-
Furthermore, a novel four-tiered prognostic system Gensler et al.7,8 were considered as well. The pattern of
places special emphasis on the invasive pattern and the invasion was defined as a pushing pattern of invasion in the
lymphocytic infiltrate (also termed ‘‘lymphocytic host presence of a cohesive homogeneous group of cells in the
response’’) of the tumors, defining in an efficient way the tumor border. A digital pattern of invasion referred to
prognosis of SCCOC. This has been validated in several neoplastic finger-like strands infiltrating the neighboring
studies.7,8,14 tissues, and an island pattern represented tissue inter-
Therefore, we performed a retrospective analysis in a spersed in the neighboring structures and clearly separated
cohort of patients with SCCOC to define the prognostic from the main tumor structure. The presence of a lym-
role of previously known clinical and histopathologic fea- phocytic infiltrate was classified as absent, moderate, or
tures, including the expression of LKB-1, and their ability plaque infiltration or intense infiltration according to its
to complement the TNM staging system, thereby increas- presence in the margins and in the tumor. The presence of
ing its performance and its suitability in the integration of a perineural invasion referred to infiltration greater than 1
prognostic model.13,15–18 mm of nerve trunks by malignant cells and was classified
as absent or present. Lymphovascular infiltration or
PATIENTS AND METHODS thrombi were identified in the neoplasm at a 9 40 to 9
100 magnification and confirmed at 9 400 magnification.
Patients This was classified as absent or present if more than two
vessels were involved.
Patients with SCCOC managed at the INCan Hospital in
Mexico City who received treatment with a curative Immunohistochemistry
intention between July 2000 and August 2017 were
included in this retrospective cohort study. Individuals We ussd both biopsy or surgical specimens fixed in 10%
were identified from clinical charts, surgical pathology formalin and embedded in paraffin cassettes. For
files, and surgery records. The inclusion criteria specified histopathologic evaluation by immunohistochemistry
patients older than 17 years of any gender who had a (IHC), tissue sections 4 lm thick were obtained, mounted
complete clinical history and physical examination, imag- in glass slides, and stained with hematoxylin and eosin.
ing studies including a computed tomography (CT) scan Tumor samples were examined for LKB-1 and p16
expression to detect the presence of HPV16.
4570 J. F. Carrillo et al.

Briefly, after deparaffinization and rehydration, tissue probability values to establish significance. The IBM SPSS
sections were boiled in 10 mmol of citrate buffer, pH 8.0, Statistics software for Mac, version 28 (IBM Corp,
at 95 °C for 40 min. The slides then were incubated in 5% Armonk, NY, USA, 2021) was used.
hydrogen peroxide to block endogenous peroxidase. After
washing, they were incubated at 37 °C for 1 h with the RESULTS
anti-LKB-1 rabbit monoclonal antibody (clone D60C5F10,
dilution 1:250, Cell-Signaling Technology, Danvers, MA, Patients
USA) or with anti-p16 antibody (1:100 dilution; JC8, Santa
Cruz, CA, USA). The slides were left overnight at 4 °C and This cohort included 201 patients (99 women [49.3%]
subsequently incubated with a biotinylated secondary and 102 men [50.7%]). The mean age of the patients was
antibody bound to a streptavidin-peroxidase conjugate. 58.5 ± 13.6 years (range, 25–85 years). A smoking history
Bound antibodies were visualized by adding hydrogen was recorded in 85 cases (42.3%) and alcohol abuse in 79
peroxide with 3,3-diamino benzidine. Sections were cases (39.3%). The neoplasm subsites were the tongue (n =
counterstained with Meyers hematoxylin. 101, 50.2%), gingivae (n = 25, 12.4%), hard palate (n = 19,
All slides were evaluated by two independent onco- 9.5%), floor of the mouth (n = 14, 7%), retromolar trigone
pathologists blinded to the clinical data, who assessed the (RMT) (n = 13, 6.5%), upper gingivae (n = 11, 5.5%), oral
intensity of nuclear and cytoplasmic staining. Scoring of mucosae (n = 10, 5%), and mandible (n = 8, 4%).
LKB-1 expression hinged on the intensity of staining in In our study, 60 patients (29.9%) had surgery as the only
malignant cells and was divided into four categories: no treatment, 110 patients (54.7%) underwent surgery and
expression or weak expression when staining was identified adjuvant radiotherapy, and 31 patients (15.5%) had radio-
at 9400 magnification, moderate expression when reaction therapy upfront. Rescue surgery was performed for 13
was identified at 9100 magnification, and intense expres- (6.46%) of these patients. Chemotherapy was administered
sion apparent at 940 magnification, as previously to only 10 patients, with great heterogeneity in the number
described.19 Representative images of the LKB-1 scoring of cycles and in agents used. Consequently, this method is
categories are presented in the Fig. S1. When staining was not described in detail.
present, p16 status was categorized as positive.10–13 Of the 201 patients, 48 (23.9%) had no LKB-1 expres-
sion, 89 (44.3%) had low expression, 42 (20.9%) had
Statistical Analysis moderate expression, and 22 (10.9%) had intense expres-
sion. The intensity of LKB-1 expression and its association
After descriptive statistics, bivariate and multivariate with clinical and histopathologic characteristics in this
analyses were performed. Between-group comparisons cohort are described in Table 1. The presence of HPV
were obtained with one-way analysis of variance infection, as determined by expression of p16, was
(ANOVA) or chi-square tests, as appropriate. The logistic observed in 21 cases (10.4%) and was not associated with
regression model was used to define factors associated with LKB-1 expression.
the event of clinical recurrence or progression after treat-
ment.20 Odds ratios (ORs) were obtained as a measure of Recurrence
association together with their 95% confidence intervals
(CIs). Recurrence-free survival (RFS) was calculated from Recurrence or progression was recorded for 104 patients
the end of treatment to recurrence or the end of the study. (51.7%), including 101 locoregional recurrences (97.1%)
Overall survival (OS) was calculated from the date of and 3 distant metastases (2.9%). Two were bone metas-
diagnosis to the last visit or death. Survival curves were tases, and one was lung metastases. In the bivariate
constructed with the Kaplan-Meier method, and differences analysis, the intensity of LKB-1 expression was associated
were estimated with the log-rank method. Multivariate with the frequency of recurrence. Recurrence developed in
survival analysis was performed with Cox’s proportional 33 patients with no LKB-1 expression (68.8%), in 47
hazards model. Hazard ratios (HRs) were obtained as a patients with low LKB-1 expression (52.8 %), in 17
measure of association together with their 95% CIs. patients with moderate LKB-1 expression (40.5%), and in 7
The model assumptions were evaluated in the final patients with intense LKB-1 expression (31.8%) (p =
models, and possible interactions were investigated. Nei- 0.011). Bivariate analyses of other variables regarding their
ther exposure (LKB-1 status) nor outcome data had missing associations with recurrence, RFS, and OS, are shown in
values. Because several covariates did have missing values Table 2.
with completely random variation, multivariate regression The intensity of LKB-1 expression, coded as a four-
imputation was performed. Two-tailed statistics were used category variable, was independently associated with
in all cases. The 95% CIs were calculated as well as the
LKB-1 in oral cavity carcinoma 4571

