Radiation-induced hypothyroidism (RHT) is a common long-term complication for nasopharyngeal carcinoma (NPC) survivors. A model using clinical and dosimetric factors for predicting risk of RHT could suggest a proper dose-volume parameters for the treatment planning in an individual level. We aim to develop a multivariable normal tissue complication probability (NTCP) model for RHT in NPC patients after intensity modulated radiotherapy (IMRT) or volumetric modulated arc therapy (VMAT). The model was developed using clinical data and dose-volume data of thyroid and pituitary gland based on a standard backwards stepwise multivariable logistic regression analysis, then internally validated using 10-fold cross validation. The final NTCP model consisted of age, pre-treatment thyroid stimulating hormone (TSH) and mean thyroid dose. The model performance was good with an area under the receiver operating characteristic curve of 0.749 on an internal (200 patients) and 0.812 on an external (25 patients) validation. The mean thyroid dose at ≤ 45Gy was suggested for treatment plan owing to an RHT incidence of 2% versus 61% in the > 45Gy group.
Research Article
A multivariable normal tissue complication probability model for predicting radiation-induced hypothyroidism in nasopharyngeal carcinoma patients in the modern radiotherapy era
https://doi.org/10.21203/rs.3.rs-1956532/v1
This work is licensed under a CC BY 4.0 License
Version 1
posted
You are reading this latest preprint version
nasopharyngeal carcinoma
normal tissue complication probability
radiation-induced hypothyroidism
radiotherapy
predictive model
Nasopharyngeal carcinoma (NPC) is an endemic disease in Thailand. According to GLOBOCAN 2020, there are 2,316 new cases and 1,482 deaths with an annual incidence of 2.6/100,000 for males and 1/100,000 for females [1, 2]. Radiation therapy (RT) with intensity-modulated radiotherapy (IMRT) and volumetric modulated arc therapy (VMAT) technique is considered as the standard treatment for NPC [3–5].
Despite these modern RT techniques, normal tissues in the head and neck region are inevitably exposed to radiation. One of the most common late complications in NPC survivors is radiation-induced hypothyroidism (RHT) that causes various degree of symptoms, including fatigue, mood changes, daytime sleepiness, weight gain, cold intolerance, constipation, dyspnea on exertion, edema, dry skin, difficulty concentrating, and increased risk of cardiovascular disease. Therefore, lifelong thyroid hormone supplementation is recommended for symptomatic and asymptomatic high-risk patients but it results in an inconvenience, poor compliance, and frequent hospital visits and investigations [6–8].
The incidence of RHT ranged 22.4–53.9% in head and neck cancer patients and was associated with thyroid exposure dose [9–18], thyroid volume [9–10, 11, 14–18], pituitary exposure dose [12], female gender [9–12], age [9–10], chemotherapy [12], and pretreatment thyroid stimulating hormone (TSH) [11]. Several studies reported the multivariable normal tissue complication model (NTCP) model for predicting RHT in head and neck cancer patients [14, 15, 17, 18], but only two were studied in NPC patients [12, 16]. The clinical and dosimetric data were combined to create a model for risk assessment using logistic regression analysis. Luo et al. studied in 174 NPC patients treated with three-dimensional conformal RT (3D-CRT) or IMRT technique and reported that the predictive factors included female gender, use of chemotherapy, and dose to thyroid and pituitary gland [12]. The other study included 69 NPC patients with 100% IMRT use demonstrated that significant risk factors were thyroid volume and mean dose [16]. Although these studies reported a good model performance with an area under the receiver operating characteristic (AUROC) curve more than 0.7, the generalizability was limited due to small dataset, different disease characteristics, inhomogeneous radiation technique, and unique treatment approach varied among institutions.
Therefore, this study aims to develop a new NTCP model that integrates clinical and dosimetric data from a uniform VMAT or IMRT technique in Thai population. Furthermore, we compared our new model with the previously published models.
We retrospectively collected data on NPC patients treated with definitive radiotherapy using the VMAT or IMRT technique 66-70Gy in 30–35 fractions with or without chemotherapy, between January 2010 and April 2019. Inclusion criteria included pathologically proven NPC, age ≥ 18 years, normal baseline TSH and free thyroxine (FT4), and at least 2 years of follow-up or sooner if RHT developed. Exclusion criteria were history of preexisting thyroid disease, abnormal or no baseline TSH and FT4, history of thyroid surgery, and history of I-131 or radiation at the neck. This study was approved by the Institutional review board (IRB no. 836/62).
Outcome measurements
The primary endpoint was RHT, defined in three ways; subclinical hypothyroidism (high TSH with normal FT4), primary hypothyroidism (high TSH with low FT4) and central hypothyroidism (low or normal TSH with low FT4), according to our institutional reference range (Supplement 1), regardless of symptoms.
