Results of a Phase II Trial of Gemcitabine Plus Doxorubicin in Patients with Recurrent Head and Neck Cancers: Serum C₁₈-Ceramide as a Novel Biomarker for Monitoring Response

This article is featured in Highlights of This Issue, p. 5839

Head and neck squamous cell carcinoma (HNSCC) is the tenth most common cancer worldwide. In 2011, 39,400 Americans developed HNSCC and 7,900 died from this disease (1). The primary risk factors for HNSCC in American men and women are tobacco and alcohol use (1). Most recently, exposure to oncogenic human papilloma virus 16 (HPV16) has been implicated in the development of HNSCC (2).

The median survival for patients with recurrent or metastatic HNSCC is about 6 months and the 1-year survival rate is around 20% (1, 2). The role of chemotherapy in the treatment of recurrent HNSCC is expanding. Single-agent chemotherapy response rates vary from 6% to 30%, and platinum-based combination therapy had response rates from 20% to 60% in phase III trials (3). However, overall survival (OS) of patients with HNSCC, even those treated with platinum-based combination regimens, has not improved over several decades. Therefore, alternative strategies to treat HNSCC and the identification of novel biomarkers to estimate response in advance of chemotherapy are desperately needed.

Gemcitabine [GEM; difluorodeoxycytidine (dFdCyd)] is an antimetabolite, whose main mechanism of action is the incorporation of dFdC triphosphate adducts into DNA, resulting in chain termination and inhibition of DNA synthesis. A phase II trial of single-agent GEM in patients with metastatic or recurrent HNSCC produced a response rate of 13% in 54 evaluable patients (4). Doxorubicin (DOX) is an anthracycline, which has the following 3 mechanisms of action: it induces the formation of covalent topoisomerase–DNA complexes, intercalates DNA, and/or causes oxidative damage. Single agent DOX has a demonstrated response rate of 13% to 23% in phase II studies (5). However, the efficacy of the GEM plus DOX combination (GEM/DOX) against HNSCC has not been previously studied.

Ceramide, a bioactive tumor suppressor sphingolipid, mediates antiproliferative effects, such as the induction of apoptosis and/or growth inhibition (6). Ceramide contains a sphingosine backbone which is amide linked to a 12- to 26-carbon containing fatty acyl chain, resulting in the generation C₁₂- to C₂₆-ceramides (7). Endogenous C₁₂- to C₂₆-ceramides and other major sphingolipids, such as sphingosine and sphingosine 1-phosphate (S1P), can be quantified in human cells and tissues using high-performance liquid chromatography/mass spectroscopy (LC/MS; refs. 8–10). Interestingly, recent data suggest that endogenous ceramides with different fatty acyl chain lengths might have biologically distinct functions (11). For example, whereas generation of C₁₈-ceramide, mainly by ceramide synthase 1, CerS1 (12), has a tumor suppressive role (11), C₁₆-ceramide generated by ceramide synthase 6 (CerS6) might induce HNSCC tumor growth (13). To this end, our previous studies have shown that C₁₈-ceramide was decreased in the majority (70%, n = 45) of human HNSCC tumors compared with their paired noncancerous head and neck tissues (14, 15). Importantly, decreased C₁₈-ceramide significantly correlated with lymphovascular invasion and nodal disease indicating the clinical relevance of C₁₈-ceramide in the overall prognosis of HNSCC patients (15). In reciprocal studies, reconstitution of C₁₈-ceramide generation via forced expression of CerS1 inhibited HNSCC xenograft-derived tumors in severe combined immunodeficient (SCID) mice compared with controls (11), supporting the tumor suppressive role of CerS1-generated C₁₈-ceramide in HNSCC.

In addition, it has been previously established that treatment of various human cancer cells with conventional chemotherapeutic agents inhibits growth via ceramide generation (16). Consistent with these data, our studies showed that the GEM/DOX combination, but not 5-fluorouracil, methotrexate, cisplatin, or paclitaxel had a synergistic effect on growth inhibition of UM-SCC-22A cells (IC₅₀ values were between 150 and 220 nmol/L) in situ (17, 18). Importantly, our preclinical studies revealed that GEM/DOX combination reconstituted levels of C₁₈-ceramide via upregulation of CerS1 expression (mRNA and protein), increasing CerS1 activity by 3.5-fold for C₁₈-ceramide generation, leading to growth inhibition in HNSCC xenografts in vivo (18). Thus, these data show that the combination of GEM/DOX inhibits HNSCC tumor growth via reconstitution of C₁₈-ceramide generation.

