An Increased Neutrophil-to-Lymphocyte Ratio Predicts Incomplete Response to Therapy in Differentiated Thyroid Cancer.

Background: Previously we have shown that an elevated baseline neutrophil-to-lymphocyte ratio (NLR) was associated with a high risk of recurrence in patients with differentiated thyroid cancer. The clinical significance of the longitudinal changes in NLR following treatment remained unestablished.
Methods: Adults patients with differentiated thyroid cancer were included in the study if the follow-up NLR data at 6 to 18 months after initial treatment were available. The response to treatment was categorized as excellent, indeterminate, biochemical incomplete, and structural incomplete as per guidelines of the American Thyroid Association.
Results: Among 151 patients with thyroid cancer, a significant decrease in NLR following treatment was observed in those with stage I disease, those with low risk of recurrence, and those with an excellent response to therapy. Patients with a structural incomplete response had a significant increase in NLR at follow-up (p = 0.012). On multivariate analysis, incomplete response to therapy was associated with male sex (odds ratio [OR] = 3.35), tumor size (OR = 1.63), lymph node metastasis (OR = 4.80), distant metastasis (OR = 12.95), and increased NLR (OR = 13.68). Conclusions: An increase in systemic inflammation following treatment as measured by NLR is independently associated with an incomplete response to therapy in differentiated thyroid cancer.

Introduction

Tumor-associated inflammation has been considered as an emerging hallmark of cancer 1. Chronic inflammation may elicit neoplastic development, progression, and metastatic dissemination. On the other hand, the acquisition and accumulation of genetic alterations, as well as the loss of normal physiological regulation, may induce the expression of tumor-associated antigens. Malignant transformation is therefore accompanied by inflammatory responses resulting from insufficient or inappropriate activation of innate and adaptive immunity. These inflammatory responses may paradoxically promote tumorigenesis and can have opposing effects on treatment efficacy 2. In this context, inflammatory biomarkers may aid in the prediction of patient outcome and the assessment of treatment response 3.

The peripheral blood neutrophil-to-lymphocyte ratio (NLR) is an inexpensive and widely available biomarker of the host immune system. An elevated NLR has been suggested to be related to poor prognosis in various solid cancers 4. Previously, we have shown that an elevated preoperative NLR was associated with a high risk of recurrence in patients with differentiated thyroid cancer 5. Subsequent studies confirmed that higher NLR values may be associated with a more aggressive tumor profile 6, 7. These observations suggest that although differentiated thyroid cancer is a relative indolent malignancy with an excellent prognosis, the phenomenon of tumor-promoting inflammation is also present in thyroid cancer patients.

Despite a great interest in the prognostic role of NLR and other inflammatory biomarkers in cancer patients, considerably less attention has been given to the dynamics of post-treatment changes in these biomarkers. Furthermore, the conventional staging and risk stratification systems based on the clinicopathological information at diagnosis may not predict persistent disease and recurrence accurately. The outcome of cancer patients is closely related to the response to treatment. It has been proposed that the initial risk estimates should be actively modified over time based on the response to therapy and the course of the disease 8. The purpose of the present study was to correlate the longitudinal NLR changes with the response to treatment in differentiated thyroid cancer. The results of this study may yield further insights into the possible link between inflammatory markers and the disease status in thyroid cancer.

Patients and Methods

Following approval of by the Institutional Review Board (17MMHIS035), a review of the prospectively maintained endocrine surgery registry was conducted to identify all thyroid surgeries performed at MacKay Memorial Hospital, Taipei, Taiwan, from July 2013 to December 2015. All adult patients undergoing thyroidectomy were eligible according to the inclusion criteria. The exclusion criteria were: major immunomodulatory conditions (including immunodeficiency and active infection) or chronic kidney disease stage 3 or greater (n = 25), operation for recurrent thyroid cancer (n = 16), non-differentiated thyroid cancer (n = 7), and age younger than 20 years (n = 6).