Radiation therapy and dosimetric analyses
Two or three planning target volumes (PTVs) were used in the radiotherapy planning. For the three PTVs plan, PTV-high risk (PTV-HR) was defined as gross tumor and gross pathologic lymph nodes (LN) that received doses of 66-70Gy. PTV-intermediate risk (PTV-IR) was defined as the subclinical disease that received prophylactic doses of 60-66Gy. PTV-low risk (PTV-LR) was defined as the elective LN region (bilateral cervical LN level II-V, VII) that received doses of 50-56Gy. For the two PTVs plan, subclinical disease was included in PTV-HR. The prescribed doses were delivered in 30–35 fractions at the rate of 5 fractions per week.
Thyroid and pituitary gland (pituitary fossa) were manually re-contoured in CT non-contrast images for all patients, using an Eclipse planning system (Varian Medical Systems, version 15.6, Palo Alto, CA, USA). Dosimetric data were obtained from a treatment planning system consisting of: thyroid volume, the maximum, mean and minimum dose to thyroid gland (Dmax, Dmean, Dmin), the percentage of thyroid volume receiving at least 25, 30, 35, 40, 45, 50, 55 and 60Gy (V25, V30, V35, V40, V45, V50, V55 and V60), the absolute thyroid volume spared from 45, 50, 55 and 60Gy (VS45, VS50, VS55 and VS60), which can also be calculated as VSx = \(\frac{(100-Vx (\%\left)\right)}{100}\)x thyroid volume (cm3), as well as the maximum, mean and minimum dose to pituitary gland (Pmax, Pmean and Pmin). In addition, the whole dose-volume histogram was reduced to a generalized equivalent uniform dose (gEUD) (Supplement 2). In practice, we did not apply specific dose constraints to the thyroid and pituitary gland.
Chemotherapy
Concurrent platinum-based chemotherapy (cisplatin or carboplatin) was administered during radiation in stage II-IV patients. Neoadjuvant or adjuvant chemotherapy regimens were based on physician discretion, including platinum-based chemotherapy (cisplatin or carboplatin), infusion fluorouracil (5FU), paclitaxel, or gemcitabine, given at 3 or 4-week intervals for a maximum of three cycles.
Statistical analyses
Clinical and dosimetric features between patients with or without RHT were compared using a χ2 test or t-test. We used three steps in developing our NTCP model. First, univariate analysis was initially performed for all variables, followed by selecting significant variables at p-value of < 0.10 to be considered in the multivariate analysis. A standard backwards stepwise logistic regression method was used to obtain the NTCP model:
with S(x) = β0 + β1x1 + β2x2 + … + βnxn,
where β0 is a constant and β1, …, βn are the logistic regression coefficients of the variable x1, x2, …, xn, respectively. Pearson correlations were also used to investigate the relationship between the variables.
Second, our model performance was investigated by calculating the area under the curve of receiver operating characteristic (AUROC) curves. An expected AUROC ≥ 0.7 was required in order to have a good predictive model. The Brier score was also used to evaluate the performance of our model that predicts a binary outcome. Finally, model calibration using the Hosmer-Lemeshow goodness-of-fit test was used, where a non-significant P-value > 0.05 indicated good prediction for our model compared with the observed cases.
Third, internal model validation was performed with a k-fold cross-validation. We used k = 10, with 9 folds as a training dataset to fit the model, and the remaining 1-fold as a test dataset to internally validate the model. Eventually, the external validation was performed in consecutive patients who were treated after the model development period. Statistical analyses were done using STATA 15.1 (StataCorp LLC, College Station, TX, USA).
Of the total 498 NPC patients, 298 were excluded due to less than a 2-year follow-up (148), no pretreatment (99) or an abnormal pretreatment thyroid function test (37), a preexisting thyroid disease (13), or a previous neck radiation (1). The remaining 200 patients (144 males and 56 females) with a mean age of 48.7 years (with a range 18 to 83 years) were analyzed. The RHT group tended to have younger ages, earlier T stages (T1-T2), more advance N stages (N2-3), and higher pretreatment TSH. The patient and treatment characteristics are shown in Table 1.
