Survival After Definitive Chemoradiotherapy With Concurrent Cisplatin or Carboplatin for Head and Neck Cancer

Background

In 2019, an estimated 65,000 cases of head and neck (H&N) cancer will be diagnosed in the United States, with approximately 14,000 cancer-related deaths.¹ Many cases of locally or regionally advanced H&N squamous cell carcinoma (HNSCC) are treated with definitive radiotherapy (RT). For these patients, multiple clinical trials and a large meta-analysis have shown superiority of concurrent chemoradiotherapy (chemoRT) compared with RT alone.² Platinum chemotherapy has the strongest evidence base and longest clinical track record as the concurrent agent.

Cisplatin and carboplatin are the primary platinum agents used in the treatment of H&N cancer. Despite early interest in carboplatin in the 1990s and its use in several large clinical trials,^3–5 cisplatin has emerged as the preferred and most widely used platinum agent in the United States,⁶ and its body of evidence is more extensive. ChemoRT with either cisplatin-based (typically monotherapy) or carboplatin-based (commonly in multiagent combinations) regimens is superior to RT alone, and both confer approximately 10% to 20% absolute overall survival (OS) benefit at 3 years.^3,7 However, cisplatin and carboplatin have significantly different adverse effect profiles. Cisplatin causes more nausea and vomiting, nephrotoxicity, ototoxicity, and neurotoxicity, whereas carboplatin causes more myelosuppression but is generally better tolerated⁸; consequently, carboplatin-based concurrent therapy is frequently administered to patients with comorbidities unable to tolerate cisplatin.

For patients who are not cisplatin candidates, often because of underlying renal dysfunction or hearing loss, concurrent treatment must be performed with a different agent or abandoned altogether. Cetuximab, a monoclonal antibody against EGFR, was widely used as an alternative concurrent agent after its combination with RT was shown to be superior to RT alone.^9,10 The RTOG 1016 and De-ESCALaTE trials directly compared cetuximab with cisplatin in patients with p16-positive oropharyngeal cancer. Recent results from both trials showed significantly inferior survival for cetuximab compared with cisplatin,^11,12 and prior studies have also suggested that cetuximab may be inferior to cisplatin for nonoropharyngeal HNSCC.^13,14 Therefore, renewed interest has been shown in identifying a chemoRT regimen more tolerable than cisplatin, without compromising efficacy.

No large or modern clinical trials have compared cisplatin- versus carboplatin-based definitive chemoRT; thus, their relative effectiveness has remained unclear. Similarly, no trials have compared concurrent cetuximab with carboplatin-based therapy. Our study analyzed national patterns of care and cancer outcomes for definitive concurrent chemoRT with cisplatin or carboplatin. Both of these regimens were also compared with RT alone and with concurrent cetuximab. The SEER-Medicare database was used because of its large sample size, national representation of real-life practice patterns and outcomes, cause-of-death information, and availability of systemic therapy, RT, and comorbidity data.

Patients and Methods

Cohort Identification

The SEER-Medicare database was queried for nonmetastatic, locally or regionally advanced (stages III–IVB according to the 6th and 7th editions of the AJCC Cancer Staging Manual) SCCs of the oropharynx, larynx, and hypopharynx diagnosed from 2004 through 2013. To ensure adequate capture of claims and to analyze a more uniform population, patients were required to have Medicare Parts A and B and no HMO coverage within 12 months of diagnosis, be enrolled in Medicare for age only, and not have a prior H&N cancer diagnosis.

All patients received definitive RT according to the SEER registry. For the RT-only cohort, patients started treatment within 120 days of diagnosis and received no chemotherapy within 6 months of diagnosis. All other patients in the study received either cisplatin (codes J9060, J9062, C9418) or carboplatin (J9045). Patients with claims for both agents within 6 months of diagnosis were excluded. Codes for RT and other chemotherapy agents and RT start date and duration were defined as described previously.¹³ To identify patients treated with concurrent chemoRT and to exclude the use of induction chemotherapy, the initial RT and chemotherapy claims were required to be within 21 days of each other. Figure 1 summarizes the schema used to determine the chemoRT and RT-only cohorts. The cohort of 609 patients treated with definitive RT and concurrent cetuximab was described previously.¹³

• Download Figure
• Download figure as PowerPoint slide

Cohort determination algorithm.