On the basis of these preclinical data, we hypothesized that GEM/DOX might provide a viable treatment option against recurrent or metastatic HNSCC via reconstitution of C₁₈-ceramide–dependent tumor suppressor signaling. To test this novel concept, a single-center phase II clinical trial was initiated and completed to evaluate the toxicity and overall response rate produced by the combination of GEM/DOX in patients with recurrent or metastatic HNSCC for whom prior platinum-based therapy failed, which is the first-line chemotherapy treatment for HNSCC (19). Reactivation/reconstitution of C₁₈-ceramide tumor suppressor signaling in these patients in response to GEM/DOX treatment was serially examined by measurements of sphingolipids in the serum samples of patients during therapy by LC/MS. We report here the results of this phase II trial.

This phase II, single-center, open-label study enrolled patients with recurrent or metastatic HNSCC who had received prior platinum therapy (cisplatin or carboplatin). Additional criteria for inclusion were measurable disease, more than 18 years of age, Eastern Cooperative Oncology Group (ECOG) performance score of 2 or less, normal organ and marrow function (absolute neutrophil count ≥1,500/mL, platelets ≥100,000/mL, total bilirubin ≤1.5 mg/dL, AST(sGOT)/ALT(sGPT) ≤2.5 × institutional upper limit of normal, creatinine ≤2 mg/dL, or creatinine clearance ≥30 mmL/min as calculated by the Cockroft-Gault formula), and lastly, the most recent platinum-based chemotherapy treatment for each patient had to be 3 or more weeks prior (or longer) to the day of entry into this phase II clinical trial. The exclusion criteria were receiving chemotherapy or radiotherapy within the past 3 or 4 weeks, respectively, prior treatment with any other investigational agents, or previous treatment with GEM and/or DOX. Patients with known brain metastases, history of allergic reactions attributed to compounds similar to GEM or DOX, lower than normal cardiac ejection fraction (because of potential cardio-toxic effects of DOX), uncontrolled intercurrent illness, clinical AIDS (or known positive HIV serology), and pregnant women were also excluded from the trial. The MUSC Institutional Review Board approved this study protocol and written informed consent was obtained from each patient prior to their enrollment into the trial at the Hollings Cancer Center. The Clinical Trial Registry ID number is NCT00509665.

Prior to initiating the treatment protocol, eligible patients were assessed by complete physical examination, computed tomography scan (CT scan) or other imaging modalities of the head and neck region, complete blood count, renal and liver function tests, and a MUGA or echocardiogram. Serum ceramide and sphingolipid were also measured. Eligible patients were treated with GEM 1,000 mg/m² and DOX 25 mg/m² on days 1 and 8 every 3 weeks (1 cycle = 21 days). GEM was administered as a 30-minute IV infusion. DOX was administered as a slow IV push or rapid IV drip over 5 to 10 minutes subsequent to the infusion of GEM. Patients were assessed on day 1 of every 21-day cycle for changes in performance status and the presence of toxicities. Toxicities were graded according to CTCAE Version 3.0 (Common Toxicity Criteria for Adverse Events, National Cancer Institute). On days 1 and 8 of every 21-day cycle, each patient's blood was evaluated, and lastly, every 2 cycles (6 weeks), patients had a CT scan for tumor restaging, and serum was collected for ceramide and sphingolipid measurement. Dose reduction or schedule change was allowed based on protocol. Adjustment was based upon clinical laboratory results on the scheduled day of treatment, and upon maximum toxicity encountered during the previous course. Discontinuation of treatment was required if the patient developed grade 1 or higher of cardiac left ventricular function toxicity, grade 2 or higher of pulmonary toxicity, and grade 4 hypersensitivity reactions. In the absence of treatment delays because of adverse events, patients continued on therapy at the discretion of the treating physician until disease progression, intercurrent illness, or unacceptable adverse events occurred. The flow chart of the treatment protocol is illustrated in Figure 1.

Patient response to therapy was assessed by CT scans at every 2 cycles (6 weeks). Response and progression was evaluated using the international criteria proposed by the Response Evaluation Criteria in Solid Tumors (RECIST 1.0) Committee (20). Measurable disease is defined as having at least tumor with a diameter of 20 mm or more measured with conventional techniques (CT, MRI, and X-ray) or as a tumor with a diameter of 10 mm or more as measured with a spiral CT scan. Evaluation of target lesions was carried out every 2 cycles and patients were removed from the study, if restaging scans showed disease progression (Fig. 1).