TABLE 1 Clinical and histopathology data of patients with squamous cell carcinoma of the oral cavity according to the intensity of LKB-1
expression by immunoperoxidase (n = 201)
None Low Moderate High p Value
(n = 48) (n = 89) (n = 42) (n = 22)
n (%) n (%) n (%) n (%)

Mean age (years) 57.8 ± 12.9 58.9 ± 13.1 57 ± 14.1 61.2 ± 16.1 0.659
Gender
Female 22 (22.2) 44 (44.4) 18 (18.2) 15 (15.2) 0.254
Male 26 (25.5) 45 (44.1) 24 (23.5) 7 (6.9)
Mean BMI (kg/m2) 27 ± 4.6 25.5 ± 4.1 25.1 ± 3.3 25.2 ± 4.1 0.116
Mean blood hemoglobin (g/dL) 15 ± 1.4 14.3 ± 1.9 14.3 ± 2.2 14.7 ± 2.1 0.125
Mean serum albumin (g/dL) 4 ± 0.4 3.9 ± 0.4 3.9 ± 0.6 4.1 ± 0.4 0.178
Smoker 24 (28.2) 32 (37.6) 16 (18.8) 13 (15.3) 0.14
Non-smoker 24 (20.7) 57 (49.1) 26 (22.4) 9 (7.8)
Alcohol use 19 (24.1) 32 (40.5) 17 (21.5) 11 (13.9) 0.682
No alcohol use 29 (23.8) 57 (46.7) 25 (20.5) 11 (9)
HPV infection
Absent 6 (28.6) 7 (33.3) 6 (28.6) 2 (9.5) 0.67
Present 42 (23.3) 82 (45.6) 36 (20) 20 (11.1)
Differentiation
Low grade 11 (22.4) 23 (46.9) 7 (14.3) 8 (16.3) 0.401
Intermediate grade 32 (23.4) 58 (42.3) 33 (24.1) 14 (10.2)
High grade 5 (33.3) 8 (53.3) 2 (13.3) 0
Invasion pattern
Pushing 6 (15) 18 (45) 9 (22.5) 7 (17.5) 0.076
Digital 13 (19.1) 37 (54.4) 10 (14.7) 8 (11.8)
Islands 29 (31.2) 34 (36.6) 23 (24.7) 7 (7.5)
Perineural infiltration
Absent 19 (23.2) 35 (42.7) 17 (20.7) 11 (13.4) 0.83
Present 29 (24.4) 54 (45.4) 25 (21) 11 (9.2)
Lymphocytic infiltration
Absent 14 (37.8) 17 (45.9) 4 (10.8) 2 (5.4) 0.129
Moderate or plaque 11 (16.7) 29 (43.9) 19 (28.8) 7 (10.6)
Intense 23 (23.5) 43 (43.9) 19 (19.4) 13 (13.3)
Perivascular infiltration
Absent 20 (19) 51 (48.6) 18 (17.1) 16 (15.2) 0.041
Present 28 (29.2) 38 (39.6) 24 (25) 6 (6.3)
TNM stage
I 4 (40) 1 (10) 2 (20) 3 (30) 0.176
II 10 (25) 17 (42.5) 12 (30) 1 (2.5)
III 12 (24.5) 26 (53.1) 7 (14.3) 4 (8.2)
IVa 20 (22.2) 38 (42.2) 20 (22.2) 12 (13.3)
IVb 2 (16.7) 7 (58.3) 1 (8.3) 2 (16.7)
Successful surgical resection 44 (24) 78 (42.6) 42 (23) 19 (10.4) 0.111
No surgical resection 4 (22.2) 11 (61.1) 0 3 (16.7)
Surgical margins
R0 23 (20.4) 49 (43.4) 29 (25.7) 12 (10.6) 0.269
R1 (\10 mm) 15 (26.3) 25 (43.9) 11 (19.3) 6 (10.5)
R2 (positive) 6 (46.2) 4 (30.8) 2 (15.4) 1 (7.7)
4572 J. F. Carrillo et al.

