Risk Factors for Early Mortality in Head and Neck Cancer Patients Undergoing Definitive Chemoradiation

Hong Kong J Radiol 2022 Jun;25(2):127-35

 

 

Department of Clinical Oncology, Queen Elizabeth Hospital, Hong Kong

 

  Full Article in PDF

Radiotherapy and chemotherapy are two of the mainstays of treatment for head and neck cancer (HNC).[1] Radiotherapy of HNC may cause mucositis, dermatitis, or dysphagia, whereas addition of concomitant chemotherapy increases the risk of myelosuppression, nausea and vomiting, neurotoxicities, and nephrotoxicities. Although these typically improve within 6 months after treatment in most patients,[2] some patients succumb to these toxicities shortly after treatment.[3] [4] [5] The United Kingdom introduced a 5% 90-day mortality rate following radical radiotherapy in HNC as a quality metric.[6] [7]

 

Treatment outcomes in HNC are influenced by a range of patient and disease factors. Co-morbidity is a known independent prognosticator for survival in HNC.[8] [9] There are several co-morbidity scores available with known prognostic significance in HNC, some of which are suitable for registry-based and retrospective assessment.[8] [10]

 

The most commonly utilised co-morbidity indices are the Charlson Comorbidity Index (CCI)[11] and the Adult Comorbidity Evaluation–27 (ACE-27).[12] These have both been validated in HNC patients.[10] [13] An ACCI (age-adjusted Charlson Comorbidity Index) is another established co-morbidity score derived from the CCI and assigns 1 additional point per decade over 40 years of age up to 4 points.[14] Although validated in HNC,[15] [16] it is less widely used than CCI and ACE-27.

 

There are several assessment tools designed specifically for HNC patients. The HN-CCI (head and neck comorbidity index score)[17] and the HNCA (head and neck cancer index)[18] are both adapted from the CCI. The WUHNCI (Washington University Head and Neck Comorbidity Index) contains seven weighted conditions. It was shown to improve prognostication over a multivariable regression model comprising age, sex, race, symptom stage, and TNM stage.[19] The Simplified Comorbidity Score (SCS) includes seven co-morbid conditions and smoking status.[20] It was originally developed for lung cancer patients, but Göllnitz et al[21] demonstrated its prognostic value in HNC along with other co-morbidity scores. The Taipei Medical University–concurrent chemoradiotherapy (TMU–CCRT) Mortality Predictor Score is the only co-morbidity score developed for prediction of 90-day mortality post radical chemoradiotherapy in locally advanced HNC patients.[3]

 

We sought to assess the 90-day mortality rate after radical radiotherapy, with or without concurrent systemic treatment, to identify significant risk factors for early mortality and to assess the predictive value of different co-morbidity scores.

 

All 1649 consecutive patients treated with radiotherapy for HNC from 1 January 2016 to 1 March 2020 at a single tertiary cancer centre were screened for eligibility. The Figure shows the flowchart for screening. Patients who were planned for radiotherapy with curative intent, with or without systemic treatment, as primary treatment for non-metastatic primary HNC from all sites and stages were included. To eliminate the effect of postoperative recovery and chemotherapy before or after radiotherapy, those who had surgery, and those who received neoadjuvant or adjuvant chemotherapy with primary treatment, were excluded. We also excluded patients who received a planned radiation dose of <66 Gy to the planning target volume. Patients who were diagnosed with lymphoma, skin malignancies, sarcomas or thyroid malignancies were excluded. All 725 patients included in our analysis had received at least 6 months of follow-up post completion of primary treatment at time of analysis.

 

Figure. Flowchart of case selection

 

Radiotherapy to the head and neck region was administered via intensity-modulated radiotherapy, volumetric mediated arc therapy, or in the case of nasopharyngeal cancers (NPC), TomoTherapy. Radiotherapy doses up to 70-72 Gy in 33 fractions for NPC and up to 70 Gy in 35 fractions for squamous HNC were prescribed. Concurrent chemoradiotherapy regimens included weekly cisplatin 40 mg/m²; cisplatin 100 mg/m² given once every 3 weeks; or cetuximab 400 mg/m² 1 week before radiotherapy and 250 mg/m² in subsequent weekly doses until completion of radiotherapy.