Abbreviations: ChemoRT, chemoradiotherapy; RT, radiotherapy; SCC, squamous cell carcinoma.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 17, 9; 10.6004/jnccn.2019.7297

Results

A total of 1,149 patients underwent chemoRT, of whom 807 (70%) received cisplatin and 342 (30%) received carboplatin. Median follow-up was 3.7 years in living patients. In addition, we identified 590 patients treated with definitive RT alone. Table 1 lists baseline characteristics of the cisplatin and carboplatin cohorts. Compared with cisplatin recipients, those treated with carboplatin were older, had higher comorbidity scores, and were less likely to receive intensity-modulated RT (IMRT). In multivariable logistic regression (Table 1), features significantly predictive of receiving carboplatin included increasing age (adjusted odds ratio, 1.07 per additional year; 95% CI, 1.04–1.10; P<.0001) and comorbidity score.

Table 1.

Baseline Patient Characteristics and Predictors of Receiving Carboplatin

Within the carboplatin cohort, 110 patients (32%) received carboplatin monotherapy and 232 (68%) received combination chemotherapy (Figure 2A), which was predominantly with paclitaxel. By contrast, 86% of patients in the cisplatin cohort received cisplatin monotherapy (Figure 2B). H&N CSM was higher for single-agent carboplatin than for multiagent chemotherapy (34% vs 26% at 3 years) (supplemental eFigure 1, available with this article at JNCCN.org), although this difference was not statistically significant in univariable (P=.43) or multivariable analysis (adjusted HR, 1.43; 95% CI, 0.93–2.22; P=.11). Among patients who received >1 dose of carboplatin, most (84%) had weekly administration, whereas only 5% received it every 3 weeks.

• Download Figure
• Download figure as PowerPoint slide

Chemotherapy regimens used in the (A) carboplatin and (B) cisplatin cohorts.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 17, 9; 10.6004/jnccn.2019.7297

Patients treated with carboplatin-based chemoRT had similar incidence of death attributable to HNSCC compared with those treated with cisplatin-based chemoRT (3-year CSM 29% vs 26%, respectively; P=.19). Subgroup analysis of oropharyngeal and nonoropharyngeal sites also showed a similar incidence of CSM between the cisplatin and carboplatin cohorts (supplemental eFigure 2). Both chemoRT cohorts had significantly lower CSM compared with RT alone (3-year CSM, 39%) (Figure 3) and RT with concurrent cetuximab (supplemental eFigure 3).

• Download Figure
• Download figure as PowerPoint slide

Cumulative incidence of CSM in patients treated with cisplatin-based chemoradiation, carboplatin-based chemoradiation, or RT alone.

Abbreviations: CSM, cancer-specific mortality; RT, radiotherapy.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 17, 9; 10.6004/jnccn.2019.7297

In the carboplatin cohort, risk of death attributable to other causes (non-CSM) was higher than in the cisplatin cohort (16% vs 11% at 3 years; P=.05) (supplemental eFigure 4A), consistent with the older age and higher comorbidity burden of carboplatin recipients. Accordingly, in the unadjusted analysis, OS favored the cisplatin cohort (62% vs 54% at 3 years; P=.003) (supplemental eFigure 4B). To adjust for baseline differences between the cisplatin and carboplatin cohorts, propensity score analysis was performed. Patient and tumor characteristics were well-balanced in the matched cohorts (supplemental eTable 1), including age and comorbidity score. Death attributable to HNSCC remained similar in the matched cohorts (3-year CSM, 29% vs 28%; P=.54) (Figure 4A), and other-cause mortality and OS were also comparable (P=.76 and P=.35, respectively) (Figure 4A, B), suggesting a similar level of fitness after propensity score adjustment.

• Download Figure
• Download figure as PowerPoint slide

Analysis of (A) CSM and other-cause mortality (non-CSM) and (B) overall survival in the matched cisplatin and carboplatin cohorts.

Abbreviation: CSM, cancer-specific mortality.

Citation: Journal of the National Comprehensive Cancer Network J Natl Compr Canc Netw 17, 9; 10.6004/jnccn.2019.7297

In multivariable Fine-Gray regression, the adjusted HR of CSM for carboplatin relative to cisplatin was 1.01 (95% CI, 0.79–1.28; P=.94). Variables significantly predictive of CSM included tumor/nodal stage, age, and nonoropharyngeal disease sites (Table 2). Fine-Gray regression within the matched cohorts yielded similar results (supplemental eTable 2). Analysis of acute toxicities in the matched cohorts revealed no significant differences in rates of dysphagia, dermatitis, mucositis, emergency department visits, or unplanned hospital admissions (supplemental eFigure 5). However, significant differences were seen in inpatient diagnoses between cisplatin and carboplatin recipients with unplanned admissions. For example, cisplatin recipients were more likely to be hospitalized with acute renal failure (P=.04), whereas carboplatin recipients had more neutropenia (P=.03) and pneumonia (P=.01) (supplemental eTable 3).