Table 1 (continued)
None Low Moderate High p Value
(n = 48) (n = 89) (n = 42) (n = 22)
n (%) n (%) n (%) n (%)
No surgical resection 4 (22.2) 11 (61.1) 0 3 (16.7)
BMI, body mass index; HPV, human papillomavirus
n, number of patients; p, probability value.

recurrence in the multivariate analysis. As independent subgroups by quintiles. Figure 2A depicts the OS curves
covariates, TNM stage and high-risk histopathology also according to these five subgroups, showing a biologic
were associated with recurrence, as described in Table 3. gradient. Figure 2B portrays the receiver operating char-
Using the logistic regression model described in acteristic (ROC) curve, depending on the observed-
Table 3, a computer software program was developed to predicted probability of recurrence (area under the ROC
obtain the predicted probability of recurrence for each curve, 0.925; standard error, 0.018; 95% CI, 0.889–0.961;
patient. p = 2.3 9 10–25).

Survival DISCUSSION

The median follow-up period for the cohort was 3.17 The TNM staging system currently is the best classifi-
years (interquartile range, 7.15 years). The median RFS cation method for establishing prognosis and defining the
was 4.49 years (95% CI, 0.061 – 8.93 years). The, median treatment strategy for SCCOC. However, it could be
OS was 8.93 years (95% CI not calculable), and 93 events improved, and several oncology groups are searching for
of death were recorded. prognostic factors to complement the TNM classification.
The patients who had SCCOC with absent or low-in- New prognostic factors could explain the uncertain prog-
tensity LKB-1 expression had median RFS periods of nosis for a significant number of patients, encouraging the
0.654 years (95% CI, 0–1.69 years) and 5.99 years (95% design of new therapeutic strategies directed against the
CI, 0–13.35 years), respectively. At this writing, the med- intrinsic molecular and histopathologic characteristics of
ian RFS for the patients with moderate or intense SCCOC.
expression of LKB-1 had not been reached (p = 0.011). The Although the presence of HPV in SCC of the orophar-
patients with SCCOC and absent expression of LKB-1 had ynx is frequently reported as contributing to a favorable
a median OS of 2.1 years (95% CI, 0.681–3.53 years). To prognosis in European and North American countries, this
date, the median OS of the patients with any LKB-1 association has not been demonstrated in reports from
expression has not been reached (p = 0.08). Figure 1A and lower-income countries because other factors such as
B depict RFS and OS depending on the intensity of LKB-1 microbiota, undernourishment, and association with a
expression. smoking habit and alcoholism are prevalent as well. The
Tables 4 and 5 show the multivariate analysis of prog- prevalence of HPV in SCCOC is 5% to 10%, and it is not
nostic factors associated with RFS or OS, supporting the associated with a better prognosis.4
independent association of the intensity of LKB-1 expres- In our study, the prevalence of HPV infection in SCCOC
sion with RFS or OS. Proportionality assumptions were was 10.4%, with no significant impact on recurrence rate,
reviewed, and all were satisfactory. No significant inter- DFS, or OS. The HPV infection prevalence was slightly
action terms were found. Figure 1C and D show the higher than that reported in European and American
constructed RFS and OS hazard curves obtained with the series.21,22
Cox models described respectively in Tables 3 and 4, The prognosis for patients with SCCOC depends on the
showing that the hazard curves depend on the intensity of subsite, according to numerous reports.23 However, by
LKB-1 expression, with adjustment for the other covariates bivariate or multivariate analysis, we found no prognostic
in both models. value of the neoplasms specific localization that could
explain the final oncologic outcome. This may reflect our
Predictive Value cohort’s inclusion of mostly intermediate- and advanced-
stage cases.
The predicted probability values for recurrence obtained
by the computer software, were categorized into five
LKB-1 in oral cavity carcinoma 4573

A new proposal on the depth of invasion by SCCOC has As stated earlier, the TNM system was found to be an
been introduced into the AJCC staging system.18 In our independent prognostic factor in terms of recurrence, DFS,
cohort, 10 patients had an invasion depth of 5 mm or less, and OS in the bivariate and multivariate analyses (Tables 2,
and 40 patients had an invasion depth of 10 mm or less, and 3, 4, 5). However, stages II and III disease had similar ORs
the statistical power for these categories was low. and HRs, underscoring the great difficulty establishing a

