Patient population and treatment
From the Cancer Registry Database of Kaohsiung Veterans General Hospital, 179 HNC patients who underwent taxane and platinum-based IC between May 1, 2014, to May 31, 2021, were identified (Fig. 1). IC was administered if a patient was under suitable conditions and had an Eastern Cooperative Oncology Group (ECOG) performance status score of 0–2. Our hospital has two main chemotherapy regimens, abbreviated as TP, and TPF. All regimens were administered every three weeks for two to three cycles. Patients who were scheduled for two to three cycles failed to undergo subsequent chemotherapy due to adverse events were also included for analysis. The TP regimen was the most used and consisted of docetaxel (60 mg/m2 on Day 1) and cisplatin (60 mg/m2 on Day 1). The TPF regimen was TP plus 5-fluorouracil (500 mg/m2 as a 4-day continuous infusion). For patients with poor renal function, cisplatin was replaced by carboplatin. After IC, no prophylactic antibiotics were prescribed unless severe toxicity occurred. The patient was re-evaluated using imaging studies to identify their overall response and determine the treatment course after two to three IC cycles.
The independent variables were categorized into three domains. One domain included the patients’ characteristics (age, sex, comorbidities, etc.) and tumor features (sub-sites, clinical TNM stages recommended by the American Joint Committee on Cancer [7th edition before 2018, and 8th edition thereafter] [16, 17], etc.). The second domain also included clinical laboratory data (complete blood cells and differential count, sodium, potassium, glutamic oxaloacetic transaminase (GOT), glutamic pyruvic transaminase (GPT), total bilirubin, albumin, glucose level, blood urea nitrogen, creatinine, etc.). Another domain included body composition measurements, including lean body mass (LBM) and skeletal muscle index (SMI). A third domain had chemotherapy dose and dose per LBM. SMI was derived from a cross-sectional area (CSA) at the third cervical vertebra of the pre-treatment computerized tomography (CT), magnetic resonance imaging (MRI), or positron emission tomography (PET) scan images, as previously described [18, 19]. Informed consent was not required due to the use of non-identifiable records. The Ethics Committee of our Institutional Review Board approved the study protocol (KSVGH22-CT1-03).
Body composition measurements and chemotherapy dose per kilogram lean body mass
Body composition was estimated using the method developed by Swartz et al. [20]. The skeletal muscles CSA was measured on the reference slice as the first axial view where all vertebral arcs could be identified, along with the transverse and spinous processes. The measurement was taken by scrolling from the caudal to the cephalad direction on the head-and-neck images. The skeletal muscles CSA was manually determined twice by a single physician (Supplementary Fig. 1). The skeletal muscles CSA at the C3 vertebra (in square centimeters [cm2]) was determined as the sum of separate measurements of the paravertebral and bilateral sternocleidomastoid muscles. The intra-rater reliability correlation coefficient for CSA at C3 was acceptable (0.94; 95% confidence interval [CI], 0.923–0.953) (Supplementary Table 1). The following formulae were also used to calculate CSA, SMI, and LBM:
$$\mathrm{CSA}\;\mathrm{at}\;\mathrm L3\;(\mathrm{cm}^2)=27.304+1.363\times\mathrm{CSA}\;\mathrm{at}\;\mathrm C3\;(\mathrm{cm}^2)-0.671\times\mathrm{age}\;(\mathrm{years})+0.64\times\mathrm{weight}\;(\mathrm{kg})+26.442\times\mathrm{sex}\;(\mathrm{wherein}\;\mathrm{sex}=\;1\;\mathrm{for}\;\mathrm{females}\;\mathrm{and}\;2\;\mathrm{for}\;\mathrm{males})$$
$$\mathrm{SMI}\;\mathrm{at}\;\mathrm L3\;(\mathrm{cm}^2/\mathrm m^2)\;=\;\mathrm{CSA}\;\mathrm{at}\;\mathrm L3\;(\mathrm{cm}^2)/\mathrm{squared}\;\mathrm{height}\;(\mathrm m^2)$$
$$\mathrm{LBM}\;(\mathrm{kg})=0.3\times\mathrm{CSA}\;\mathrm{at}\;\mathrm L3\;(\mathrm{cm}^2)+6.06$$
Chemotherapy dose per kilogram lean body mass was determined as:
$$\mathrm{Docetaxel}\;\mathrm{dose}\;\mathrm{per}\;\mathrm{kilogram}\;\mathrm{lean}\;\mathrm{body}\;\mathrm{mass}\:=\:\mathrm{Docetaxel}\;\mathrm{dose}\;(\mathrm{mg})/\mathrm{LBM}(\mathrm{kg})$$
$$\mathrm{Cisplatin}\;\mathrm{dose}\;\mathrm{per}\;\mathrm{kilogram}\;\mathrm{lean}\;\mathrm{body}\;\mathrm{mass}\:=\:\mathrm{Cisplatin}\;\mathrm{dose}(\mathrm{mg})/\mathrm{LBM}(\mathrm{kg})$$
Dose-limiting toxicity events and patient follow-up
The dose-limiting toxicity (DLT) events included (1) grade 3 or higher acute hematological adverse events like neutropenia (absolute neutrophil count [ANC] < 1000 with or without fever), anemia (hemoglobin < 8 mg/dL, or with life-threatening consequences), and thrombocytopenia (platelet < 25,000/μL), and non-hematological adverse events, such as acute liver injury (total bilirubin > 1.5×upper limit of normal (ULN); GOT or GPT > 5 × ULN) and acute kidney injury (creatinine > 3× ULN) according to the Common Terminology Criteria for Adverse Events version 5.0 or death after IC (https://ctep.cancer.gov/protocoldevelopment/electronic_applications/docs/CTCAE_v5_Quick_Reference_5x7.pdf), (2) death after chemotherapy, (3) subsequent dose reduction of > = 50% for taxane or platinum, or (4) postponement of subsequent therapy of ≥ 4 days due to bone marrow suppression [14]. DLT events after the first cycle of induction chemotherapy was defined as the study endpoint.
Responses after 2–3 cycles of induction chemotherapy were analyzed according to Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 criteria by CT/MRI images. Subsequent treatment modality was also reported.
Statistical analyses
SAS version 9.4 (SAS, Inc., Cary, NC) was used to analyze the data. Categorical variables, such as sex, primary tumor site, clinical stage, and comorbidities, were analyzed using Pearson’s chi-square or Fisher’s exact test. Continuous variables, such as age, LBM, SMI, and chemotherapy dose per LBM, were compared using one-way ANOVA. Univariate logistic regression was performed at first, and variable with a P-value less than 0.1 were candidates for multivariate analysis. Multivariate logistic regression with backward stepwise method was applied to determine the significant variables. Rounded number of beta coefficient or summation of risk factors in the final model was used to establish a prediction model. Another cohort composed of 34 HNC treated with taxane and platinum-based induction chemotherapy between June 1, 2021, and May 31, 2022, were recruited in order to validate the above-mentioned risk score. A two-sided P-value < 0.05 was considered statistically significant.