All procedures were performed under general anesthesia. The complete blood count and leukocyte differential count were routinely measured one day before surgery. The NLR was calculated by dividing the absolute neutrophil count by the lymphocyte count 9. The follow-up NLR data were obtained, if available, at 6 to 18 months after treatment when the patients were taking thyroxine suppression or replacement therapy. The difference in the NLR (ΔNLR) was calculated as the NLR at follow-up minus the NLR before surgery. The standard deviation of the NLR distribution was determined for the study cohort. Increased NLR was defined as an NLR increase greater than or equal to the standard deviation 10. Decreased NLR was defined as an NLR decrease more than or equal to the standard deviation. Otherwise, the patients were considered to have a stable NLR.

Patients with differentiated thyroid cancer underwent either lobectomy or total thyroidectomy and, if indicated, therapeutic central and/or lateral neck lymph node dissection. Unilateral prophylactic central neck dissection was performed in clinically node-negative patients 11. The decision to have radioactive iodine (RAI) therapy was at the discretion of treating physicians on the basis of clinical need. Patient management was regularly discussed at multidisciplinary team meetings. Thereafter, patients were followed up with periodic clinical examinations, biochemical tests, and imaging studies 12.

The tumor staging upon initial treatment was based on the seventh edition of the American Joint Committee on Cancer/Union for International Cancer Control tumor-node-metastasis (TNM) system. The mutational status of BRAF was routinely determined by Sanger sequencing on cytological cell blocks or surgical specimens 13. The risk of recurrence was evaluated using the modified American Thyroid Association (ATA) initial risk stratification system, which considers low-volume lymph nodal metastases as low risk 14. Patients who had a total thyroidectomy and RAI therapy were reclassified by the response to the initial therapy as per the 2015 ATA guidelines 14. In our institute, the response to therapy was typically assessed using thyroid-stimulating hormone (TSH)-stimulated thyroglobulin (Tg) values obtained 6 to 18 months after initial therapy. The restratification scheme included the results of imaging studies (neck ultrasound with or without RAI scintigraphy) and stimulated or suppressed Tg and anti-Tg antibody levels during the follow-up period and divided the patients into four categories: excellent, indeterminate, biochemical incomplete, and structural incomplete response. For those who underwent lobectomy or total thyroidectomy without RAI therapy, the response to therapy was evaluated using the dynamic risk stratification proposed by Momesso and colleagues 15.

Continuous variables are presented as medians and interquartile ranges (IQRs), while categorical variables are presented as numbers and percentages. The TNM stage, ATA risk of recurrence, and the response to therapy were categorized as ordinal variables. Differences in medians between two groups were compared with nonparametric Mann-Whitney U tests. The Wilcoxon rank signed tests were used to assess NLR differences before and after treatment 16. Trends were detected using Spearman's nonparametric correlation analysis. Logistic regression analysis was performed to determine the odds ratios (ORs) of variables with 95% confidence intervals (CIs) for an incomplete (biochemical or structural) response to therapy. Multivariate analysis using stepwise logistic regression was conducted on parameters shown to be significant in the univariate analysis. All statistical tests were two-sided at the 5% significance level. Analyses were performed using Stata statistical software (version 14.0; StataCorp, College Station, TX, USA).

Results

A total of 587 patients met the inclusion and exclusion criteria. Among these patients, 357 (61%) underwent thyroidectomy for benign thyroid lesions and 230 (39%) for differentiated thyroid cancer. Benign thyroid lesions included nodular goiter (n = 252), follicular adenoma (n = 53), Hashimoto's thyroiditis (n = 28), and Graves' disease (n = 24). During the follow-up period, 87 (24%) patients with benign lesions and 151 (66%) patients with differentiated thyroid cancer had available NLR data for analysis. These patients constituted the study cohort (Table ). At a median follow-up of 19 months, 10 patients had locoregional recurrence (neck lymph nodes, n = 9; thyroid bed, n = 1), and one patient died of locally advanced disease and lung metastases. Two additional patients died of irrelevant diseases (cervical cancer and sepsis, respectively).

