Nasopharyngeal carcinoma (NPC) is a kind of malignancy arising from the nasopharyngeal mucosal lining. Different from other head and neck cancers, the incidence of NPC is obviously unbalanced across the world, with the highest incidence rate observed in South China (1). Because of its radiosensitivity, radiotherapy (RT) with or without chemotherapy is the standard treatment method for non-metastatic NPC (2, 3). Nowadays, distant metastases have become the main treatment failure and cause of death in NPC (4). Besides, approximately 15% of patients are detected to have distant metastases at the point of primary diagnosis (5). Once distant lesions are present, the prognosis is poor, and treatment mainly relies on systemic palliative chemotherapy (PCT) (6).

Various of platinum-based PCTs are widely applied in metastatic patients (7–9). However, it remains unknown which PCT regimen is the best, considering the trade-off between efficacy and toxicity. A randomized trial has verified that patients receiving gemcitabine plus cisplatin (GP) achieved better survival outcomes when compared with fluorouracil plus cisplatin (PF) (10). Meanwhile, in head and neck cancer, several large-scale phase III trials have shown the statistically significant survival benefits of adding docetaxel to the PF (TPF) induction regimen, and the superiority has also been observed in locally advanced NPC (11–15). Therefore, the comparison between the GP and TPF regimens is of great clinical significance. For non-metastatic NPC, Zhu et al. (16) demonstrated that the GP regimen was equivalent to TPF in treatment outcomes but with less toxicity. Up to now, no study has compared the efficacy and toxicity between TPF and GP in metastatic NPC.

In this study, we retrospectively enrolled 266 metastatic NPC patients receiving TPF or GP regimens. Based on the relatively large sample size, we compared the survival condition and acute toxicity of patients between these two PCT groups, in order to provide important information for determining the proper PCT regimen for metastatic NPC patients.

The patients' baseline characteristics between the two groups were evaluated by the χ²-test. Kaplan–Meier survival curves were used to estimate the OS curves, and the difference was compared by log-rank test. A multivariable Cox regression model was utilized to calculate the hazard ratios (HRs) and 95% confidence intervals (CIs). The propensity score matching (PSM) method was applied to balance potential confounders, using a 1:1 matching protocol with a greedy-matching algorithm, and the caliper width equaled 0.05. The following factors were included in the matching process: age, gender, smoking history, primary or secondary metastases, number of metastatic organs, chemotherapy cycles, and local treatment. To evaluate the predictive accuracy of Epstein-Barr virus (EBV) DNA, the time-dependent (3-year) receiver operating characteristic (ROC) analysis was applied. The area under the ROC curve (AUC) was calculated to assess the sensitivity and specificity of EBV DNA to predict death. The cutoff value of EBV DNA was selected based on the minimum P (highest χ²) value defined by log-rank test (17). Kaplan–Meier survival curve and Cox regression model were used to explore the prognostic value of EBV DNA. Interaction analysis based on the Cox proportional hazards model was performed between chemotherapy regimens and the state of metastasis, primary or secondary. The efficacy of two regimens was compared in the subgroups of either primary or secondary disease. All analyses were two-sided, and a two-tailed P < 0.05 indicated a difference with statistical significance. The statistical analyses were performed using SPSS for Mac version 23.0 (SPSS Inc., Chicago, IL, USA), R 3.5.1 (R Project, Vienna, Austria) and X-tile software (V.3.6.1; Yale University, New Haven, CT, USA).

To our knowledge, this is the first study to compare the efficacy and safety of TPF to GP in metastatic NPC. Compared with TPF, the administration of GP was a protective factor for OS with fewer G3–G4 AEs. According to the results, GP could serve as first-line treatment for metastatic NPC patients, in particular for those with secondary metastases. Prospective and large-sample studies are needed to validate our results.

Nowadays, with the development of RT technology, NPC patients have obtained satisfactory local regional control. Distant metastasis remains a major challenge in the management of NPC and also the leading cause of death (4, 18). Because of the huge tumor burden for metastatic patients, platinum-based systemic palliative therapy has become the main treatment method with objective response rates of 55 to 80% (7–9). However, the duration of response is short, and the long-term survival is still poor (19). Which PCT regimen is the best choice is still under discussion.

For locally advanced NPC, patients receiving TPF achieved better survival when compared with PF and (docetaxel plus cisplatin) TP, which was considered as the most effective chemotherapy regimen (14, 15). Unfortunately, the triple regimen also brought a higher incidence of treatment-related AEs (15). Because multiple cycles of chemotherapy were indispensable, the tolerability of chemotherapy regimens became a big issue faced by metastatic patients. Therefore, an effective and tolerable PCT regimen for these patients was in urgent need.