 

In our study, early mortality was defined as 135 days from start of radiotherapy as a surrogate for 90-day mortality post radical radiotherapy, assuming a maximum planned overall treatment time of 45 days.[4]

 

Manual review and data extraction were performed for each patient. Demographic information, including age, sex, body weight and height, smoking history, and alcohol use were extracted. Eastern Cooperative Oncology Group (ECOG) performance status was documented. Tumour site, histology and staging, the results of baseline swallowing assessment, treatment details including treatment technique, dose, and concurrent systemic therapy were logged. Disruptions or unplanned events warranting medical attention during treatment, including unplanned feeding tube insertion, admission, or need for adaptive replanning of radiotherapy were recorded. Start and end dates of radiotherapy, last date of follow-up, and date and cause of death were recorded.

 

Baseline laboratory values including haemoglobin, white cell count, platelet count, albumin, creatinine, calcium, and lactate dehydrogenase were retrieved. Symptom score at presentation, scored as the number of items present from the following list: dysphagia, otalgia, weight loss and presence of neck mass, was calculated.[22]

 

Patient records were reviewed for co-morbidities comprising seven of the eight co-morbidity scores mentioned — CCI, ACCI, HN-CCI, SCS, WUHNCI, HNCA and TMU-CCRT. The calculation of ACE-27 was assessed directly from electronic medical records. The full list of co-morbidities assessed is shown in Table 1.

 

Table 1. Co-morbidity information

 

The full cohort was analysed for demographic information. Univariate analysis was performed using the Chi-square test for categorical data and the Mann-Whitney U test for continuous or ordinal data to compare the 90-day mortality and survival groups. Multivariable logistic regression was performed to evaluate the effect of different variables on 90-day mortality after chemoradiotherapy, controlling for other covariates. A p value <0.05 was taken as significant.

 

In total, among 1649 consecutive patients, 1292 were treated with radiotherapy for HNC during the study period from 1 January 2016 to 1 March 2020 (Figure). After exclusions for prior surgery (n = 281), neoadjuvant or adjuvant chemotherapy (n = 163), metastatic disease (n = 118), or death before treatment (n = 5), finally 725 patients were included in our cohort and received ≥6 months of follow-up after completion of radiotherapy at time of analysis (Figure).

 

The median age at diagnosis was 59 years. A total of 19.6% were aged >70 years, and 76.3% were male. Lifetime smokers and drinkers made up 55% and 37% of cases, respectively. The most common tumour site was nasopharynx (70.2%), followed by larynx (13.9%), oropharynx (7.6%), oral cavity (3.7%), hypopharynx (2.9%), and other sites (1.6%). Locally advanced disease (stage III/IV) made up 90.6% of cases. A total of 74.1% received chemotherapy and radiotherapy. The median follow-up was 23.1 months as of 30 November 2020 (Table 2).

 

Table 2. Patient demographics and treatment details

 

In all, 33 out of 725 patients (4.6%) did not survive beyond 90 days from the end of treatment. Of these, 14 patients (42.4%) died due to pneumonia or sepsis, nine (27.3%) developed sudden cardiac arrest. Four died due to distant relapse, and five died due to other causes including acute kidney injury, pneumoperitoneum, or deterioration after early termination of treatment. Cause of death was not available for one patient.

 

Poorer radiation tolerance was noted in the 90-day mortality group, with significantly higher rates of adaptive radiotherapy replanning (30.3% vs. 13.0%, p = 0.005), feeding tube insertion during treatment (21.2% vs. 9.2%, p = 0.024), radiotherapy suspension (36.4% vs. 5.9%, p < 0.001) and unplanned admissions (51.5% vs. 5.9%, p < 0.001).