|
All (N = 200) |
No RHT (N = 74) |
RHT (N = 126) |
||
|---|---|---|---|---|
|
Age (year) |
Mean (SD) |
48.7 (11.5) |
51 (11.7) |
47.4 (11.2) |
|
Gender |
||||
|
- Male |
N (%) |
144 (72) |
52 (70.3) |
92 (73) |
|
- Female |
N (%) |
56 (28) |
22 (29.7) |
34 (27) |
|
T stage - T1-2 - T3-4 |
N (%) N (%) |
130 (65) 70 (35) |
41 (55.4) 33 (44.6) |
89 (70.6) 37 (29.4) |
|
N stage - N0-1 - N2-3 |
N (%) N (%) |
64 (32) 136 (68) |
30 (40.5) 44 (59.5) |
34 (27) 92 (73) |
|
Group Stage - Stage I - Stage II - Stage III - Stage IV |
N (%) N (%) N (%) N (%) |
3 (1.5) 33 (16.5) 107 (53.5) 57 (28.5) |
0 (0) 12 (16.2) 37 (50) 25 (33.8) |
3 (2.4) 21 (16.6) 70 (55.6) 32 (25.4) |
|
WHO type |
||||
|
- 1 : Keratinizing - 2 : Non-keratinizing |
N (%) |
2 (1) |
1 (1.4) |
1 (0.8) |
|
2A: Differentiated |
N (%) |
32 (16) |
12 (16.2) |
20 (15.9) |
|
2B: Undifferentiated |
N (%) |
165 (82.5) |
60 (81.0) |
105 (83.3) |
|
- Other |
N (%) |
1 (0.5) |
1 (1.4) |
0 (0) |
|
RT Technique - IMRT - VMAT - IMRT + VMAT |
N (%) N (%) N (%) |
36 (18) 159 (79.5) 5 (2.5) |
12 (16.2) 59 (79.7) 3 (4.1) |
24 (19) 100 (79.4) 2 (1.6) |
|
Pre-treatment TFT |
Mean (SD) Mean (SD) |
1.3 (0.2) 1.7 (0.9) |
1.3 (0.2) 1.4 (0.9) |
1.3 (0.2) 1.9 (0.8) |
|
- FT4 (ng/dl) - TSH (µU/ml) |
||||
|
Post-treatment TFT (worst during 2-year follow-up) |
||||
|
- FT4 (ng/dl) |
Mean (SD) |
1 (0.2) |
1 (0.2) |
1 (0.2) |
|
- TSH (µU/ml) |
Mean (SD) |
9.5 (19.4) |
2.2 (1) |
13.9 (23.4) |
|
Thyroid gland - Volume (cm3) |
Mean (SD) |
14.8 (7.3) |
15.1 (6.1) |
14.6 (7.9) |
|
- gEUD (Gy) |
Mean (SD) |
47.2 (6.9) |
45.5 (6.5) |
49.2 (6.7) |
|
- Dmean (Gy) |
Mean (SD) |
52.9 (5.3) |
51 (5.3) |
54 (5.1) |
|
- Dmin (Gy) |
Mean (SD) |
34.4 (5.9) |
32.6 (6.1) |
35.4 (5.5) |
|
- Dmax (Gy) |
Mean (SD) |
66.3 (6.8) |
64.1 (6.1) |
67.6 (6.8) |
|
- V35 (%) |
Mean (SD) |
96.8 (7.2) |
94.6 (10) |
98.2 (4.6) |
|
- V40 (%) |
Mean (SD) |
91.5 (12) |
87.8 (15) |
93.7 (9.2) |
|
- V45 (%) |
Mean (SD) |
82.3 (16.7) |
77.2 (19.2) |
85.3 (14.4) |
|
- V50 (%) |
Mean (SD) |
67.9 (22) |
61.9 (23.6) |
71.4 (20.3) |
|
- V55 (%) |
Mean (SD) |
44.2 (27) |
37.1 (25.9) |
48.3 (26.9) |
|
- V60 (%) |
Mean (SD) |
16.3 (20.2) |
10.3 (15.3) |
19.8 (21.9) |
|
- VS45 (cm3) |
Mean (SD) |
2.7 (2.8) |
3.4 (2.9) |
2.3 (2.6) |
|
- VS50 (cm3) |
Mean (SD) |
4.8 (3.9) |
5.7 (3.8) |
4.3 (3.7) |
|
- VS55 (cm3) |
Mean (SD) |
8.3 (5.4) |
9.4 (4.8) |
7.7 (5.6) |
|
- VS60 (cm3) |
Mean (SD) |
12.4 (6.5) |
13.5 (5.5) |
11.8 (6.9) |
|
Pituitary gland |
||||
|
- Pmean (Gy) - Pmin (Gy) - Pmax (Gy) |
Mean (SD) Mean (SD) Mean (SD) |
53.6 (16.7) 45.3 (19.5) 61.3 (14.4) |
55.5 (17.8) 46.7 (20.8) 62.8 (14.9) |
52.4 (15.9) 44.5 (18.7) 60.5 (14.1) |
|
Chemotherapy |
N (%) |
194 (97) |
72 (97.3) |
122 (96.8) |
Abbreviations : RHT = Radiation-induced hypothyroidism; WHO = World Health Organization; RT = Radiotherapy; IMRT = Intensity-modulated radiotherapy; VMAT = Volumetric Modulated Arc Therapy; TFT = Thyroid function test; TSH = Thyroid-stimulating hormone; FT4 = Free thyroxine; Dmin = Minimal dose of the thyroid (Gy); Dmean = Mean dose of the thyroid (Gy); Dmax = Maximum dose of the thyroid (Gy); Pmin = Minimal dose of the pituitary (Gy); Pmean = Mean dose of the pituitary (Gy); Pmax = Maximum dose of the pituitary (Gy); gEUD = generalized equivalent uniform dose (Gy); Vx = Volume of thyroid gland receiving x Gy or more (%); VSx = absolute thyroid volume spared from x Gy or more (cm3)
With a median follow-up time of 29 months (with a range 5-120 months), RHT occurred in 126/200 patients (63%) at a median latency period of 18 months (5–97 months). The 1, 2 and 3-year cumulative incidences of RHT were 17.5%, 38.5% and 50.5%, respectively. The most common type of RHT was subclinical hypothyroidism (108/126, 85.7%), followed by primary hypothyroidism (11/126, 8.7%), and central hypothyroidism (7/126, 5.6%). There were patients with subclinical hypothyroidism who progressed to primary hypothyroidism (15/108, 13.9%) and central hypothyroidism (1/108, 0.9%) after a longer follow-up period. Thyroid hormone supplementation was prescribed to 77/126 patients (61.1%). Generally, thyroid hormone replacement therapy was prescribed when TSH level ≥ 10 µU/mL or FT4 ≤ 0.8 ng/dL. The median time from complete radiation to thyroid hormone supplementation was 26 months (with a range 6-108 months).
