Dose-volume derived nomogram as a reliable predictor of radiotherapy-induced hypothyroidism in head and neck cancer patients

A significant proportion of patients with head and neck squamous cell carcinoma (HNSCC) are treated with radiation therapy at some point during their disease. Despite the use of modern techniques and advances in radiotherapy delivery, the dose at organs at risk (OARs) in the head and neck region is not negligible, and may result in significant side effects, such as oral mucositis, xerostomia, osteoradionecrosis, radiodermatitis and dysphagia.1 The thyroid gland, as an OAR, is frequently involved in the treatment field due to its midline neck position, and the dose it receives often exceeds 50 Gy.2 Thus, one of the potential consequences of neck irradiation is development of various thyroid disorders, mainly hypothyroidism (HT), which occurs in up to 50% of patients treated with radiotherapy for HNSCC.3, 4, 5 HT is a subacute or chronic side effect of radiotherapy, with the time to development usually between 6 and 24 months after treatment. However, HT is often unpredictable and can develop at any time after completion of radiotherapy, with data showing a rising incidence with longer follow-up periods.6

Despite the growing body of evidence regarding possible damage to the thyroid during radiotherapy, there is still no generally accepted consensus on dose-volume parameters and constraints that could be used to spare the thyroid gland.2,7, 8, 9 Moreover, the threshold dose on the thyroid gland, and the relationship between the radiation dose and induction of HT as the most important thyroid disorder resulting from radiotherapy is not well established.10, 11, 12

The aim of this study was to determine the possible predictive value of various dosimetric (dose-volume) parameters and clinical characteristics on the development of HT in patients with HNSCC treated with radiotherapy. Also, an additional effort was made to identify a subgroup of patients at high risk for HT development, using a combination of different factors (dose-volume nomogram).

The mean age of the study population was 59.8 ± 9.7 years and 137 (87.8%) were males. After a median follow-up of 23.0 (12.0–38.5) months at the study cut-off date, 70 (44.9%) patients developed HT (Figure 1, Table 1). The differences in general and dosimetric parameters between patients with and without HT are presented in Tables 2 and 3.

Figure 1

Since no significant differences in general characteristics between patients with and without HT were found, ROC analysis was performed for all dosimetric parameters (Table 4), and the strongest predictive variable was added into the multivariate model. Overall, the best predictive factor was VS65 with an area under curve (AUC) of 0.684 (95% confidence interval (CI) 0.600–0.768). Other predictive factors were Dmin - AUC of 0.673 (95% confidence interval (CI) 0.589–0.757) and V50 -AUC of 0.630 (95% confidence interval (CI) 0.543–0.717), while total thyroid volume (TTV) was analysed separately. Thyroid volume was analysed separately because it is an independent variable and is not associated with other dosimetric parameters.

Firstly, the interaction between these mentioned four parameters was analysed. A strong correlation was found between all variables. The strongest positive correlation was found between V50 and Dmin, and between VS65 and TTV (Table 5). In order to analyse their independent association with HT, variables were logarithmically transformed, after which a multivariate backwards conditional stepwise Cox regression was employed. In a multivariate model, HT was associated with lower TTV (OR 0.312, 95% CI 0.112–0.868, P = 0.026) and Dmin (OR 9.832, 95% CI 1.894–108.082, P = 0.042).

Secondly, we constructed a regression equation entitled hypothyroidism risk score (HRS) comprised of TTV and Dmin:

HRS substantially increased the accuracy of TTV in predicting HT (Figure 2). Log Rank X² of TTV was 8.73 (P = 0.003) and it increased to 25.68 (P < 0.001) when we replaced TTV with HRS. HRS had an AUC of 0.709 (95% CI 0.627–0.791) and a cut-off of > 7.1 had a sensitivity of 75.7% and a specificity of 64.0%. HT occurred in 13 (20.0%) patients with a score < 7.1 and in 57 (62.6%) patients with a score > 7.1.

Figure 2

Finally, we aimed to demonstrate the actual improvement in predictive accuracy when we applied HRS to a previously established predictive factor like V50. In our cohort of patients, V50 was also associated with HT as previously reported, but this association was not independent when considering other variables. V50 was capable of delineating only 14 patients with V50 < 60% as those with a lower risk of HT (Figure 3A). When we employed HRS in patients with V50 > 60%, additional 52 patients were categorized as those with lower risk of HT (Figure 3B).