TABLE 2 Bivariate Recurrence RFS OS

association of clinical and OR (95% CI)a HR (95% CI)a HR (95% CI)a
histopathology data of patients
with squamous cell carcinoma
Age (years) 0.987 (0.966–1.007) 0.998 (0.984–1.012) 1.001 (0.985–1.016)
of the oral cavity according to
the three different outcome Male gender 1.352 (0.776–2.355) 1.399 (0.95–2.059) 1.432 (0.951–2.155)
measures (n = 201) BMI (kg/m2) 0.981 (0.917–1.05) 0.978 (0.931–1.026) 0.976 (0.928–1.027)
Blood hemoglobin (g/dL) 1.081 (0.932–1.254) 1.046 (0.941–1.163) 1.059 (0.947–1.185)
Serum albumin (g/dL) 1.228 (0.691–2.183) 1.068 (0.693–1.645) 1.059 (0.68–1.648)
Smoker Yes 1.179 (0.673–2.067) 1.355 (0.917–2.001) 1.328 (0.88–2.003)
Alcohol abuse Yes 1.413 (0.8–2.498) 1.472 (0.996–2.176) 1.378 (0.911–2.084)
HPV infection Present 0.832 (0.336–2.056) 1.005 (0.523–1.933) 1.02 (0.513–2.028)
Differentiation
Low grade 1 1 1
Intermediate grade 1.844 (0.953–3.609) 1.553 (0.938–2.573) 1.517 (0.9–2.558)
High grade 4.342 (1.207–15.63) 2.762 (1.31–5.825) 2.405 (1.11–5.215)
Invasion pattern
Pushing 1 1 1
Digital 3.505 (1.36–9.034) 3.04 (1.33–6.947) 2.812 (1.149–6.882)
Islands 12.82 (5.031–32.68) 8.93 (4.074–19.57) 8.639 (3.716–20.09)
Perineural infiltration present 7.44 (3.923–14.109) 4.61 (2.815–7.551) 4.371 (2.57–7.432)
Lymphocytic infiltration
Absent 1 1 1
Moderate or plaque 0.074 (0.009–0.581) 0.324 (0.205–0.513) 0.299 (0.186–0.479)
Intense 0.007 (0.001–0.055) 0.066 (0.037–0.117) 0.077 (0.043–0.14)
Perivascular infiltration Present 5.622 (3.061–10.327) 3.733 (2.461–5.663) 3.242 (2.089–5.03)
TNM stage
I 1 1 1
II 0.9 (0.218–3.715) 1.029 (0.341–3.105) 1.854 (0.418–8.222)
III 1.696 (0.425–6.767) 1.858 (0.646–5.338) 2.947 (0.69–12.59)
IVa 1.875 (0.495–7.102) 2.246 (0.808–6.24) 3.983 (0.965–16.45)
IVb 4.5 (0.73–27.73) 5.995 (1.821–19.74) 16.72 (3.609–77.49)
LKB-1-negative expression 1 1 1
Low LKB-1 expression 0.509 (0.243–1.065) 0.644 (0.413–1–007) 0.69 (0.429–1.11)
Moderate LKB-1 expression 0.309 (0.13–0.736) 0.461 (0.257–0.829) 0.575 (0.315–1.043)
Intense LKB-1 expression 0.212 (0.072–0.628) 0.348 (0.154–0.788) 0.355 (0.158–0.8)
Successful surgical resection 0.116 (0.026–0.518) 0.27 (0.156–0.467) 0.381 (0.158–0.918)
Surgical margins
R0 1 1 1
R1 (\10 mm) 0.823 (0.433–1.566) 0.854 (0.527–1.384) 1.033 (0.627–1.703)
R2 (positive) 6.226 (1.32–29.37) 2.825 (1.456–5.48) 2.77 (1.382–5.552)
No surgical resection 9.057 (1.989–41.23) 3.966 (2.225–7.071) 6.024 (3.314–10.95)
n, number of patients; RFS, recurrence free survival; OS, overall survival; OR, odds ratio; HR, hazard ratio;
CI, confidence interval; BMI, body mass index; HPV, human papillomavirus; TNM, tumor-node-metas-
tasis; LKB-1, liver kinase B1
a
Numbers in bold characters represent significant bivariate associations.
4574 J. F. Carrillo et al.