The mean dose, Dmin, Dmax, and Vx of the thyroid gland in RHT group were significantly higher than non-RHT group, while the VS45, VS50, and VS55 were significantly lower. The maximum correctly classified cutoff for the mean thyroid dose to differentiate between patients with no RHT and RHT group was 45Gy. The incidence of RHT was 2% and 61% in the thyroid Dmean of ≤ 45Gy and > 45Gy subgroups, respectively. The gEUD was also a significant predictor for RHT. Compared to the mean thyroid dose, all patients had a lower gEUD with a mean absolute difference of 5Gy (with a range range 0–8Gy).
Other dosimetric parameters, such as thyroid volume and pituitary exposure dose, were not found to be significant in this study. The mean thyroid volume was 14.8 cm3 (15.1 vs 14.6 cm3, with a non-significant P-value = 0.64) and the mean pituitary dose was 53.6 Gy (55.5 vs 52.4 Gy, with a non-significant P-value = 0.22). In subgroup analysis, patients with central RHT received a significantly higher pituitary dose that those without central RHT (a median of Pmean 72.3 Gy vs 57.4Gy with a significant P-value = 0.009, and a median of Pmax 74.7Gy vs 63.5Gy with a significant P-value = 0.019) as shown in Supplement 3.
NTCP model development
In the univariate logistic regression analyses, age, T and N stage, pre-treatment TSH, Dmean, Dmin, Dmax, V35, V40, V45, V50, V55, V60, VS45, VS50, and VS55 of the thyroid gland were significantly associated with RHT at any time point. In the multivariate logistic regression analysis, age, pre-treatment TSH ,and Dmean were the only significant parameters, as shown in Table 2.
|
Variables |
Reference |
Univariate |
Multivariate* |
|||||||
|---|---|---|---|---|---|---|---|---|---|---|
|
Odds Ratio |
95%CI |
P-value |
Odds Ratio |
Regression coefficient |
95%CI |
P-value |
||||
|
Age (year) |
0.97 |
0.95-1.00 |
0.035 |
0.97 |
-0.03 |
-0.06 to -0.002 |
0.034 |
|||
|
Gender |
Male |
0.87 |
0.46–1.65 |
0.676 |
||||||
|
T stage |
T1-2 |
0.52 |
0.28–0.94 |
0.030 |
||||||
|
N stage |
N0-1 |
1.84 |
1.0-3.39 |
0.048 |
||||||
|
Pre-treatment FT4 |
1.61 |
0.37–7.06 |
0.525 |
|||||||
|
Pre-treatment TSH |
2.44 |
1.61–3.71 |
< 0.001 |
2.49 |
0.91 |
0.48–1.33 |
< 0.001 |
|||
|
Thyroid volume (cm3) |
0.99 |
0.95–1.03 |
0.643 |
|||||||
|
gEUD (Gy) |
1.09 |
1.04–1.14 |
< 0.001 |
|||||||
|
Dmean (Gy) |
1.12 |
1.05–1.18 |
< 0.001 |
1.12 |
0.12 |
0.05–0.19 |
0.001 |
|||
|
Dmin (Gy) |
1.09 |
1.03–1.14 |
0.001 |
|||||||
|
Dmax (Gy) |
1.09 |
1.04–1.14 |
0.001 |
|||||||
|
V35 (%) |
1.07 |
1.03–1.12 |
0.002 |
|||||||
|
V40 (%) |
1.04 |
1.02–1.07 |
0.002 |
|||||||
|
V45 (%) |
1.03 |
1.01–1.05 |
0.001 |
|||||||
|
V50 (%) |
1.02 |
1.01–1.03 |
0.004 |
|||||||
|
V55 (%) |
1.02 |
1.01–1.03 |
0.005 |
|||||||
|
V60 (%) |
1.03 |
1.01–1.05 |
0.002 |
|||||||
|
VS45 (cm3) |
0.87 |
0.78–0.96 |
0.009 |
|||||||
|
VS50 (cm3) |
0.92 |
0.85–0.98 |
0.022 |
|||||||
|
VS55 (cm3) |
0.94 |
0.89–0.99 |
0.032 |
|||||||
|
VS60 (cm3) |
0.96 |
0.92-1.00 |
0.078 |
|||||||
|
Pmean (Gy) |
0.99 |
0.97–1.01 |
0.219 |
|||||||
|
Pmin (Gy) |
0.99 |
0.98–1.01 |
0.456 |
|||||||
|
Pmax (Gy) |
0.99 |
0.97–1.01 |
0.282 |
|||||||
|
Chemotherapy |
Yes |
0.85 |
0.15–4.74 |
0.850 |
||||||
|
Constant |
-5.62 |
-9.33 to -1.901 |
0.003 |
|||||||
Abbreviations: CI = confidence interval; TSH = Thyroid-stimulating hormone; Dmean = Mean dose of the thyroid (Gy); Pmean = Mean dose of the pituitary (Gy); Vx = Volume of thyroid gland receiving x Gy or more (%); VSx = absolute thyroid volume spared from x Gy or more (cm3).