Before treatment, the median NLR was 1.93 (IQR, 1.53-2.61) and 1.96 (IQR, 1.51-2.59) for the benign and cancer groups, respectively (p = 0.633). In the benign group, 72 patients had thyroidectomy for nodular goiter and 15 for follicular adenoma. As shown in Figure , there was no significant change in NLR after thyroidectomy (p = 0.113).

Among 151 patients in the cancer group, the majority (n = 145, 96%) were treated for papillary thyroid cancer, while 6 patients had follicular thyroid cancer. Interestingly, the overall NLR significantly decreased to 1.78 (IQR, 1.34-2.48) following treatment (p = 0.037). The difference was only significant in the total thyroidectomy plus RAI therapy group (n = 128, p = 0.020) but not in the lobectomy group (n = 14, p = 0.778) or total thyroidectomy only group (n = 9, p = 0.678). In parallel to the decrease in NLR, the median total white blood cell count decreased from 7100 before surgery to 6300 at follow-up (p < 0.001). There was no difference in hemoglobin levels before and after treatment (p = 0.845).

We further analyzed the changes in NLR among different subgroups. The significant decrease in NLR following treatment was observed in patients with stage I disease (p = 0.004) but not those with stage II to IV disease. Additionally, the significant decrease in NLR was observed in patients with low risk of recurrence (p = 0.014) but not those with intermediate risk (p = 0.051) or high risk (p = 0.412). Interestingly, we found a significant decrease in NLR among patients with an excellent response to therapy (p < 0.001, Figure ). Of interest, 12 patients with a structural incomplete response had a significant increase in NLR from a median of 2.01 (IQR, 1.58-2.49) before surgery to 2.73 (IQR, 1.96-3.84) at follow-up (p = 0.012).

The median ΔNLR in patients with differentiated thyroid cancer was -0.15 (IQR, -0.62 to 0.45), which was significantly different from that of the benign group (median, +0.09; p = 0.008). The ΔNLR was not associated with the TNM stage (p for trend = 0.098). However, the ΔNLR was positively associated with the ATA risk of recurrence (p for trend = 0.017) and response to therapy (p for trend < 0.001). Scatter plots are depicted in Figure .

The standard deviation of preoperative NLR was 1.08 in our cancer patients. By definition, 12 (8%) patients with a ΔNLR ≥ 1.08 had an increased NLR, and 17 (11%) with a ΔNLR ≤ -1.08 had a decreased NLR. The remaining 81% of the patients had stable NLR following treatment. As shown in Figure , the proportions of patients having an increased NLR significantly increased in those with a higher risk of recurrence or incomplete response to therapy (both p < 0.001).

To validate whether an increased NLR was an independent predictor for incomplete response to therapy, univariate and multivariate logistic regression analyses were performed. We found that male sex, larger tumor size, nodal metastasis, distant metastasis, and increased NLR independently predicted the risk of incomplete response to therapy (Table ). When the increased NLR was replaced by the ΔNLR as a continuous variable in the model, the results remained unchanged. Male sex (OR, 3.65, 95% CI, 1.13-11.77), tumor size (OR, 1.84, 95% CI, 1.14-2.96), lymph node metastasis (OR, 5.90, 95% CI, 1.55-22.49), distant metastasis (OR, 8.17, 95% CI, 1.37-48.82), and the absolute change in NLR (OR, 3.04, 95% CI, 1.49-6.19) were independent risk factors for incomplete response to therapy.

Discussion

Tumor-associated inflammatory cells are not uncommon in thyroid cancer on microscopic examination 17. Various and distinct patterns of immune infiltrates may have different prognostic implications 18. Immune dysregulation is one of the probable mechanisms involving in worse prognosis in older patients with papillary thyroid cancer 19. Recently, we reported that the expression of interleukin-8 (IL-8) was progressively upregulated during disease progression 20. IL-8 is a proinflammatory CXC chemokine which induces neutrophil chemotaxis and promotes angiogenesis. Contrary to anaplastic thyroid cancer, intratumoral neutrophil infiltration is rarely seen in the absence of necrosis. Nonetheless, neutrophil infiltration in differentiated thyroid cancer, when present, may portend an aggressive phenotype 21.