TABLE 3 Multivariate b SE p Value Exp b 95% CI

analysis of prognostic factors
associated with the event of Age (years) - 0.031 0.017 0.075 0.969 0.937–1.003
clinical recurrence using the
LKB-1-negative expressionb – – 0.002 1 –
logistic regression model (n =
201)a Low LKB-1 expression - 1.852 0.647 0.004 0.157 0.044–0.557
Moderate LKB-1 expression - 2.613 0.75 0.0005 0.073 0.017–0.319
Intense LKB-1 expression - 3.06 0.954 0.001 0.047 0.007–0.304
Perineural infiltration 1.583 0.475 0.0009 4.869 1.92–12.35
No lymphocytic infiltrationb – – \0.00001 1 –
Moderate or plaque infiltration - 2.043 1.098 0.063 0.13 0.015–1.115
Intense lymphocytic infiltration - 5.182 1.135 0.000005 0.006 0.001–0.052
TNM stage
Ib – – 0.005 1 –
II 1.063 1.23 0.387 2.896 0.26–32.29
III 1.432 1.213 0.238 4.4187 0.388–45.14
IVa 2.115 1.205 0.079 8.292 0.782–87.9
IVb 4.82 1.522 0.002 123.95 6.273–2449
Constant 2.671 1.162 0.022 14.449 –
b, regression coefficient; SE, standard error of b; p, probability value; Exp b, exponent of beta (or odds
ratio); CI, confidence interval; LKB-1, liver kinase B1; TNM, tumor-node-metastasis
a
The estimators for the final model have been adjusted for neoplasm site (–2LL 138.04; Cox & Snell r2,
0.503; Nagelkerke r2 , 0.67; Hosmer & Lemeshow Goodness-of-fit test, p = 0.177).
b
Reference category

differential prognosis between these stages and suggesting Expression of the serine/threonine LKB-1 gene, also
a requirement to consider additional prognostic factors known as STK11, has been linked to prognosis in several
such as the histopathology features and molecular markers malignancies including breast, non-small cell lung cancer,
described in this study,18 which also could help in and larynx squamous cell cancer.10–12 In most studies, a
designing new treatment strategies for SCCOC. suppressed expression status predicts a worse prognosis in
The bivariate and multivariate analyses found surgical terms of shorter DFS and OS as well as metastasis
resection to be a robust prognostic factor in terms of development.24,25
recurrence, RFS, or OS. Although most of the patients in Our analysis found that categorization of LKB-1
our cohort had intermediate- or advanced-stage disease, expression status after a four-grade scale, as previously
this fact underscores the leading role of surgical therapy, reported,11 is associated with recurrence, RFS, and OS,
especially compared with patients subjected to radiation even after adjustment for TNM stages and histopathology
alone or combined with chemotherapy as upfront therapy. factors, as reported by Brandwein et al.7 Absence of LKB-1
Brandwein-Gensler et al.7 have proposed a prognostic expression or low LKB-1 expression clearly implies worse
system based on histopathologic parameters such as the oncologic outcomes, and a clear biologic gradient was
invasion pattern, perineural and perivascular invasion, and observed.
the lymphocytic infiltration of the host’s tissue. This sys- Notably, the expression of LKB-1 is sufficiently
tem has been validated in several reports and by our group, homogeneous so that the biopsy accurately represents
with some modifications,14 proving its reproducibility and expression of the protein throughout the tumor, a fact that
validity. In the current study, however, we established a can be devised from comparison of preoperative biopsies
major impact on recurrence rate, RFS, and OS only for and histology analyses of resected specimens in patients
lymphocytic infiltration and perineural invasion, under- subjected to surgery. Findings show that LKB-1 intervenes
scoring the need to identify molecular markers that could in several cellular processes including metabolic balance,
be used as prognostic markers as well as potential biologic DNA integrity maintenance, proliferation, polarity estab-
targets. However, previous reports from our group and lishment, and interaction with tissue milieu.11,13 In our
Brandwein-Gensler et al.7 that found a significant prog- study, local and systemic spread of neoplastic cells could
nostic impact for invasion pattern, perineural invasion, and be driven by deficient expression of LKB-1, as evident by
lymphocytic infiltration regarding recurrence and OS the worst pattern of invasion (islands) and the scarcity of
included larger samples (292 and 634 cases, lymphocytic infiltrate at the histopathologic level. Invasive
respectively).7,14 islands, particularly those smaller than 15 cells, represent
LKB-1 in oral cavity carcinoma 4575

(A) (B)
1.0 1.0
Cumulative Recurrence-free Survival

Cumulative Overall Survival

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0.0 0.0
0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14
Subjects at risk Time (years) Subjects at risk Time (years)
21 13 12 10 8 3 0 0 21 14 13 10 8 3 2 1
41 26 15 11 8 5 2 2 41 29 21 12 10 5 2 2
88 44 38 32 23 14 7 3 88 49 40 35 27 14 10 6
47 17 14 9 9 6 4 3 47 24 18 12 10 6 5 4

(C) (D)
2.0 1.0

0.8
1.5
Cumulative Hazard

Cumulative Hazard 0.6

1.0
0.4

0.5
0.2

0.0 0.0
0 2 4 6 8 10 12 14 0 1 2 3 4 5 6 7 8 9 10
Time (years) Time (years)

FIG. 1 A Kaplan-Meier recurrence-free survival curves of patients D Hazard curves for any-cause mortality over time of patients
depending on the intensity of liver kinase B1 (LKB-1) expression (p = depending on the intensity of LKB-1 expression, with adjustment for
0.011). B Kaplan-Meier overall survival curves of patients depending the covariates described in Table 5. No expression (purple lines), low
on the intensity of LKB-1 expression (p = 0.08). C. Hazard curves for expression (red lines), moderate expression (green lines), intense
recurrence over time of patients depending on the intensity of LKB-1 expression (blue lines)
expression, with adjustment for the covariates described in Table 4.