* The multivariate results were obtained using a requirement of a P-value < 0.10 from the univariate and backward stepwise logistic regression analyses to select the final model.
The multivariable NTCP model for RHT is given by the following equation:
with S(x) = -5.62 + (-0.03×age(yrs)) + (0.91×pre-treatment TSH (µU/ml)) + (0.12×Dmean thyroid (Gy)).
It was noted that the negative coefficient for age (-0.03) indicated that patients with an older age had a reduced risk of RHT. However, the positive coefficients for pre-treatment TSH (0.91) and Dmean (0.12) indicated that these two variables were positively associated with a risk of RHT. Pearson correlation tests showed no significant correlations between age and Dmean, but there were high correlations between the dose-volume variables of thyroid gland as shown in Supplement 4.
T and N stage were significant predictors of RHT in univariate analysis, but not in multivariate analysis. The supraclavicular lymph node involvement was not significantly different between the non-RHT and RHT groups (6.8% vs 14.3% with a non-significant P-value = 0.107). Additionally, there was no difference of the presence of PTV 70Gy in the supraclavicular region between the non-RHT and RHT groups (21.6% vs 32.5% with a non-significant P-value = 0.099).
Model performance and validation
Our NTCP model for predicting RHT had a good performance with an AUROC of 0.749 (95% CI 0.677–0.821) (Fig. 1). The Hosmer-Lemeshow goodness-of-fit test was used to assess the performance of our NTCP model. There was a good degree of fit with a non-significant P-value = 0.837. Supplement 5 shows the calibration plot of expected and observed event rates. Internal validation was performed with a 10-fold cross-validation and resulted in a mean AUROC of 0.764 (95% confidence interval (CI) 0.634–0.795). The model performance on external validation dataset (N = 25) was excellent with an AUROC of 0.812. We compared our model to two previously reported NTCP models for predicting RHT in NPC patients, Luo[12] and Shen[16], and revealed the low AUROC in both training and external validation dataset (Table 3 and Fig. 2).
|
N |
Model |
AUROC from original dataset |
AUROC from our training dataset |
AUROC from external validation dataset |
|
|---|---|---|---|---|---|
|
Our NTCP model |
200 |
βthyroid mean dose = 0.12 βAge = -0.03 βTSH pre RT = 0.91 Constant = -5.62 |
0.749 |
0.749 |
0.812 |
|
Luo et al.[12] |
174 |
βPmax= -0.026 βthyroid V50 = 0.05 βfemale = 1.28 βCMT = 2.902 Constant = -2.695 |
0.793 |
0.606 |
0.571 |
|
Shen et al.[16] |
69 |
βthyroid mean dose = 0.093 βthyroid Volume = -0.188 Constant = -1.385 |
NR |
0.619 |
0.487 |
Abbreviations: AUROC = area under the curve of receiver operating characteristic (ROC) curves; NR=not reported; RT=radiotherapy; CMT=chemotherapy
We developed a new NTCP model as a useful tool to estimate the risk of hypothyroidism after radiation in NPC patients treated with IMRT or VMAT using a standard multivariate logistic regression analysis. The final model with three significant variables comprised of age, pre-treatment TSH, and Dmean of thyroid gland. Although there were high correlations between dose-volume variables of the thyroid gland, our model selected Dmean thyroid because it represented the best predictor as shown in the multivariate analysis, and was easily obtained from the treatment planning system. The model performance of our NTCP model was good in both training and validation dataset.