Generally, differentiated thyroid cancer is associated with a low degree of systemic inflammatory response. Based on our experience and review of the literature, a baseline NLR is not helpful to distinguish thyroid cancer from benign thyroid lesions 5, 22. However, the peripheral blood NLR consistently showed a positive correlation with tumor size among patients with differentiated thyroid cancer 5-7. It suggests that differentiated thyroid cancer exhibits a weak but distinct antigenicity which induces a certain level of systemic inflammatory response. Nonetheless, the baseline NLR appeared to have no association with disease-free survival in patients with differentiated thyroid cancer 6, 7. This may be due to the fact that patient with differentiated thyroid cancer had relatively low NLR values compared with those of patients with other solid tumors 4, or that the analyses were underpowered to detect small differences in the context of the overall good prognosis of thyroid cancer.

Depending on the response to therapy, the initial staging evaluation may not be as accurate as dynamic risk stratification in predicting patient outcome. Similarly, the dynamics of inflammatory biomarkers may be of more prognostic value than the pretreatment level. The importance of the changes in NLR after treatment, as seen in this study, is not unique to thyroid cancer. Peng et al. demonstrated that an increased NLR, but not preoperative or postoperative NLR, was an independent predictor of overall and recurrence-free survival in patients with small hepatocellular carcinoma 23. The association of the changes in NLR with survival has also been reported in other types of malignancy, such as glioblastoma, lung, and gastric cancers 24-26.

Interestingly, the dynamics of inflammatory biomarkers can also be exploited as useful surrogates for response to various therapies. An early decline of NLR was associated with favorable outcomes during targeted therapy or immune checkpoint blockade for metastatic renal cell carcinoma 27, 28. By contrast, Saied and colleagues demonstrated that increased NLR levels were associated with poor responses following hepatic artery infusions of chimeric antigen receptor-modified T cells for unresectable liver metastases 29. In borderline resectable pancreatic cancer, an increased NLR after neoadjuvant therapy was associated with worse survival 10. The increase in NLR following treatment may reflect the disease progression or indicate an inadequate modulation of the host immune response 30. Taken together, the change in NLR before and after treatment is not only a prognostic but a predictive indicator which may tailor follow-up and future therapeutic approaches.

We observed a small but significant reduction in NLR among patients with an excellent response. Strikingly, the post-treatment NLR levels in these patients were even lower than the postoperative NLR levels in patients who underwent thyroidectomy for benign thyroid lesions. A reasonable interpretation of these results is that the decline in NLR was likely caused by the effects of RAI therapy. Asymptomatic drops in white blood cell count are frequently seen with usual doses of RAI and may persist for one year 31, 32. Nonetheless, serious bone marrow suppression is rare and occurs exceptionally in patients who had both RAI therapy and external radiation therapy to bone metastases. Lymphocytes appear to be more susceptible to RAI than do neutrophils 33. As such, one would expect a rise in NLR following RAI therapy. In this respect, Demir and collaborators demonstrated that the NLR level significantly increased two months after RAI and returned nearly to baseline by six months after RAI 34. The timing of obtaining the follow-up NLR data in this study was 6 to 18 months post treatment and prior to TSH stimulation to verify the absence of disease. Based on our results, we speculate that for patients on thyroid hormone therapy who have an undetectable Tg level with the use of a sensitive Tg assay and a stable or decreased NLR at follow-up, TSH stimulation testing may be no longer necessary.

Limitations of this study include its retrospective design. About one-third of our patients were excluded from the analysis because they did not have available follow-up NLR data during the desired time frame. The number of patients with an incomplete response to therapy is relatively small (n = 34). The rule of thumb for logistic and Cox models recommends a minimum of 10 outcome events per predictor variable (EPV) to avoid the risk of overfitting. However, the validity of this thumb rule has been questioned, and the rule of 10 EPV could be relaxed, depending on the research question 35. The application of bootstrap sampling procedure in our regression analysis gave similar results (data not shown).

