Intratumoral heterogeneity as measured using the tumor-stroma ratio and PET texture analyses in females with lung adenocarcinomas differs from that of males with lung adenocarcinomas or squamous cell carcinomas.

We compared intratumoral stromal proportions and positron emission tomography (PET) textural features between females and males with lung adenocarcinoma (ADC) or squamous cell carcinoma (SCC).We retrospectively evaluated 167 consecutive patients (male 122, female 45) who underwent pretreatment fluorodeoxyglucose PET/CT and surgical resection. The tumor-stroma ratios (TSRs) of primary tumors were estimated on hematoxylin-and-eosin-stained histological sections, and higher-order textural features were extracted on PET. We compared the histological and PET features between the sexes.More females than males had ADC. Age and pathological tumor size did not significantly differ between females and males. Females with ADC had more stroma-rich tumors than males with ADC (P = .016) or SCC (P = .047). In addition, some PET textural features significantly differed between females with ADC and males with ADC and SCC; short run emphasis, long run emphasis, coarseness, strength, short-zone emphasis, zone percentage and high-intensity large-zone percentage were the commonly differed textural features. However, the TSRs and PET textural features did not significantly differ between males with ADC or SCC.Intratumoral heterogeneity in females with lung ADC differs from that in males with lung ADC or SCC.

Introduction

Non-small cell carcinoma (NSCLC) is a heterogeneous disease. The biological features and treatment-related outcomes of lung adenocarcinoma (ADC) and squamous cell carcinoma (SCC) differ, although both are categorized as NSCLCs. In addition, the clinical characteristics of NSCLC differ between females and males in terms of median age at diagnosis, smoking history, predominant histological subtype, and treatment outcomes.[ Although the reasons are complex, a large and growing body of literature has emphasized the important roles of estrogen and estrogen receptors in lung cancer development.[ The effects of estrogen on the tumor microenvironment have been investigated in several types of cancer including lung cancer.[

Interplay among cancer, stromal, and immune cells and extracellular molecules of the tumor microenvironment are critical in cancer growth. Increasing evidence indicates that interactions between cancer cells and the neighboring stroma are important in cancer progression and metastasis.[ Some investigations have focused on the intratumoral stromal proportion as a key regulator of cancer biology. The clinical significance of a high stromal proportion (the so-called high tumor-stroma ratio) in terms of cancer progression has been investigated in many types of cancers including those of the esophagus, stomach, breast, colon, and lung.[

F-18 fluorodeoxyglucose (FDG) positron emission tomography (PET)/computed tomography (CT) measures tumor glucose metabolism noninvasively, and is widely used to stage NSCLC. In addition, PET textural analyses have emerged as novel tools for assessing intratumoral heterogeneity. Texture analysis involves a series of calculations that evaluate the position and intensity of a digital image and extract information about the relationship between adjacent pixels.[ Increasing evidence suggests that intratumoral heterogeneity can be characterized by PET textural analysis and that these features are related to clinical outcome such as treatment response or survival prediction in lung cancer.[

We previously reported that heterogeneity imaging parameters are significantly associated with the intratumoral stromal proportion in head and neck squamous cell carcinoma.[ We hypothesized that heterogeneity imaging parameters might reflect the tumor microenvironment, and thus differ between sexes in NSCLCs. Therefore, we measured tumor-stroma ratios (TSRs) and PET textural features of primary NSCLCs, and evaluated the differences between females and males.

Materials and methods

Patients

We retrospectively reviewed the pretreatment FDG PET/CT scans of 275 consecutive NSCLC patients who underwent surgical treatment at a single institution (Ajou University Hospital, Suwon, Korea). Those with small malignant nodules (<1 cm in diameter) and histological cancer subtypes other than adenocarcinoma (ADC) or squamous cell carcinoma (SCC) were excluded. We ultimately evaluated 167 patients with lung ADC or SCC. Our ethics committee approved this retrospective study and waived the requirement for informed patient consent.

FDG PET/CT acquisition

FDG PET/CT was performed using the Discovery STE PET/CT scanner (GE Healthcare, Milwaukee, WI). All patients fasted for at least 6 hours before FDG PET/CT; their serum glucose levels at the time of FDG injection were <150 mg/dL. Unenhanced CT was performed 60 minutes after injection of 5 MBq/kg FDG, using a 16-slice helical CT scanner (120 keV; 30–100 mA in the AutomA mode; section width=3.75 mm). Emission PET data were acquired from the thigh to the head for 3.0 minutes per frame in the 3-dimensional mode. Attenuation-corrected PET images using CT data were reconstructed by an ordered-subset expectation maximization algorithm (20 subsets, 2 iterations).