Figure 3

Severe side effects may occur in the treatment field in different organs at risk in patients with HNSCC who are receiving (chemo)radiotherapy with curative intent. The dose-volume constraints for the thyroid gland as an OAR are a matter of debate and not clearly defined.10 Although the introduction of IMRT has marked a great advance in radiotherapy for HNSCC patients, with its ability to spare normal surrounding tissue from high radiation doses while delivering a highly conformal dose to the tumor, a significant radiation dose is still unavoidably delivered to the thyroid gland.2 Furthermore, IMRT used for locally advanced head and neck cancer may increase the radiation dose to the thyroid compared with conventional conformal radiotherapy unless appropriate thyroid dose constraints are used.3 Due to its proximity to other structures that are commonly included in the radiation field of these patients, sparing the thyroid gland can compromise final oncological outcome, especially in cases of suboptimal doses to areas which are at high-risk for recurrence. Thus, a significant proportion of patients receiving IMRT develops thyroid dysfunction, with HT being the most common complication. We observed a high rate of HT (44% of patients) after a 2-year-follow-up period, which is consistent with previous reports.4, 5, 14

The relationship between the radiation dose delivered to the thyroid and the development of radiation-induced HT has been a matter of debate. Although previous studies have suggested that higher radiation doses to the thyroid gland are associated with a higher HT rate, a clear threshold radiation dose has not been defined.15, 16 Recent studies on patients with Hodgkin’s lymphoma, breast cancer and head and neck cancer, proposed different dose-volumetric parameters for the prediction of HT development, mostly including the proportion of thyroid volume receiving some X dose (VX). In these studies, the defined VX threshold parameters ranged from V10 to V50.2,4,11,17, 18, 19, 20 However, the proposed proportion of thyroid volume that could receive VX dose was different among studies, making it difficult to draw strong conclusions.

In the study by Kim et al. V45 was the only parameter that independently predicted HT in multivariate analysis, and V45 of 50% was a threshold value.4 In contrast, Akgun et al. found V30 to be useful in evaluating the risk of HT, since it was a statistically significant predictor of HT development.17 In a large meta-analysis that examined dose-response data in 4 studies including a total of 1027 patients10, final analysis showed that although there was a radiation dose-response relation with a 50% risk of HT at a dose of 45 Gy, there was considerable variation in the dose-response between studies (the dose of 50% radiotherapy induced HT probability varied from 33 Gy to 65 Gy), which was explained by the differences in follow-up. All authors agreed that although the threshold dose at which most patients will develop HT is still not defined, even low radiation doses have the potential to induce thyroid dysfunction. Thus, it is currently not possible to completely eliminate the risk of HT in patients treated with radiotherapy.

In the present study, V50 was found to be the best predictive factor for the development of HT among VX parameters. V50 is commonly reported in the literature as the parameter which is the most valuable in this setting; however, threshold V50 levels for the development of HT differ significantly between studies. The results of the present study are similar to those by Ling et al. which analysed radiotherapy dose parameters that corresponded with radiotherapy-induced thyroid dysfunction in 102 patients.8 Their data showed that the incidence of HT was reduced when achieving D50 < 50 Gy, V50 <50%, and a mean dose of < 54.58 Gy. Sachdev et al. also reported that V50 was highly correlated with HT development, while other dosimetric parameters did not reach statistical significance. After a 50-month follow-up, the total rate of HT was 33%, and the proposed threshold was V50 > 60%.11 A V50 threshold of <75% was proposed by Lin et al. as a useful guideline to avoid HT.12 In the paper from Xu, the threshold level of V50 was set to 54.5%.20 When all these results are analysed collectively, it can be concluded that the rate of HT is small in patients receiving < 50 Gy. In this study, it was shown that by employing our novel scoring system (HRS) it is possible to additionally stratify a cohort of patients with V50 > 60%, in order to predict the risk of HT development more precisely. Thus, this dose-volume derived nomogram could be a valuable tool in addition to the presently used parameters in everyday clinical practice.