Table 2.

Multivariable Fine-Gray Regression for Predictors of Cancer-Specific Mortality

Discussion

The optimal treatment of patients with HNSCC who cannot receive cisplatin is highly controversial. Recently, direct comparisons of cisplatin and cetuximab in RTOG 1016¹¹ and De-ESCALaTE¹² have shown cetuximab to be significantly inferior to cisplatin in patients with p16-positive oropharyngeal SCC. Prior studies have also suggested that cetuximab is inferior to cisplatin for nonoropharyngeal HNSCC.^13,14 Currently, the question of which concurrent regimen to use in patients with contraindications to cisplatin is of paramount importance. Our study suggests that patients treated with carboplatin-based regimens have cancer-specific outcomes virtually equivalent to those of patients treated with cisplatin-based regimens; hence, patients with a contraindication to cisplatin may be better served with carboplatin-based therapy. Overall rates of acute toxicities did not differ significantly, but among patients who were ill enough to warrant unplanned hospital admission, the inpatient diagnoses reflected the expected adverse event profiles of the concurrent agent that was used.

In our study, patients treated with carboplatin had several key differences in baseline characteristics compared with those treated with cisplatin, as expected from clinical experience. Overall, carboplatin recipients were older and had a higher comorbidity burden than cisplatin recipients; however, no significant difference was seen in risk of death attributable to HNSCC according to the choice of platinum agent, whereas chemoRT with either cisplatin or carboplatin was superior to RT alone and RT with concurrent cetuximab. Furthermore, the propensity score–adjusted analysis showed no difference in OS between the matched cisplatin and carboplatin cohorts. In the multivariable model, the adjusted HR for CSM was 1.01, indicating that outcomes with carboplatin and cisplatin were essentially identical. Moreover, the upper 95% confidence limit of 1.28 was well within the 1.45 noninferiority threshold for the HR of cetuximab relative to cisplatin prespecified in RTOG 1016.

As the largest, most comprehensive comparison of cisplatin- versus carboplatin-based chemoRT to date, our study adds significantly to the existing literature, given that prior studies have largely been single-institution retrospective series and limited in sample size and external validity. In one study, a matched-pair analysis of 65 patients treated with carboplatin and 65 with cisplatin found similar 3-year rates of locoregional control (87% vs 79%; P=.54), freedom from distant metastasis (88% vs 85%; P=.79), and OS (59% vs 68%; P=.24).¹⁶ Similarly, 2 series of patients who received carboplatin/5-FU or cisplatin showed comparable locoregional control, disease-free survival (DFS), and OS with either regimen.^17,18 A prior SEER-Medicare study restricted to patients with oropharyngeal carcinoma¹⁹ reported no difference between cisplatin- and carboplatin-based chemoRT, but also no difference between carboplatin and cetuximab; this may be because of the limited number of patients treated with carboplatin (n=69). Compared with that study, ours had a substantially larger sample size (finding carboplatin to be superior to cetuximab), included multiple H&N sites, and investigated the patterns of care corresponding to carboplatin monotherapy and combination regimens.

In addition, our results are consistent with findings from 2 small prospective studies that predate modern conformal RT techniques. In both, patients were randomly assigned to RT alone (70 Gy), chemoRT with cisplatin, or chemoRT with single-agent carboplatin.^20,21 Both trials had approximately 40 to 50 patients in each arm and found that the chemoRT arms had superior outcomes compared with RT alone. In the first trial, which used high-dose chemotherapy given every 3 weeks, disease control and survival appeared to favor cisplatin over carboplatin, but the differences were not statistically significant (3-year survival, 52% vs 42%; P=.28).²⁰ The other trial, which used low-dose daily chemotherapy, reported no differences between cisplatin and carboplatin for 5-year rates of local relapse-free survival (51% vs 48%; P=.75) or survival (32% vs 29%; P=.82).²¹ A randomized trial of 206 patients restricted to nasopharyngeal carcinoma also showed similar outcomes for cisplatin and carboplatin,²² although the relevance of this study to non–nasopharyngeal carcinoma HNSCC is unclear due to differences in disease biology.