Our results confirmed that thyroid exposure dose was clearly associated with the risk of RHT, which was consistent with the results of previous studies [9–18]. The volume of thyroid that spared from high dose was a recent indicator. Chyan et al. indicated that if thyroid VS45 ≥ 3cm3, then RHT rates at 3 years post-RT was reduced (38% vs 55%) [19]. Lertbutsayanukul et al. discovered that VS60 ≥ 10cm3 was associated with lower risk of RHT (2-year incidence of 68.4% vs 40.4%) and recommended for treatment planning [11]. In this study, we also found the association of VS45 and VS60 with the risk of RHT in a univariate analysis, however, they had strong correlation with Dmean and were not included in the final model. Vogelius et al. revealed a 50% risk of RHT at a dose of > 45Gy [20], and Kim et al. suggested a V45 of 50% as a possible dose-volumetric threshold of RHT [21]. Lin et al. reported that more than 80% of the RHT patients received Dmean of over 43Gy [22]. Similarly, our study also concluded that a Dmean >45Gy was associated with an increased risk of developing RHT, thus we recommended using this constraint for the treatment planning.
Age had a negative effect on the risk of RHT in our patients which suggested that younger patients were more susceptible to develop RHT than older patients. The result agreed with previous reports [9–10] and the finding in Sommat et al. who showed that the odds of RHT decreased by a factor of 0.94 with each additional year of age [23]. This implied that radiosensitivity of the thyroid gland decreased with age.
TSH is one of the vital endocrine hormones in the hypothalamic-pituitary-thyroid axis and plays an important role to maintain thyroid homeostasis. However, the association between pre-treatment TSH level and RHT is scarce. Ronjom et al. indicated that increasing baseline TSH was a significant risk factor for developing RHT in a univariate analysis, but not in a multivariate analysis [17–18]. Lertbutsayanukul et al. demonstrated that pre-treatment TSH levels of ≥ 1.55 µU/ml was associated with smaller thyroid volume and significantly increased risk of developing RHT by 2.5 times [11]. It was postulated that the small thyroid gland might have less functional subunit to produce thyroid hormone and send negative feedback by increasing TSH level from the pituitary gland. Our finding was in line with these studies and included the pre-treatment TSH in the final NTCP model. However, it was worth noting that the absolute TSH level was different according to the laboratory units, so the model needed to be calibrated prior to application in another population.
There were two studies from China developing the NTCP models for predicting RHT in NPC patients. Luo et al. reported the incidence of RHT in 39/174 patients (22.4%) and 81.6% of them were treated with IMRT technique [12]. The predictors in NTCP model were female gender, use of chemotherapy, radiation dose to thyroid gland (V50) and pituitary gland (Pmax). Another model developed by Shen et al. from 69 IMRT-treated NPC patients consisted of two parameters, pretreatment thyroid volume and Dmean thyroid [16]. In comparison to their studies, our patient population were also in an endemic area of NPC and had similar proportion of locally advanced disease. However, there was a difference in treatment approach. For example, neoadjuvant chemotherapy was routinely administered in China but it is selected in very high-risk patients in Thailand. In addition, chemotherapy regimens in Luo et al. consisted of neoadjuvant cisplatin and docetaxel combined with concurrent and adjuvant cisplatin and 5-fluorouracil [12]. Whereas concurrent cisplatin combined with adjuvant cisplatin and 5-fluorouracil was mostly administered in our patients. Hence, in the external validation process of those two models [12, 16] using our data, the established NTCP models did not perform well, as shown by low AUROC values. The reduction in the performance of the previous two NTCP models is probably due to the differences in treatment strategies.
Higher N stage was a significant factor in univariate analysis but not in multivariate analysis. We hypothesized that patients with advanced nodal disease had larger nodes or had lower neck involvement (N3b). Thus, they received higher radiation doses to the supraclavicular region that was adjacent to the thyroid gland. However, we found that no difference of supraclavicular lymph node involvement or PTV 70Gy in the supraclavicular region between non-RHT or RHT groups. This could be due to the small number of cases to demonstrate the difference.
In our study, 7 patients had central RHT and 1 one of them was initially diagnosed with subclinical RHT and later progressed to central RHT. The median time to develop central RHT was 40 months (with a range 18–97 months). Pituitary dose was significantly higher in central RHT group. This finding was contradicted to Luo et al. who reported that a higher pituitary dose was a protective factor of biochemical RHT by preventing TSH from rising [12]. But it should be noted that the median follow-up time in their study was 24 months which was in the period of a decrease of TSH level in NPC patients who were treated with radiotherapy [24]. The incidence of central RHT might increase after a longer period of observation. Consequently, we suggested to constrain pituitary doses as low as possible while not compromising target dose in order to achieve optimal disease control.