PET texture analyses

The PET texture analysis method was previously described.[ For tumor delineation, we used the gradient-based segmentation method (PET Edge) in MIM version 6.4 (MIM Software Inc., Cleveland, OH). Texture analysis for the quantification of intratumor heterogeneity on PET was performed using the Chang-Gung Image Texture Analysis toolbox (CGITA, https://code.google.com/archive/p/cgita), an open-source software package implemented in MATLAB (version 2012a; MathWorks Inc., Natick, MA).[ Of the many textural matrices that can be derived using CGITA, 27 higher-order textural features were analyzed, because these features are commonly evaluated by medical imaging researchers.[ Higher order textural features included gray level run length matrix, neighborhood gray level difference matrix, and gray level size zone matrix. After resampling with 64 discrete bins, we derived 27 higher order textural features. We also measured conventional PET parameters including the maximum standardized uptake value (SUVmax), metabolic tumor volume (MTV), and total lesion glycolysis (TLG) of the primary tumors.

Histopathology

Histological subclassification was performed by 2 pathologists (YWK and DL) based on the 2015 World Health Organization Classification of Lung Tumors.[ To perform a differential diagnosis between ADC and SCC, we conducted immunohistochemical stainings for TTF-1, Napsin A and p40. TSRs were obtained by consensus. The stromal proportion was quantified as previously described.[ In brief, the largest invasive tumor block was selected using the 1.25× objective. Subsequently, using the 10x objective, fields were only scored where both stroma and tumor were present, and the tumor cells were visible on all slides in the microscopic image field. Then the stromal proportion was estimated and scored in increments of 5 percentage points (e.g., 15%, 20%, and 25%). Representative photomicrographs are presented in Figure 1. All slides were reviewed in a blinded manner.

Figure 1

Representative photomicrographs of nonsmall cell carcinoma. (A) Female adenocarcinoma with a stromal proportion estimated as 60%. (B) Male adenocarcinoma with a stromal proportion estimated as 20%. (C) Male squamous cell carcinoma with a stromal proportion estimated as 20%.

Statistical analyses

We expressed data as means ± SDs for continuous variables and as percentages for categorical variables. All analyses were conducted using SPSS for Windows software (ver 18.0; IBM Inc., New York, NY). Differences between the 2 groups were compared using the t-test for continuous variables and the chi-square test for dichotomous variables. P values <.05 were considered to indicate statistical significance.

Results

Patient characteristics

Of the 167 patients, 45 were females (26.9%) and 122 males (73.1%) of a mean age of 63.3 ± 9.2 years. The mean diameter of resected primary tumors was 3.5 ± 1.3 cm (range, 1.2–7.0 cm). The clinical characteristics of females and males are summarized in Table 1. More males than females had smoking history. More females than males had ADC. Age, pathological tumor size, and TSR did not significantly differ between females and males. Stroma-rich and stroma-poor tumors were distinguished using the median TSR value (30). The proportion of stroma-rich tumors (TSR > 30) was significantly higher in females than in males (42.2 vs 24.6%, P = .026). We created 4 subgroups by sex and histology: females with ADC (FADC, n = 37), females with SCC (FSCC, n = 8), males with ADC (MADC, n = 58), and males with SCC (MSCC, n = 64). Figure 2 shows the differences in proportions of stroma-rich and stroma-poor tumors. We excluded the FSCC subgroup from statistical analyses because of the small number of patients. The FADC subgroup had more stroma-rich tumors than the MADC subgroup (45.9 vs 22.4%, P = .016) and the MSCC subgroup (45.9 vs 26.6%, P = .047); however, the MADC and MSCC subgroups did not significantly differ (P = .679).

Table 1

Clinical characteristics of 167 patients.

Figure 2

The difference of stroma-rich tumors among 4 subgroups. FADC = females with adenocarcinoma, FSCC = females with squamous cell carcinoma, MADC = males with adenocarcinoma, MSCC = males with squamous cell carcinoma.

Differences in PET textural features according by sex and histology

The mean SUVmax, MTV, and TLG of females were 8.6 ± 4.8, 20.7 ± 30.5, and 107.1 ± 214.7, respectively; and 11.0 ± 6.6, 30.0 ± 38.4, and 181.9 ± 246.6, for males, respectively, who had a significantly higher SUVmax than females (P = .033). The MTV and TLG did not significantly differ between the sexes.