Although the aforementioned trials showed superiority of concurrent carboplatin monotherapy compared with definitive RT alone,^20,21 most evidence has focused on carboplatin combination regimens, and a previous randomized trial with 72 patients found no benefit to postoperative carboplatin monotherapy compared with RT alone.²³ The carboplatin regimen most extensively studied prospectively has been carboplatin/5-FU, which was used in the seminal GORTEC 94-01 trial that showed superiority of chemoRT versus RT alone in locoregionally advanced oropharyngeal cancer.³ More recently, this combination was used in GORTEC 99-02, which compared conventional and accelerated chemoRT with accelerated RT alone,⁴ and in GORTEC 2007-01, which showed superiority of adding carboplatin/5-FU to RT with concomitant cetuximab.⁵ As of early 2019, NCCN listed both high-dose cisplatin and carboplatin/5-FU as category 1 recommendations for concurrent chemoRT, although cisplatin is preferred.⁶ In accordance with the NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for H&N Cancers, our results support the use of carboplatin-based therapy on par with cisplatin.

Of note, our analysis shows that <5% of the carboplatin cohort received 5-FU or every-3-week administration, indicating a significant departure from national guidelines. Instead, most patients received weekly carboplatin, and the most common combination regimen was carboplatin/paclitaxel, which is an NCCN category 2B recommendation (in the same category as weekly cisplatin, 40 mg/m²) and is the only concurrent carboplatin regimen besides carboplatin/5-FU listed in the NCCN Guidelines.⁶ Outcomes of this regimen have been reported in older single-institution retrospective series, which found approximately 50% DFS and OS at 3 years, and 30% DFS and OS at 5 years.^24,25 Recently, an institutional series using modern RT techniques (IMRT) with concurrent carboplatin/paclitaxel in locoregionally advanced oropharyngeal cancer reported promising 5-year locoregional recurrence-free survival, DFS, and OS rates of 89%, 65%, and 71%, respectively.²⁶ Although our study was not powered to compare different carboplatin regimens, our finding of overall equivalency between cisplatin and carboplatin (which was mostly combined with paclitaxel) is consistent with the listing of carboplatin/paclitaxel as an alternative option for concurrent therapy in the NCCN Guidelines.⁶

Other, less-studied carboplatin-based regimens described in small series include carboplatin/docetaxel^27,28 and carboplatin/cetuximab.²⁹ We also observed these regimens in a small number of patients, although these combinations are not listed in national guidelines. Overall, most patients treated with carboplatin received some form of combination chemotherapy. Only approximately one-third received carboplatin as monotherapy, and single-agent carboplatin tended to be inferior to combination chemotherapy, although this did not reach statistical significance (multivariable P=.11). Firm conclusions regarding the comparative effectiveness of single-agent versus multiagent carboplatin regimens are not possible, given the small size of the single-agent cohort (∼100 patients).

The primary limitation of our study is its retrospective nature; therefore, we used competing risks analysis, propensity score analysis, and multivariable regression to adjust for potential confounders. Because there are no large or modern randomized comparisons of cisplatin- versus carboplatin-based chemoRT, our study (although hypothesis-generating) still addresses a crucial gap in knowledge. The SEER-Medicare database also does not contain information regarding RT details (fields, dose, fractionation), recurrence/local control, HPV status for oropharyngeal cancers, and pharmacologic dosing data. Even carboplatin monotherapy can be subject to variations in dosage, depending on clinical practice. However, our study can still be interpreted as an intention-to-treat analysis as in other population-based studies,³⁰ regardless of the actual amount of chemotherapy received in either cohort. Although HPV status was not available, our results were consistent in subgroup analyses of oropharyngeal carcinoma (which may or may not be HPV-related) and nonoropharyngeal carcinoma (ie, laryngeal and hypopharyngeal cancers), which are not HPV-related. Finally, our study comprised a Medicare-aged population and may not be directly applicable to younger patients (in whom oropharyngeal SCC is more likely to be HPV-positive); however, most patients who are ineligible for cisplatin are older and would be encompassed by our study. Appropriate patient selection for non–cisplatin-based therapy remains challenging and is subject to variations in provider preferences and institutional practices; toxicity may also not be fully captured in a claims-based analysis.

Conclusions

Our study suggests the potential equivalence of cisplatin- and carboplatin-based definitive chemoRT in the treatment of locoregionally advanced HNSCC. Carboplatin-based chemoRT was superior to RT alone and RT with concurrent cetuximab. Interestingly, most carboplatin recipients received combination therapy with paclitaxel and not 5-FU, representing a significant departure from national guidelines. Considering the results of RTOG 1016 and De-ESCALaTE showing cetuximab to be inferior to cisplatin in p16-positive oropharyngeal SCC, our study suggests that carboplatin-based regimens should be investigated further in a prospective fashion as an alternative to cisplatin for patients who are not cisplatin candidates.