 

A total of 57% of cases had at least one co-morbid condition (Table 1). The 90-day mortality group had higher median scores across all of the co-morbidity indices except SCS. A history of stroke (21.2% vs. 5.3%, p < 0.001), diabetes (27.3% vs. 13.2%, p = 0.022) or diabetes with complications (9.1% vs. 0.9%, p < 0.001), dementia (9.1% vs. 1.0%, p < 0.001), hemiplegia (18.2% vs. 1.0%, p < 0.001) and moderate-to-severe renal disease (18.2% vs. 5.2%, p = 0.002) were more common in the 90-day mortality group. Within 1 year of diagnosis of HNC, electrolyte disturbance (6.1% vs. 1.3%, p = 0.029), sepsis (12.1% vs. 1.4, p < 0.001), pneumonia (18.2% vs. 1.2%, p < 0.001), and urinary tract infections (6.1% vs. 0.4%, p < 0.001) were more common in the 90-day mortality group.

 

Results of simple logistic regression analysis are shown in Table 3. Multivariable logistic regression analysis (Table 4) identified age >60, ECOG performance status, and pre-treatment haemoglobin as significant independent predictors of 90-day mortality. Of the co-morbidity indices investigated, when controlled for clinical parameters, only ACE-27 and TMU-CCRT remained significant on multivariable regression analysis.

 

Table 3. Results of simple logistic regression analysis for 90-day mortality

 

Table 4. Results of multivariable logistic regression analysis for 90-day mortality

 

Our study found a 4.6% 90-day post-treatment mortality rate, which was within the 5% cut-off per the National Institute of Clinical Excellence and the Scottish Cancer Taskforce recommendations.[6] [7] Other real-world cohorts have reported early mortality rates of 5% to 18%.[3] [23] [24] Historical randomised controlled trials, which reported early mortality rates of 1% to 5%, defined the early mortality period differently, ranging from 30 days from end of treatment to 90 days from the start of treatment.[25] [26] [27] As in our cases, the addition of chemotherapy in these studies was associated with a lower early mortality rate, although this difference is likely attributable to confounding due to selection of fitter patients for more intensive treatment.[25] [26] [27]

 

Early mortality in NPC is less studied compared to locally advanced squamous cell carcinomas. Despite sharing similar radiotherapy doses, organs at risk and systemic treatment regimens, NPC, having a different aetiology and clinical course, has been purposely excluded from HNC studies to facilitate long-term prognostic analysis.[3] [4] [5] In our study, we included NPC, which made up 69.9% of our cases, due to similarities in primary treatment between NPC and locally advanced HNCs. A similar acute toxicity profile is expected. Our results confirm that the risk of early mortality, after adjusting for different clinical parameters, is not significantly different in patients with squamous cell HNC of other primary sites.[4]

 

Multivariable logistic regression analysis identified age, ECOG performance status, and haemoglobin levels as significant predictors of early mortality. The effect of age and performance status on prognosis in HNC is well-known.[28] Elderly and frail patients were traditionally excluded from HNC trials,[25] [26] [27] although exclusion based on age alone is no longer recommended on the basis of evidence showing no significant differences in objective acute toxicity measurements or late toxicity across different age cut-offs for radical radiotherapy.[29] In patients aged >70 years, however, the addition of chemotherapy yielded no survival benefit.[1] In our study, we showed that increasing age remained an important predictor for early mortality. Performance status is often used as a proxy for burden of co-morbidity, but other studies have shown that it provides prognostic information independent of co-morbidity.[30] [31] The retention of ECOG performance status with certain co-morbidity indices in multivariable logistic regression analysis confirms that both have important implications in predicting early mortality.