Several strengths of our study included the uniform treatment including radiotherapy technique and regimen as well as chemotherapy use, the relatively large number of patients with longer follow-up periods, and our comprehensive analyses of dose-volume variables. We attempted to analyze gEUD which was a theoretically good surrogate for thyroid gland dose. However, we discovered that it performed similar to Dmean thyroid, possibly because the thyroid gland was a small organ with a lower non-uniform dose than a large organ. Consequently, we recommended the use of Dmean thyroid in the NTCP model because it was easier obtained than gEUD.
The retrospective design of our study resulted in missing data which could possibly lead to a selection bias. However, we attempted to reduce the bias by recruiting all consecutive patients in the study period and clearly defining inclusion and exclusion criteria. It should also be noted that the data used for our study was gathered exclusively from a single institution. Therefore, larger dataset and external validation of our model with NPC patients from other institutions is future work that would be useful.
RHT is a dose-limiting late toxicity in NPC survivors and can be predicted using our multivariable NTCP model. We demonstrated that our three-variable model, comprising of age, pre-treatment TSH, mean thyroid dose, was effective in predicting RHT in both internal and external validation cohort. Furthermore, we recommended keeping the mean thyroid dose below 45Gy in order to minimize the risk of toxicity.
3D-CRT: Three-dimensional conformal radiation therapy; 5FU: Fluorouracil; AUROC: area under the receiver operating characteristic; Dmin: Minimal dose of the thyroid (Gy); Dmean: Mean dose of the thyroid (Gy); Dmax: Maximum dose of the thyroid (Gy); FT4: Free thyroxine; gEUD: generalized equivalent uniform dose (Gy); HR: High risk; IMRT: Intensity modulated radiotherapy; IR: intermediate risk; IRB: Institutional review board; LN: Lymph node; LR: Low risk; NPC: Nasopharyngeal carcinoma; NTCP: Normal tissue complication probability; Pmin: Minimal dose of the pituitary (Gy); Pmean: Mean dose of the pituitary (Gy); Pmax: Maximum dose of the pituitary (Gy); PTV: Planning target volume; RHT: Radiation-induced hypothyroidism; RT: Radiation therapy; TSH: Thyroid stimulating hormone; VMAT: Volumetric modulated arc therapy; VSx: absolute thyroid volume spared from x Gy or more (cm3); Vx: Volume of thyroid gland receiving x Gy or more (%)
Ethical approval and consent to participate
Institutional review board approval, Faculty of Medicine, Chulalongkorn University, Bangkok (IRB no. 836/62).
Consent for publication
Not applicable
Availability of supporting data
Research data are stored in an institutional repository and will be shared upon request to the corresponding author.
Competing interests
The authors declare that they have no competing interests.
Funding
This study received no funding.
Authors’ contributions
C.C. and A.P. conceived the study and developed the theoretical framework. S.W. and N.R. collected data. N.S., A.H., S.K., P.V., and J.S. designed the computational framework and analyzed the data. S.W., A.P., C.C., N.S., A.H., and S.K. contributed to the interpretation of the results. S.W. and A.P. wrote the first draft of the manuscript. All authors provided critical feedback and helped shape the research, analysis, and manuscript.
Acknowledgements
This research article would not have been possible without the assistance and support of key advisors, colleagues, and statisticians. Thank you to the Faculty of Medicine, Chulalongkorn University and King Chulalongkorn Memorial Hospital, Thai Red Cross Society for providing resources.
- Sung H, Ferlay J, Siegel RL, Laversanne M, Soerjomataram I, Jemal A, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2021;71:209–49. https://doi:10.3322/caac.21660.
- National Cancer Institute Thailand, Cancer in ThailandVol. IX 2013–2015, 2018: 17–19
- Kam MK, Leung S-F, Zee B, Chau RM, Suen JJ, Mo F, et al. Prospective randomized study of intensity-modulated radiotherapy on salivary gland function in early-stage nasopharyngeal carcinoma patients. J Clin Oncol. 2007;25(31):4873–9.
- Peng G, Wang T, Yang K-y, Zhang S, Zhang T, Li Q, et al. A prospective, randomized study comparing outcomes and toxicities of intensity-modulated radiotherapy vs. conventional two-dimensional radiotherapy for the treatment of nasopharyngeal carcinoma. Radiother Oncol. 2012;104(3):286–93.
- Zhang M-X, Li J, Shen G-P, Zou X, Xu J-J, Jiang R, et al. Intensity-modulated radiotherapy prolongs the survival of patients with nasopharyngeal carcinoma compared with conventional two-dimensional radiotherapy: a 10-year experience with a large cohort and long follow-up. Eur J Cancer. 2015;51(17):2587–95.
- Biondi B, Cappola AR, Cooper DS. Subclinical hypothyroidism: a review. J Am Med Assoc. 2019;322(2):153–60.