Furthermore, our study was limited by its relatively short follow-up period, which restricts the ability to make meaningful conclusions with respect to recurrence or mortality. A power calculation suggested that a sample size of 360 to 800 patients with a follow-up of 6 to 10 years is required for the recurrence analysis in papillary thyroid cancer 36. Nonetheless, this study supports the hypothesis that the magnitude of systemic inflammation reflects the disease burden and the status of the response to therapy in differentiated thyroid cancer. The findings reported herein offer a number of potentially fruitful avenues for future research.

In conclusion, our present study found that an increase in systemic inflammation as measured by NLR is independently associated with an incomplete response to therapy in differentiated thyroid cancer. NLR dynamics may provide a readily available biomarker to guide individualized patient care.

Table 1

Patient characteristics

		Study cohort (n = 151)
Female		124 (82%)
Age (years)		47 (37 - 56)
Body mass index (kg/m2)		23.4 (21.2 - 27.7)
Hashimoto's thyroiditis		26 (17%)
Papillary thyroid cancer		145 (96%)
BRAF mutation		112 (74%)
Treatment
	Lobectomy	14 (9%)
	Total thyroidectomy	9 (6%)
	Total thyroidectomy + RAI	128 (85%)
Tumor size (cm)		1.8 (1.2 - 2.5)
Extrathyroidal extension
	No	77 (51%)
	Microscopic	58 (38%)
	Macroscopic	16 (11%)
Multifocality		41 (27%)
Lymphovascular invasion		50 (33%)
Lymph node metastasis
	N0	70 (46%)
	N1a	68 (45%)
	N1b	13 (9%)
Distant metastasis		10 (7%)
TNM stage
	Stage I	85 (56%)
	Stage II	7 (5%)
	Stage III	36 (24%)
	Stage IV	23 (15%)
ATA risk of recurrence
	Low risk	39 (26%)
	Intermediate risk	87 (58%)
	High risk	25 (17%)
Response to therapy
	Excellent	92 (61%)
	Indeterminate	25 (17%)
	Biochemical incomplete	22 (15%)
	Structural incomplete	12 (8%)

Note: Data are given as frequencies (percentages) or medians (interquartile ranges). Abbreviations: ATA, American Thyroid Association; RAI, radioactive iodine.

Table 2

Univariate and multivariate logistic regression analyses of factors predicting an incomplete response to therapy in differentiated thyroid cancer

	Univariate			Multivariate
	Odds ratio	95% CI	p-value	Odds ratio	95% CI	p-value
Age †	1.010	0.987 - 1.032	0.396
Male sex	3.348	1.620 - 6.921	0.001	3.593	1.133 - 11.394	0.030
Body mass index †	1.056	0.977 - 1.143	0.171
Hashimoto's thyroiditis	0.631	0.248 - 1.606	0.334
Tumor size †	1.629	1.258 - 2.110	< 0.001	1.704	1.068 - 2.719	0.025
Extrathyroidal extension	4.785	2.424 - 9.448	< 0.001	1.990	0.698 - 5.673	0.198
Multifocality	2.124	1.096 - 4.115	0.026	0.558	0.171 - 1.825	0.335
Lymphovascular invasion	4.048	2.092 - 7.835	< 0.001	1.856	0.589 - 5.848	0.291
Lymph node metastasis	4.800	2.383 - 9.671	< 0.001	4.678	1.238 - 17.683	0.023
BRAF mutation	0.686	0.277 - 1.696	0.414
Distant metastasis	12.951	3.357 - 49.962	< 0.001	10.965	1.817 - 66.161	0.009
Preoperative NLR †	1.293	0.868 - 1.926	0.206
Increased NLR ‡	13.680	3.454 - 54.185	< 0.001	18.662	3.260 - 106.839	0.001

Note: Variables shown to be significant in the univariate analysis were included in the multivariate analysis. † Modeled as a continuous variable. ‡ As a binary variable (increased NLR versus stable/decreased NLR). Abbreviation: NLR, neutrophil-to-lymphocyte ratio