Blood samples were collected from the patients at every 2 cycles and LC/MS measurements were carried out to quantify serum ceramide (C₁₂- to C₂₆-ceramides), C₁₆-dihydroceramide, sphingosine, and S1P (Lipidomics) during the course of treatment. In brief, blood was collected from patients and centrifuged at 3,000 rpm for 5 minutes. Serum was used (0.1 mL) for LC/MS analysis (21). Sphingolipids were extracted via a single-phase extraction method using ethyl acetate:isopropanol:water, and sphingolipid concentrations were calculated as pmol/100 μL of serum (21).

Simon's optimal 2-stage design was used to evaluate the efficacy of GEM/DOX. The null and alternative response rates [complete response (CR) and partial response (PR)] were 0.20 and 0.45, respectively. Specifically, the first stage enrolled 7 patients with at least 2 responses necessary to move to the second stage. In the second stage, 10 patients were enrolled. At the end of the study, 6 or more responses in 17 patients would provide evidence to reject the null hypothesis with 81% power and an overall 1-sided alpha of 0.09. The point estimate for the response rate, P value, and 95% CI for the response rate was calculated adjusting for early stopping design (22). The primary outcome was overall response rate and the secondary outcomes were progression-free survival (PFS), OS and toxicity. The Kaplan–Meier method was used to estimate PFS, and OS. Baseline (prior to cycle 1 treatment) and changes from cycle 1 to cycle 3 in ceramide levels were evaluated by response categories [CR/PR, stable disease (SD), and progressive disease (PD)] using graphical displays and Kruskal–Wallis tests. Although many patients had data collected on ceramide markers beyond cycle 3, those who progressed did not have follow-up for ceramide measurement beyond cycle 3. As a result, comparisons between response categories (CR, PR, and SD compared with PD) were limited to changes from cycle 1 to cycle 3. Moreover, ceramide was measured for patients with CR, PR, or SD until treatment was stopped at subsequent cycles.

Between August 2005 and September 2010, 18 patients were enrolled on protocol at the Hollings Cancer Center, Medical University of South Carolina. Baseline characteristics are summarized in Table 1. The median age was 57 years with a range of 45 to 77 years. Stage IVa disease (67%) was the most common stage at diagnosis and oropharynx was the most common location of the primary disease (44.4%). Seventeen out of 18 patients were smokers, with an average of 35 pack-years (py; which was defined as smoking 20 cigarettes per day for 1 year). Moreover, 8 patients were heavy drinkers, whereas 6 patients were moderate or social drinkers, and 4 patients did not drink alcoholic beverages.

All patients enrolled in our trial had previously been treated with a platinum-based therapy prior to enrollment. Upon enrollment, 5/18 patients had locally advanced disease, 9/18 patients had distant metastases only, and 4/18 patients had locally recurrent disease and distant metastases. As primary treatment, 10 patients had definitive chemoradiation treatment, 7 patients had primary surgery followed by either adjuvant radiation or chemoradiation treatment, and 1 patient had neoadjuvant chemotherapy followed by surgery. Prior to enrollment into this phase II study, 5 patients had salvage surgery and 3 patients had reirradiation treatment with platinum agents as local therapy. Ten patients received GEM/DOX as first line, and 8 patients had GEM/DOX as second- or third-line recurrent/metastatic chemotherapy. Patients on our trial received a median of 2 cycles of GEM/DOX (range 1 to 19 cycles), with 8 patients (44.5%) who had 4 or more cycles of GEM/DOX. Post-GEM/DOX treatment, 13 patients went on to subsequent chemotherapy including docetaxel (n = 6), cetuximab (n = 3), cisplatin + cetuximab (n = 1), paclitaxel + cetuximab (n = 1), carboplatin + docetaxel + cetuximab (n = 1), and bortezomib + celebrex (n = 1).

Toxicity information was collected for every patient prior to initiating each cycle of therapy while on GEM/DOX treatment and was graded according to CTCAE Version 3.0. A summary of the toxicities can be found in Table 2. In general, the main toxicities seen were hematologic. The most common toxicity was neutropenia, which was present as a grade 3 toxicity in 4/18 patients, and as a grade 4 toxicity in 5/18 patients. Also, of significance, thrombocytopenia was observed in 1 patient as grade 3, and 2 patients as grade 4. Twelve of the 18 patients received granulocyte colony-stimulating factor (G-CSF) during the course of the treatment, and 13 patients required dose delay or reduction per protocol. Only 1 patient had febrile neutropenia. Thus, this regimen had an acceptable nonhematologic toxicity profile. For future studies, primary prophylaxis with growth factors would be recommended.