The higher order PET textural features of the 3 subgroups (FADC, MADC, and MSCC) are summarized in Tables 2–4 (gray-level run length matrix, neighborhood gray-level difference matrix, and gray-level size zone matrix, respectively). Significant differences were evident between FADC and MADC and FADC and MSCC, but not between MADC and MSCC. The textural features that differed commonly were the short run emphasis, long run emphasis, coarseness, strength, short-zone emphasis, zone percentage and high-intensity large-zone percentage.

Table 2

Comparison of gray-level run length matrix.

Table 4

Comparison of gray-level size zone matrix.

Discussion

We found that intratumoral heterogeneity as measured using TSR and PET textural analyses significantly differed between females and males. Histopatholgically, FADC had more tumors with high proportions of stroma. Interestingly, we found no significant difference between MADC and MSCC. The PET data yielded similar results: some textural features significantly differed between FADC and MADC and FADC and MSCC, but not between MADC and MSCC.

Intratumoral heterogeneity measured by imaging modalities such as PET or CT may be associated with differences in regional tumor cellularity, proliferation, hypoxia, necrosis, and angiogenesis.[ These heterogeneous biological characteristics may be related with the tumor–stroma interaction. Cancer-associated fibroblasts (CAFs), the main cellular components of the stroma, may play a critical role in the tumor–stroma interaction. The accumulation of fibrotic extracellular matrix is induced by CAFs and the structure of the extracellular matrix can be further stimulated and altered by proteases, which degrade the stroma. Together, these cause the destruction of epithelial tissue and remodeling of the extracellular matrix, thereby promoting the invasion of tumor cells. CAFs also secrete growth factors, inflammatory factors, and angiogenic factors, all of which contribute to intratumoral heterogeneity.[ It has been suggested that increased image heterogeneity is associated with more aggressive tumor behavior, poorer response to treatment and worse prognosis.[ Because clinical characteristics between sexes in NSCLC are apparent, we hypothesized that the intratumoral heterogeneity between females and males is different. Therefore, we chose TSR as a possible histopathologic feature to compare PET textural features between females and males although little is known what the biological correlates of textural features are.

NSCLC-specific research has demonstrated that stroma in NSCLC is associated with tumor invasiveness, metastasis and poor prognosis. Nakamura et al[ showed that the expression of carbonic anhydrase IX by CAFs contributed to an increase in aggressive behavior in lung ADC with hypoxic regions. Takahashi et al[ reported that SCC with a fibrous stroma displayed higher invasive phenotype and were associated with a significantly poor prognosis.

Recently, texture analysis has been identified as a tool for “radiomics” on medical images. Recent investigations using texture analyses by FDG PET and chest CT showed that some textural features of ADC and SCC differed, thus radiomic signature can be used to distinguish lung ADC from SCC.[ Here, we found differences in PET textural features by sex and histological subtype, consistent with the different proportions of high-stroma tumors among the subgroups. Thus, tumor biology may differ by sex in addition to histological subtype; females with lung ADC differed from that of males with lung ADC or SCC. This result suggests that lung ADC and SCC may require a sex-specific PET radiomic approach.

Our study had several limitations. Patient numbers were relatively small and all had relatively early stage lung cancers; we only included surgically treated patients. We excluded females with SCC from statistical analyses because of the small number of patients; the female SCC incidence is rather low. Further studies with larger cohorts are required.

Conclusion

Intratumoral heterogeneity measured using TSR and PET textural analyses in females with lung ADC differed from that of males with lung ADC or SCC.

Author contributions

Conceptualization: Young Wha Koh, Dakeun Lee.

Data curation: Dakeun Lee.

Formal analysis: Young Wha Koh, Dakeun Lee, Su Jin Lee.

Funding acquisition: Su Jin Lee.

Methodology: Young Wha Koh, Dakeun Lee, Su Jin Lee.

Supervision: Su Jin Lee.

Visualization: Young Wha Koh, Su Jin Lee.

Writing – original draft: Young Wha Koh, Su Jin Lee.

Writing – review & editing: Dakeun Lee, Su Jin Lee.

Su Jin Lee orcid: 0000-0002-2096-6115.

Table 3

Comparison of neighborhood gray-level difference matrix.