VS is another less used dosimetric parameter which is important in estimating the amount of spared tissue. In paper by Lee et al. VS60 and VS45 of the thyroid were significant predictors of biochemical hypothyroidism.21 Freedom from biochemical hypothyroidism was longer for those whose VS60 was ≥ 10 cm³. Furthermore, in a paper by Chyan et al. VS30 Gy, VS40 Gy, and VS50 Gy were dosimetric parameters found to be statistically significant predictors of HT development.7 In the present study, VS65 was independently associated with HT in multivariate analysis, and therefore could be a useful predictive factor for radiation-induced hypothyroidism.

Data from previous studies examining the role of chemotherapy on HT development are inconclusive. Some studies have found that concurrent chemotherapy application increased the probability of HT development.8,19,22 In the study from Luo et al. chemotherapy was one factor contributing to the development of radiotherapy induced hypothyroidim, and was selected as one of the variables (risk factors), using the least absolute shrinkage and selection operator (LASSO).19 On the other hand, in the literature-based meta-analysis, which included 15 studies, chemotherapy did not affect the risk of hypothyroidism.10 However,these discrepancies may arise due to different chemotherapy doses and sequences and multi-agent chemotherapy regimens. We did not find a correlation between chemotherapy and HT, or an association between other patient and clinical characteristics (age, gender, extent of surgery, tumor site, thyroid surgery, and TNM status) and HT development. Earlier studies found that patients that underwent thyroid lobectomy had a higher incidence of HT.23 However, prior lobectomy was not found to be a predictive factor of HT in our study.

This study has several limitations including its retrospective design and relatively small number of patients. On the contrary, we believe that developing a formula that determines the risk of hypothyroidism based on the radiated thyroid volume and the dose received is of great interest. In the end, we would like to comment on the statistics used in this paper. Significant correlation exists between all dosimetric variables, making classical multivariate analysis impossible. We were aware of the fact that LASSO statistics was the most appropriate method to do the multivariate analysis in this case19. It might have been the most appropriate method from mathematical point of view, but it would substantially impair clarity of the results from a clinical point of view. We have performed an inferior type of LASSO statistics and decided to choose only one variable from each group which showed the strongest correlation with the onset of HT in univariate (ROC) analysis. Afterwards, we made a classical multivariate regression in order to calculate regression equation. Hence, we have made a less sophisticated variant of LASSO statistic in order to get clearer results from clinical point of view.

The nomogram presented in this paper can help in treatment decision-making, especially in HNSCC patients with a relative indication for post-operative radiotherapy. Omitting radiotherapy in some of these patients with low-risk disease (i.e. histologically negative neck, clinically negative contralateral neck in N1/N2a disease from well lateralized and small tumors) eliminates the unnecessary risk of radiation-induced side effects, including HT. In such cases, risk of radiation-induced HT predicted with the help of the nomogram (together with dosimetric constraints for other OAR) could be an important factor in treatment planning, in order to determine the best strategy for each individual patient.

The volumes and doses in head and neck radiotherapy depend mainly on the localization and extension of the local tumor as well as the levels of lymph node involvement. It is important to emphasize that sparing thyroid gland is not an option in cases when control of the tumor may be compromised. In these clinical circumstances, based on proposed nomogram, follow-up could be adjusted in those that are at highest risk for early HT development, which could lead to earlier management with hormone replacement therapy and subsequent higher quality of life.

In conclusion, the thyroid gland as an OAR remains a gray zone in radiotherapy. The thyroid gland is often neglected and has no priority in contouring, and because of its proximity to the tumor bed it has a high exposure to radiation, especially when it is used as primary therapy. Our comprehensive analysis of dose volume parameters suggests that proposed nomogram (HRS) may be a useful tool in predicting radiation-induced hypothyroidism, which may aid in the treatment planning process. Thyroid sparing may be optimized by using V50 < 60% as a dose-volumetric threshold when possible. In cases when V50 > 60% HRS may be helpful in predicting HT risk more precisely. Dose volume parameters should be incorporated into routine clinical practice. Although thyroid sparing should never compromise tumor coverage, hypothyroidism, as the most common manifestation of thyroid dysfunction should be anticipated and treated appropriately. These results should be tested in other studies and if validated through prospectively designed trials, could be incorporated in treatment planning and decision making in a subgroup of patients which are at low risk of recurrence when omitting radiotherapy.