- Stott DJ, Rodondi N, Kearney PM, Ford I, Westendorp RG, Mooijaart SP, et al. Thyroid hormone therapy for older adults with subclinical hypothyroidism. N Engl J Med. 2017;376(26):2534–44.
- Jonklaas J, Bianco AC, Bauer AJ, Burman KD, Cappola AR, Celi FS, et al. Guidelines for the treatment of hypothyroidism: prepared by the american thyroid association task force on thyroid hormone replacement. Thyroid. 2014;24(12):1670–751.
- Huang CL, Tan HW, Guo R, Zhang Y, Peng H, Peng L, et al. Thyroid dose-volume thresholds for the risk of radiation‐related hypothyroidism in nasopharyngeal carcinoma treated with intensity‐modulated radiotherapy—A single‐institution study. Cancer Med. 2019;00:1–7.
- Xu Y, Shao Z, Tang T, Liu G, Yao Y, Wang J, et al. A dosimetric study on radiation–induced hypothyroidism following intensity–modulated radiotherapy in patients with nasopharyngeal carcinoma. Oncol Lett. 2018;16(5):6126–32.
- Lertbutsayanukul C, Kitpanit S, Prayongrat A, Kannarunimit D, Netsawang B, Chakkabat C. Validation of previously reported predictors for radiation-induced hypothyroidism in nasopharyngeal cancer patients treated with intensity-modulated radiation therapy, a post hoc analysis from a Phase III randomized trial. J Radiat Res. 2018;59(4):446–55.
- Luo R, Wu VW, He B, Gao X, Xu Z, Wang D, et al. Development of a normal tissue complication probability (NTCP) model for radiation-induced hypothyroidism in nasopharyngeal carcinoma patients. BMC cancer. 2018;18(1):575.
- Sachdev S, Refaat T, Bacchus ID, Sathiaseelan V, Mittal BB. Thyroid V50 highly predictive of hypothyroidism in head-and-neck cancer patients treated with intensity-modulated radiotherapy (IMRT). Am J Clin Oncol. 2017;40(4):413–7.
- Boomsma MJ, Bijl HP, Christianen ME, Beetz I, Chouvalova O, Steenbakkers RJ, et al. A prospective cohort study on radiation-induced hypothyroidism: development of an NTCP model. Int J Radiat Oncol Biol Phys. 2012;84(3):e351-e6.
- Cella L, Liuzzi R, Conson M, D’Avino V, Salvatore M, Pacelli R. Development of multivariate NTCP models for radiation-induced hypothyroidism: a comparative analysis. Radiat Oncol. 2012;7(1):224.
- Shen G, Peng Y, Li J, Wu H, Zhang G, Zhao C, et al. Multivariate NTCP Model of Hypothyroidism After Intensity-Modulated Radiotherapy for Nasopharyngeal Carcinoma. Front Oncol. 2021:3336.
- Rønjom MF, Brink C, Bentzen SM, Hegedüs L, Overgaard J, Johansen J. Hypothyroidism after primary radiotherapy for head and neck squamous cell carcinoma: normal tissue complication probability modeling with latent time correction. Radiother Oncol. 2013;109(2):317–22.
- Rønjom MF, Brink C, Bentzen SM, Hegedüs L, Overgaard J, Petersen JB, et al. External validation of a normal tissue complication probability model for radiation-induced hypothyroidism in an independent cohort. Acta Oncol. 2015;54(9):1301–9.
- Chyan A, Chen J, Shugard E, Lambert L, Quivey JM, Yom SS. Dosimetric predictors of hypothyroidism in oropharyngeal cancer patients treated with intensity-modulated radiation therapy. Radiat Oncol. 2014;9(1):1–7.
- Vogelius IR, Bentzen SM, Maraldo MV, Petersen PM, Specht L. Risk factors for radiation-induced hypothyroidism: A literature‐based meta‐analysis. Cancer. 2011;117(23):5250–60.
- Kim MY, Yu T, Wu H-G. Dose-volumetric parameters for predicting hypothyroidism after radiotherapy for head and neck cancer. Japan J Clin Oncol. 2014;44(4):331–7.
- Lin Z, Yang Z, He B, Wang D, Gao X, Tam S-y, et al. Pattern of radiation-induced thyroid gland changes in nasopharyngeal carcinoma patients in 48 months after radiotherapy. PLoS One. 2018;13(7):e0200310.
- Sommat K, Ong WS, Hussain A, Soong YL, Tan T, Wee J, et al. Thyroid V40 predicts primary hypothyroidism after intensity modulated radiation therapy for nasopharyngeal carcinoma. Int J Radiat Oncol Biol Phys. 2017;98(3):574–80.
- Prayongrat A., Lertbutsayanukul C. Hypothyroidism after radiotherapy for nasopharyngeal carcinoma. Annals of Nasopharynx Cancer; Vol 4 (June 2020): Ann Nasopharynx Cancer 2020
Figure 2 is not available with this version.
No competing interests reported.