Gemcitabine is a ribonucleotide reductase inhibitor, which shows a broad-spectrum antitumor activity, including NPC (20). More importantly, gemcitabine can enhance the activity of cisplatin by increasing the formation of DNA adducts induced by platinum and inhibiting the activity of ERCC1, which is an important mechanism of platinum resistance (21). The synergistic effect of cisplatin and gemcitabine has been verified in a variety of human tumor cells in vitro (22, 23). At present, this combination is used in the treatment for a variety of malignant tumors (24, 25). Among locally advanced NPC patients, a multicenter phase III trial demonstrated that GP induction chemotherapy (IC) significantly improved the survival condition when compared with concurrent chemoradiotherapy alone (26). Compared with a PF regimen, cost-effectiveness analysis proved that GP was more cost-effective than the traditional regimen (27). Moreover, in patients with recurrent or metastatic NPC, Zhang et al. (10) verified the superiority of GP over PF in terms of efficacy and toxicity. However, no study directly compared the efficacy of GP to TPF, and the latter one has been considered a stronger PCT regimen when compared to PF in metastatic NPC patients.

In the present study, we compared these two PCT regimens in 266 metastatic patients and found that the OS was similar between the two groups (50.3 vs. 52.6%; P = 0.929). However, it should be noted that the patient's characteristics were unbalanced between the two groups, and we observed higher proportions of multiple metastatic organs (P = 0.028) and secondary metastases in the GP group (P < 0.001). After adjusting for important variables, GP was an independent protective factor with a 14.0% lower risk of death as compared to the TPF group in multivariable analysis. In the matched cohort, a higher 3-year OS in the GP group was also achieved (52.6 vs. 35.6%, P = 0.042). Comparison between the two regimens indicated that the triple regimen resulted in a higher grade 3–4 AEs in neutropenia (P = 0.012) and mucositis (P < 0.001). The incidences of grade 3–4 AEs in the GP group were similar to a previous clinical trial (10).

Additionally, our group verified that secondary metastases were associated with worse prognosis than primary metastases. Interaction analysis showed that interaction effect existed between chemotherapy regimens and the state of metastases occurrence. Therefore, we performed subgroup analysis in patients with different state of metastases. Interestingly, the superiority of GP over TPF was consistently seen only in patients with secondary metastases. This result could be partially explained by the following. On the one hand, the oral health status of NPC patients after RT is generally poor, and it can be further aggravated by receiving fluorouracil and cisplatin regimen because it may lead to severe mucositis. As the adverse reactions are intolerable, the patients are more likely to discontinue the treatment (10). In addition, deep vein catheterization for fluorouracil infusion also increases the risk of catheter-related infection and thromboembolism, which also compromised their survival rates (28). These conditions affect the efficacy of TPF regimen. However, as for GP regimen, the most common blood toxicity can be more easily identified and dealt with. On the other hand, fluorouracil or taxol-containing regimens are widely used as the IC for locally advanced NPC. For patients who developed metastases after primary treatment, the previous use of fluorouracil or taxol-containing IC might have led to chemotherapy resistance of the corresponding drugs and will weaken the efficacy of the subsequent palliative TPF regimen to some degrees (29, 30). Unfortunately, as some patients did not receive the initial treatment in our center, the previous treatment details were inaccessible; thus, we could not provide the relevant information in current study.

Our study showed the advantage of GP in improving efficacy and reducing toxicities when compared with TPF in metastatic NPC, and we recommended the use of GP as first-line treatment, especially for metastases after primary treatment. Besides, the administration of GP regimen is also simpler than TPF, and outpatient treatment is feasible, which could reduce costs for patients and the stress of hospital stay.

There were several limitations to the current study. First, this was a retrospective study, and the selective bias was unavoidable. Therefore, we performed multivariable analysis to eliminate confounding factors to some extent, and the well-balanced cohort using the PSM was considered to eliminate potential confounders between patients who received TPF and GP. Besides, it is difficult to determine the accurate disease progression time because of the retrospective design. Therefore, OS was the only endpoint in current study. Second, in view of the different types of failure pattern, cause of death, and prognosis between recurrent and metastatic NPC, patients with local recurrence were not included in this study. Third, some patients only used chest X-ray and abdominal ultrasound to evaluate their metastatic conditions, which offered limited evaluative value to treatment response. Finally, all patients involved in this study were from an endemic area, and the predominant histology is the undifferentiated non-keratinizing carcinoma. Whether the results could be applied to non-endemic areas needs to be verified. A well-designed prospective clinical trial is necessary to validate our results.

Our study suggests that the GP PCT regimen achieved better efficacy compared to the TPF PCT regimen among patients with secondary metastatic NPC. The incidence of grade 3–4 AEs was relatively lower in the GP group than in the TPF group. Prospective studies are awaited to validate our results.

The datasets generated for this study are available on request to the corresponding author.

This retrospective study was approved by the Clinical Research Committee of Sun Yat Sen University Cancer Center. The patients/participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

H-QM, L-QT, and Q-YC: study concepts. X-SS, X-HW, and S-LL: study design. X-SS, X-HW, S-LL, D-HL, RS, L-TL, and S-SG: data acquisition. X-SS, X-HW, and S-LL: quality control of data and algorithms. X-SS, X-HW, RS, D-HL, L-TL, S-SG, and S-LL: data analysis and interpretation. X-SS, X-HW, and S-LL: statistical analysis. X-SS, X-HW, S-LL, RS, D-HL, L-TL, S-SG, H-QM, L-QT, and Q-YC: article preparation. X-SS, X-HW, and S-LL: article editing. X-SS, X-HW, S-LL, RS, D-HL, L-TL, S-SG, L-QT, Q-YC, and H-QM: article review. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.