All patients received at least 1 cycle of GEM/DOX. However, only 17/18 patients were included in the analysis of objective response rate because 1 patient did not have week 6 evaluations, and was therefore not evaluable for response. The patient was replaced to complete the required 17 patients for the Simon 2-stage design evaluation. Of the remaining 17 patients, CR or PRs were observed in 4 patients. This fails to meet the criteria for rejecting the null hypothesis (P = 0.26). The adjusted estimated response rate is 0.26 with a 90% CI (0.1–0.52). In addition, 8 patients had SD (47%) for an average of 2.4 months (95% CI: 0.97–3.8), and 5 patients (29%) showed disease progression at the time of first assessment. The median OS was 5.6 months (95% CI: 3.8–18.2) for all patients on trial, and the median PFS was 1.6 months (95% CI: 1.4–4.2; Fig. 2). The OS at 1 year was 30% (Fig. 2).

To determine whether response to GEM/DOX therapy correlated with the elevation of C₁₈-ceramide generation, an indication of ceramide reconstitution, peripheral blood was collected from all patients in the trial after every 2 cycles of treatment for measurement of sphingolipids by LC/MS. C₁₂- to C₂₆-ceramides, sphingosine, S1P, and C₁₆-dihydroceramide were measured in serum samples using LC/MS, as described previously (21). A baseline blood sample was collected before the first treatment, and subsequent blood samples were collected prior to cycle 3, and thereafter prior to every 2 cycles, until the patient was taken off the trial. Serum ceramide and responses were then correlated. Baseline sphingolipid measurements and changes from baseline to cycle 3 in these patients are summarized in Figure 3. Baseline ceramides and S1P measurements did not vary by eventual response category (Fig. 3A). Importantly, sphingolipid profiles changed during the course of GEM/DOX treatment in some patients. Of significance, after 2 cycles of treatment there were significant differences in the mean percent increase in C₁₈-ceramide in patients who were classified as CR, PR, and SD when compared with PD patients (P = 0.05, Wilcoxon rank sum test; Fig. 3B). Note that of the 6 patients who progressed, only 3 had ceramide measured at cycle 3. Comparisons in other markers did not yield statistical significance (total ceramide, P = 0.19; S1P, P = 0.92; and C₁₆-ceramide, P = 0.28). There were no significant correlations between the baseline levels of ceramides (prior to cycle 1 treatment) and response to GEM/DOX. Ceramide and S1P levels from cycle 1 through 6 are shown in Figure 4 for patients who had CR/PR or SD compared with PD. In most patients with CR, PR, or SD, C₁₈-ceramide was elevated above baseline as early as after cycle 2, whereas in the 3 PD patients, C₁₈-ceramide slightly decreased in 2 patients, or did not change in 1 patient over 2 cycles of GEM/DOX treatment (Fig. 4, bottom left), suggesting that increases in C₁₈-ceramide may precede favorable clinical outcomes.

In this report of our phase II clinical trial, we provide preliminary evidence that the combination of GEM/DOX has activity in some patients with recurrent or metastatic HNSCC. In our trial, 12 of the 17 treated patients either responded to treatment or had SD as a best response. Significantly, 1 of the 17 patients had a CR, and 3 patients exhibited PR, giving CR/PR rates of 26%. Importantly, our data also suggests that elevation of serum C₁₈-ceramide, but not other ceramides, sphingosine or S1P, which was only observed in patients with CR/PR/SD compared with patients with PD, might provide a novel biomarker, which can be serially measured, to help estimate response during therapy.

Currently, platinum-based chemotherapy ± cetuximab is considered to be the standard of care for patients with unresectable recurrent or metastatic HNSCC despite the fact that most of them have been previously treated with a platinum-containing regimen (19, 23). For the majority of patients with recurrence, these responses are not durable and disease progression is inevitable; most patients will eventually succumb to the disease. Hence, development of alternate nonplatinum regimens like GEM/DOX, which has not been previously reported in the setting of refractory HNSCC, is critically necessary. Our studies showed that the GEM/DOX combination was reasonably well tolerated in most patients without major nonhematologic toxicities. However, because of the observed hematologic toxicities, we recommend primary prophylaxis with G-CSF. Both the toxicity profile and clinical benefit of this regimen could be better understood in a larger phase II/III trial. To this end, in this phase II trial, CR/PR rate of GEM/DOX treatment was 26%, which is higher than CR/PR rates of single GEM or DOX treatments with around 13% (4, 5). Importantly, when compared with various other published combination studies in phase II clinical trials for the treatment of recurrent or metastatic HNSCC (23–26), CR/PR rate obtained in response to GEM/DOX seems to be more efficacious. For example, recent trials using combinations of pemetrexed plus GEM, or bertozomib plus docetaxel reported response rates of 16% and 5%, respectively, in recurrent or metastatic HNSCC (24,25). Moreover, a recent review of cetuximab therapy reported that the response rate of 5 collective clinical trials in recurrent or metastatic HNSCC was 18.7% (23). In addition, our CR/PR rates of GEM/DOX (26%) was comparable with the response rate obtained using triweekly reduced dose docetaxel and cisplatin for the treatment of HNSCC with 24.4% (26). Thus, our data revealed that GEM/DOX treatment showed limited but important efficacy at least in some recurrent or metastatic HNSCC patients, for whom there are no successful treatment options to date. In fact, in this trial, the 1-year survival was 30%, which is encouraging given that the reported 1-year survival for this patient population in average is about 20% (19, 23–26).