the buds or sprouts at the tumor invasion front, with the recurrence rate or DFS, probably because of its confounder
neoplastic cells gaining the ability to adhere and migrate in effect exposed by advanced TNM stages, the aforemen-
the context of epithelial-mesenchymal transition. On the tioned histopathologic factors, and the LKB-1 expression
other hand, inactivation of LKB-1 promotes immune eva- status, which emerges as a new prognostic
sion and immune suppression by activating immune check determinant.25,26,28
points and by releasing proinflammatory cytokines and Our analysis included a significant number of cases with
chemokines into the tumor microenvironment. This long follow-up periods, allowing accurate prediction of the
mechanism of damage is enhanced in presence of con- prognosis for patients with SCCOCO.
comitant mutations such as KRAS but not TP53 in the case The novel prognostic factors analyzed in this study were
of NSCLC. independent, and at the same time, the TNM classification
To our knowledge, except for a report by our group in system was validated. These factors alluded mostly to
which LKB-1 expression was mediated through miR-100- intrinsic characteristics of the malignancy, which might
3p,12,26 this is the first comprehensive communication on explain the behavior of a disease that often deviates from
the association of LKB-1 and recurrence, DFS, and OS in previously established prognostic factors. As a result, we
SCCOC. Notably, surgical margin27 was found to be an were able to construct a software program that could aid in
independent determinant of prognosis in our multivariate establishing treatment intensification or even de-escalation
analysis, with very low ORs only for OS and not for of primary and adjuvant therapies in an expeditious and
4576 J. F. Carrillo et al.

1.0 1.0

>0.933 (n = 40; 2 events)

0.8 0.8
Cumulative Overall Survival

0.777-0.932 (n = 40; 7 events)

0.6 0.6

Sensitivity
0.274-0.776 (n = 41; 13 events)

0.4 0.4

0.2 0.2
0.093-0.273 (n = 40; 31 events)

< 0.092 (n = 40; 40 events)

0.0 0
0 2 4 6 8 10 12 0 0.2 0.4 0.6 0.8 1.0
Time (years) 1 - Specificity

FIG. 2 A. Kaplan-Meier overall survival curves depending on the showing the prognostic accuracy, depending on the observed and
predicted probability of recurrence estimated with the logistic expected probability of recurrence (area under the ROC curve, 0.925;
regression model described in Table 3. The predicted probability standard error, 0.018; 95% confidence interval, 0.889–0.961; p = 2.3
values are categorized by quintiles (p = 8.9 910–28). B Receiver 9 10–25).
operating characteristic (ROC) curve of all the patients in the cohort

TABLE 4 Multivariate b SE p Value Exp b 95% CI

analysis of prognostic factors
associated with recurrence-free LKB-1-negative expressionb – – 0.083 1 –
survival using Cox’s
Low LKB-1 expression – 0.301 0.249 0.227 0.74 0.454–1.206
proportional hazard model (n =
201)a Moderate LKB-1 expression – 0.729 0.339 0.032 0.483 0.248–0.939
Intense LKB-1 expression – 0.899 0.461 0.051 0.407 0.165–1.005
Perineural infiltration 1.031 0.313 0.00098 2.803 1.519–5.172
Perivascular infiltration 0.661 0.254 0.009 1.937 1.176–3.189
No lymphocytic infiltrationb – – \0.00001 1 –
Moderate or plaque infiltration – 0.588 0.266 0.027 0.556 0.33–0.936
Intense lymphocytic infiltration – 2.139 0.317 \0.00001 0.118 0.063–0.219
TNM stage
Ib – – 0.0007 1 –
II 0.763 0.583 0.19 2.146 0.684–6.73
III 1.359 0.561 0.016 3.869 1.288–11.62
IVa 1.605 0.551 0.004 4.978 1.692–14.65
IVb 2.962 0.779 0.00014 19.328 4.197–89.008
Pushing pattern of invasionb – – 0.00041 1 –
Digital pattern of invasion 0.953 0.43 0.027 2.594 1.116–6.03
Diffuse pattern of invasion 1.58 0.426 0.00021 4.854 2.105–11.19
Surgical resection – 0.881 0.436 0.044 0.414 0.176–0.975
b, regression coefficient; SE, standard error of b; p, probability value; Exp b, exponent of beta (or hazard
ratio); CI, confidence interval; LKB-1, liver kinase B1; TNM, tumor-node-metastasis
a
Final model –2 log likelihood = 816.34; p = 1.13 x10-31)
b
Reference category

precise manner. This software is available in the following mediated through miRNA-100-3p12 and could be used as a
link: http://www.incan.salud.gob.mx. In addition, under- novel biomarker to design targeted therapies against this
expression of LKB-1, demonstrated by our group, is specific miRNA in SCCOC.
LKB-1 in oral cavity carcinoma 4577