 

In squamous cell HNCs, pre-treatment co-morbidity has consistently been identified as an important independent prognosticator. The role of co-morbidity in NPC is less pronounced, as the Epstein–Barr virus, rather than alcohol and tobacco exposure, is the dominant aetiological factor. Despite this, we found ACE-27 and TMU-CCRT to be independent predictors of early mortality after adjusting for baseline demographic factors. ACE-27 is widely used for co-morbidity assessment across different cancers; it has been shown to predict severe acute toxicity, early mortality, as well as long-term prognosis in HNC.[4] [10] [32] [33] ACE-27 differs from other co-morbidity indices covered here in that it grades overall severity of co-morbidity from 0 (no co-morbidity) to 3 (severe co-morbidity).[12] It also covers a wider scope of co-morbid conditions compared with other co-morbidity indices under review.[10] Currently, it is the recommended standard for recording co-morbidity data in the United Kingdom.[34] TMU-CCRT scores patients for early mortality risk based on stratified age and six other co-morbid conditions.[3] Our results validate the role of TMU-CCRT as a predictor of early mortality in HNC.

 

Based on median scores in the 90-day mortality and survival groups, ACE-27 was best able to differentiate between severity levels of co-morbidity. An ACE-27 score of 2 corresponds to a moderate level of co-morbidity, while a score of 0 corresponds to no co-morbidity. In contrast, the median scores for TMU-CCRT in the 90-day mortality and survival groups both fell into the low-risk category based on the proposed stratification by Lin et al.[3] This illustrates the need for universally appropriate cut-offs for clinical use. Our review of the current literature found that, apart from ACE-27, different cut-offs have been utilised by different groups to delineate the co-morbidity burden.

 

Pre-treatment laboratory values are relatively less investigated as prognostic markers.[24] These haematological and biochemical markers may reflect baseline co-morbidity, but not all mechanisms by which they affect survival or treatment outcomes are well understood. In our study, we found that haemoglobin levels below the lower limit of normal was associated with early mortality. In HNC, anaemia is associated with higher recurrence rates and poorer overall survival,[21] [35] an effect attributed to decreased efficacy of treatment due to tumour hypoxia and decreased radiosensitivity.[36] [37] Other studies have also demonstrated that early mortality was more likely in patients with baseline anaemia.[4] After controlling for other factors, other blood parameters were not correlated with early mortality in our cases.

 

Sepsis and pneumonia were the most common causes of death among patients who died within 90 days after completing treatment. Radiotherapy to the head and neck often causes acute toxicities such as mucositis, dysphagia, and odynophagia, increasing the risk of aspiration.[38] Eisbruch et al[39] reported aspiration rates up to 65% within 3 months after radiotherapy, while other studies have found aspiration pneumonia rates up to 17.6%.[40] This may be exacerbated by concomitant use of chemotherapy or biologics. Collapse or cardiac events were the second most common, supporting previous work that suggested higher risk of cardiovascular events in cancer patients.[41]

 

Despite attempts to identify risk factors for early mortality following radical treatment for HNC, the optimal case selection criteria for intensive treatment remains elusive. Substandard treatment is independently associated with poorer survival after adjusting for other factors such as performance status, age, and mild co-morbidity,[42] emphasising the need for comprehensive assessment of fitness for treatment.

 

There are several limitations to our study. Our data were obtained from a single tertiary cancer centre and consisted of predominantly NPC patients. Analysis from a population-based database and further validation in squamous cell HNC population is necessary to confirm generalisability of our results. Second, information regarding baseline functional status or socioeconomic status was not available from retrospective review of patient records. These are potentially important factors in predicting tolerance to treatment.[43] [44] Crucially, although presence of one or more these factors indicates a higher odds of mortality, this does not definitively identify patients who will suffer early death within 90 days after treatment; as such, these factors should not be used to exclude patients from potentially curative treatment. The best course of action in an individual with multiple risk factors is not known and the formulation of any treatment plan requires careful discussion with the patient.

 

Minimising early treatment mortality is important for optimising outcomes of patients with HNC. The 90-day mortality rate after radical radiotherapy, with or without concurrent systemic treatment, in our cohort was 4.6%. Age, pre-treatment Hb, ECOG performance status, and two co-morbidity indices: ACE-27 and TMU-CCRT, were found to be independent predictors of early mortality. Development of a more comprehensive prediction model from these factors may help with case selection of patients for intensive HNC treatment.