Ceramides are a family of bioactive sphingolipids with tumor suppressor functions (13, 27–29). The unique and novel finding of our study is that changes in serum C₁₈-ceramide correlated with response to GEM/DOX treatment, showing that increasing serum C₁₈-ceramide may serve as an estimate of therapeutic response, which has important clinical implications. It has been well established that one of the mechanisms by which treatment with chemotherapeutic agents induce cell death is via induction of ceramide or inhibition of sphingosine kinase in various cancers (30–32). Moreover, previous studies have shown that GEM/DOX inhibits HNSCC tumor growth in SCID mice via induction of C₁₈-ceramide generation (18). The results of this phase II clinical trial reported here corroborate the finding of the previously reported preclinical studies. To our knowledge, ours is the first trial to suggest that serum C₁₈-ceramide could be a potential serum biomarker of therapeutic response in HNSCC (or any other) cancer. To this end, a limitation of this study is the fact that our protocol did not require serum sample collection at the time of progression. This made it difficult to analyze the exact change of lipid profile at progression. The next step is to analyze the timing and degree of C₁₈-ceramide changes in relation to treatment effects. Our future investigations will include a study with HNSCC patients treated with surgery, definitive chemoradiation, and palliative/salvage chemotherapy to address the above-mentioned relationships.

Currently, there are a few established biomarkers in use for HNSCC. EGFR and HPV are the two main biomarkers in use, and confirming their tissue expression is now a part of standard pathology for HNSCC at various institutions (33, 34). However, obtaining serial biopsies for biomarker assessment is not practical in the clinical setting. Serum LDH is used in the clinic to determine the metastatic spread of melanoma and non-Hodgkin's lymphoma; however, it tends to be nonspecific (35). Some potential diagnostic serum markers for HNSCC have been identified, such as interleukin 6, to monitor the development of secondary primary cancer (36), MMP-13 to diagnose lymphatic metastasis in HNSCC (37), and hypoxia-related factors to predict recurrence in HPV-negative HNSCC patients (38). While advances in serum markers may contribute to better HNSCC diagnosis, our data suggest that elevation of serum C₁₈-ceramide might be a novel biomarker for monitoring response to therapy.

It should be noted also that reconstitution of ceramide generation and/or elevation has been implicated in radiosensitization of fibrosarcoma tumor xenografts (39), and treatment with nanoliposomal ceramide enhanced effectiveness of sorafenib causing synergistic inhibition of growth in situ and in vivo (40). Thus, these studies are also in agreement with our data, which implicated GEM/DOX-mediated ceramide generation in the control of tumor progression in some HNSCC patients. More importantly, assessment of an RNA interference screen revealed mitotic and ceramide pathways as potential markers of pathologic CR in primary triple-negative breast cancers in a retrospective analysis of 5 clinical trials (41), supporting our data regarding the identification of serum C₁₈-ceramide elevation as a potential biomarker for response to GEM/DOX therapy in HNSCC patients.

In summary, our data suggest that the combination of GEM/DOX might add to the therapeutic armamentarium against HNSCC. More importantly, this study suggests for the first time that serum C₁₈-ceramide presents a novel serum biomarker for monitoring treatment response, which could be conveniently and serially measured during therapy. However, these findings need to be confirmed in studies with larger patient cohorts, in which measurement of serum ceramides would be carried out earlier and more often to determine the relationship between C₁₈-ceramide changes and timing of response or progression during treatment.

No potential conflicts of interest were disclosed.

We thank the members of the Ogretmen Laboratory for helpful discussions. In addition, we are grateful to the MUSC Lipidomics Core for the sphingolipid analysis. Lastly, we thank the Hollings Cancer Center and the patients who participated on this trial.

This work was supported by research grants from the U.S. NIH (DE016572, CA088932, and CA097132) to B. Ogretmen. S.A. Saddoughi was supported by an F30 grant from the National Institute of Environmental Health Sciences (ES019464), and Wachovia and Abney Foundation scholarships.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.