TABLE 5 Multivariate b SE p Value Exp b 95% CI

analysis of prognostic factors
associated with overall survival LKB-1-negative expressionb – – 0.064 1 –
using Cox’s proportional hazard
Low LKB-1 expression – 0.437 0.264 0.098 0.646 0.385–1.084
model (n = 224)a
Moderate LKB-1 expression – 0.178 0.35 0.611 0.837 0.422–1.662
Intense LKB-1 expression – 1.252 0.311 0.012 0.286 0.108–0.759
Perineural infiltration 1.445 0.311 \0.00001 4.241 2.305–7.803
No lymphocytic infiltrationb – – – – –
Moderate or plaque infiltration –0.803 0.275 0.004
Intense infiltration –2.268 0.313 \0.00001
TNM
Ib – – 0.00049 1 –
II 1.069 0.776 0.168 2.912 0.637–13.32
III 1.379 0.763 0.071 3.972 0.89–17.72
IVa 1.748 0.751 0.02 5.743 1.318–25.1
IVb 3.571 0.935 0.00013 35.56 5.691–222.2
Surgical margins – – 0.031 1 –
R0
R1 (\10 mm) 0.15 0.267 0.574 1.162 0.689–1.959
R2 (positive) 0.322 0.371 0.386 1.379 0.666–2.856
No surgical resection 1.311 0.443 0.003 3.711 1.558–8.838
b, regression coefficient; SE, standard error of b; p, probability value; Exp b, exponent of beta (or hazard
ratio); CI, confidence interval
a
The estimators for the final model have been adjusted for neoplasm site (final model –2 log likelihood =
146.575; p = 1.007 9 10-24).
b
Reference category

The main weaknesses of our study included the retro- Supplementary Information The online version contains
spective nature of the data and the inclusion of patients supplementary material available at https://doi.org/10.1245/s10434-
022-11544-x.
treated initially with radiotherapy because some of these
patients were not subjected to surgery afterward. The ACKNOWLEDGMENTS The authors thank to Professor Déborah
computer software can be downloaded by request to the Aleman-Hoey profusely for her English-language editorial review.
corresponding author. DISCLOSURE There are no conflicts of interest.

CONCLUSION REFERENCES

Absence or low expression of LKB-1 is associated with 1. Ferlay, J. Ervik M, Lam F, et al. Global Cancer Observatory:
poor a prognosis for patients with SCCOC. We have Cancer Today (2018). International Agency for Research on
Cancer, Lyon, France. Retrieved 4 November 2021 at IARC h
designed a new and useful model that considers ttps://gco.iarc.fr/today.
histopathology and biologic markers in association with 2. Johnson DE, Burtness B, Leemans CR, Lui VWY, Bauman JE,
previously described clinical prognostic factors in the Grandis JR. Head and neck squamous cell carcinoma. Nat Rev
TNM staging system. This model has a high predictive Dis Primers. 2020;6:92. https://doi.org/10.1038/s41572-020-002
24-3.
value in terms of tumor recurrence, RFS, and OS for 3. Stein AP, Saha S, Kraninger JL, et al. Prevalence of human
patients with SCCOC, but it requires external validation. papillomavirus in oropharyngeal cancer: a systematic review.
The prevalence of HPV in cases of SCCOC was low, Cancer J. 2015;21:138–46. https://doi.org/10.1097/PPO.
although slightly higher than previously reported, and not 0000000000000115.
4. Menezes FDS, Fernandes GA, Antunes JLF, Villa LL, Toporcov
associated with recurrence, RFS, or OS, possibly due to TN. Global incidence trends in head and neck cancer for HPV-
other underlying etiopathogenic factors. related and -unrelated subsites: a systematic review of popula-
tion-based studies. Oral Oncol. 2021;115:105177. https://doi.org/
10.1016/j.oraloncology.2020.105177.
5. Almangush A, Pirinen M, Youssef O, Mäkitie AA, Leivo I. Risk
stratification in oral squamous cell carcinoma using staging of the
eighth American Joint Committee on Cancer: systematic review
4578 J. F. Carrillo et al.

and meta-analysis. Head Neck. 2020;42:3002–17. https://doi.org/ classification system in oral cavity squamous cell cancers. Int J
10.1002/hed.26344. Cancer. 2020;146:3379–84. https://doi.org/10.1002/ijc.32720.
6. Almangush A, Mäkitie AA, Triantafyllou A, et al. Staging and 18. Amin MB, Edge SB, Greene FL, et al. American Joint Committee
grading of oral squamous cell carcinoma: an update. Oral Oncol. on Cancer, Cancer Staging Manual. 8th edn. New York, NY:
2020;107:104799. https://doi.org/10.1016/j.oraloncology.2020. Springer Science; 2017.
104799. 19. Hasegawa T, Yanagitani N, Ninomiya H. Association between
7. Brandwein-Gensler M, Teixeira MS, Lewis CM, et al. Oral the efficacy of pembrolizumab and low STK11/LKB1 expression
squamous cell carcinoma: histologic risk assessment, but not in high-PD-L1-expressing non-small-cell lung cancer. In Vivo.
margin status, is strongly predictive of local disease-free and 2020;34:2997–3003. https://doi.org/10.21873/invivo.12131.
overall survival. Am J Surg Pathol. 2005;29:167–78. https://doi. 20. Christodoulou E, Ma J, Collins GS, Steyerberg EW, Verbakel JY,
org/10.1097/01.pas.0000149687.90710.21. Calster BV. A systematic review shows no performance benefit of
8. Li Y, Bai S, Carroll W, et al. Validation of the risk model: high- machine learning over logistic regression for clinical prediction
risk classification and tumor pattern of invasion predict outcome models. J Clin Epidemiol. 2019;110:12–22. https://doi.org/10.10
for patients with low-stage oral cavity squamous cell carcinoma. 16/j.jclinepi.2019.02.004.
Head Neck Pathol. 2013;7:211–23. https://doi.org/10.1007/s121 21. Arsa L, Siripoon T, Trachu N, et al. Discrepancy in p16
05-012-0412-1. expression in patients with HPV-associated head and neck
9. Taberna M, Mena M, Pavón MA, Alemany L, Gillison ML, squamous cell carcinoma in Thailand: clinical characteristics and
Mesı́a R. Human papillomavirus-related oropharyngeal cancer. survival outcomes. BMC Cancer. 2021;21:504. https://doi.org/10.
Ann Oncol. 2017;28:2386–98. https://doi.org/10.1093/annonc/md 1186/s12885-021-08213-9.
x304. 22. Augustin JG, Lepine C, Morini A, et al. HPV Detection in head
10. Jiang L, Liang X, Liu M, et al. Reduced expression of liver kinase and neck squamous cell carcinomas: what is the issue? Front
B1 and Beclin1 is associated with the poor survival of patients Oncol. 2020;10:1751. https://doi.org/10.3389/fonc.2020.01751.
with non-small cell lung cancer. Oncol Rep. 2014;32:1931–8. h 23. Nair S, Singh B, Pawar PV, et al. Squamous cell carcinoma of
ttps://doi.org/10.3892/or.2014.3432. tongue and buccal mucosa: clinico-pathologically different enti-
11. He SS, Chen Y, Wang HZ, et al. Loss of LKB1 expression ties. Eur Arch Otorhinolaryngol. 2016;273:3921–8.
decreases the survival and promotes laryngeal cancer metastasis. 24. Lynette M, Sholl LM. Protein correlates of molecular alterations
J Cancer. 2017;8:3548–54. https://doi.org/10.7150/jca.20535. in lung adenocarcinoma: immunohistochemistry as a surrogate
12. Figueroa-González G, Carrillo-Hernández JF, Perez-Rodriguez I, for molecular analysis. Semin Diagn Pathol. 2015;32:325–33. h
et al. Negative regulation of serine threonine kinase 11 (STK11) ttps://doi.org/10.1053/j.semdp.2015.02.019.
through miR-100 in head and neck cancer. Genes Basel. 25. Calles A, Sholl LM, Rodig SJ, et al. Immunohistochemical loss of
2020;11:1058. https://doi.org/10.3390/genes11091058. LKB1 is a biomarker for more aggressive biology in KRAS-
13. Ren YH, Zhao FJ, Mo HY, et al. Association between LKB1 mutant lung adenocarcinoma. Clin Cancer Res.
expression and prognosis of patients with solid tumours: an 2015;21:2851–60. https://doi.org/10.1158/1078-0432.CCR-14-
updated systematic review and meta-analysis. BMJ Open. 3112.
2019;9:e027185. https://doi.org/10.1136/bmjopen-2018-027185. 26. Skoulidis F, Heymach JV. Co-occurring genomic alterations in
14. Carrillo JF, Carrillo LC, Cano A, et al. Retrospective cohort study non-small cell lung cancer biology and therapy. Nat Rev Cancer.
of prognostic factors in patients with oral cavity and oropha- 2019;19:495–509. https://doi.org/10.1038/s41568-019-0179-8.
ryngeal squamous cell carcinoma. Head Neck. 2016;38:536–41. h 27. Kubik MW, Sridharan S, Varvares MA, et al. Intraoperative
ttps://doi.org/10.1002/hed.23914. margin assessment in head and neck cancer: a case of misuse and
15. Sinha N, Rigby MH, McNeil ML, et al. The histologic risk model abuse? Head Neck Pathol. 2020;14:291–302. https://doi.org/10.
is a useful and inexpensive tool to assess risk of recurrence and 1007/s12105-019-01121-2.
death in stage I or II squamous cell carcinoma of tongue and floor 28. Li D, Zhou Y, Liu Y, et al. Decreased expression of LKB1
of mouth. Mod Pathol. 2018;31:772–9. https://doi.org/10.1038/ predicts poor prognosis in pancreatic neuroendocrine tumor
modpathol.2017.183. patients undergoing curative resection. Onco Targ Ther.
16. Sridharan, S., Thompson, L.D.R., Purgina, B. et al. (2019) Early 2018;11:1259–65. https://doi.org/10.2147/OTT.S154168.
squamous cell carcinoma of the oral tongue with histologically
benign lymph nodes a model predicting local control and vetting Publisher’s Note Springer Nature remains neutral with regard to
of the eighth edition of the American Joint Committee on cancer jurisdictional claims in published maps and institutional affiliations.
pathologic stage. Cancer.125:3198–207, https://doi.org/10.1002/c
ncr.32199
17. Singhavi HR, Chakrabarti S, Singh A, et al. Comparison of the
seventh and eighth editions American Joint Committee Cancer