The Utility of Glasgow Prognostic Score and Palliative Prognostic Index in Patients With Head and Neck Squamous Cell Carcinoma Under Palliative Care

Abstract

Objectives:

Palliative care patients with head and neck squamous cell carcinoma (HNSCC) often experience dysphagia and airway trouble; thus, each patient requires a specific palliative care plan based on their prognostication. However, no established specific prognostic tool performed on the day of starting end-of-life care is available for such patients. We assessed the accuracy of Glasgow prognostic score (GPS) and palliative prognostic index (PPI) and their combination to establish a specified prognostic tool for patients with HNSCC in end-of-life setting.

Methods:

A retrospective clinical chart review was undertaken on patients with HNSCC in end-of life setting who were decided in Kagawa University Hospital and National Hospital Organization Shikoku Cancer Center between April 2011 and March 2019. The patients were divided into 2 categories according to GPS (0-1 and 2) and PPI (groups A-B and C). These were combined into 4 categories (PPI group A-B and GPS score 0-1: good; PPI group A-B and GPS score 2: intermediate; PPI group C and GPS score 2: poor; and PPI group C and GPS score 0-1: others). The survival curves were compared for the former 3 categories.

Results:

The median survival of the scores 0-1 and 2 on GPS were 114 (72-148) and 39 (25-52) days, respectively (P < .01). These of groups A-B and C on PPI were 79 (64-99) and 16 (9-29) days, respectively (P < .01). The median survival of the good, intermediate, and poor categories was 127 (73-149), 64 (44-80), and 15 (9-27) days, respectively (P < .01 among all categories).

Conclusions:

In this study, the survival of terminally ill patients with HNSCC can be predicted by the GPS, PPI, and their combination with sufficient probability.

Keywords

Glasgow prognostic score palliative prognostic index head and neck cancer palliative care terminal care

Introduction

Head and neck cancers are responsible for more than 550 000 cases and approximately 300 000 deaths annually worldwide.¹ The male-to-female ratio ranges from 2:1 to 4:1. Approximately 90% of all head and neck cancers are head and neck squamous cell carcinomas (HNSCCs). By incidence worldwide, HNSCC is the sixth leading cancer. Most HNSCCs are formed in the epithelial lining of the oral cavity, oropharynx, larynx, and hypopharynx.^2,3 During the end-of-life stage of HNSCC, we often experience difficulty with the patient’s care and survival prediction due to local tumor growth, followed by upper airway stenosis, hemorrhage, and reduced oral intake.

In patients with HNSCC receiving palliative care, survival must be predicted to determine the place of care. For head and neck oncologists, predicting prognosis on the day of initiation of end-of-life care is vital. However, no specific prognostic tools that meet our demands are available. Generally, the clinical prediction of survival (CPS) based on the experience of physicians is not accurate and tends to overestimate survival.⁴ Furthermore, the lack of perception regarding the death of a bereaved person may cause major depressive disorders.⁵ Therefore, a predictive prognostic tool for patients under palliative care is necessary.

Our previous study⁶ suggested that 14 days (IQR, 8-16) are necessary from the day end-of-life care is initiated to the day home care is introduced. Additionally, 30 days (IQR, 8-41) are necessary from the day end-of-life care is initiated until the day the patient can be transferred to another medical facility for palliative care (P = .04). Therefore, head and neck oncologists must be able to predict 14- and 30-day survival from the day end-of-life care is initiated.

In 2005, the European Association for Palliative Care made recommendations regarding the use of prognostic tools in advanced cancers.⁷ Since then, several prognostic tools have been developed for predicting survival under palliative care.^8-11 The Glasgow prognostic score (GPS) is a systemic inflammation-based prognostic tool that assesses only 2 objective factors (serum C-reactive protein [CRP] and albumin).¹² The palliative prognostic index (PPI), established in 1999, is a prognostic tool that assesses 5 subjective symptoms (reduced oral intake, palliative performance scale [PPS], edema, dyspnea at rest, and delirium).¹³ Sophisticated prognostic tools are difficult to use routinely in daily clinical situations. However, GPS and PPI are simpler than other prognostic tools (described in section “Discussion”). We consider GPS and PPI feasible for daily use in real-world clinical situations.

We assessed the application of the GPS and PPI to establish a specified prognostic tool for patients with HNSCC on the day of initiation of end-of-life care to accurately predict survival. We attempted to establish a more appropriate prognostic tool by combining these 2 prognostic indexes.

Methods

Study Design

A retrospective clinical chart review was performed on patients with HNSCC in end-of life settings. Patients were recruited from the Kagawa University Hospital and National Hospital Organization Shikoku Cancer Center between April 2011 and March 2019. Data on basic demographic and clinical parameters were collected on the day of initiation of end-of-life care for patients with HNSCC in the head and neck oncology division. The term “end-of-life care” in this study was defined that the care given to people who have stopped treatment to cure. It includes not only the hospice care but also the best supportive care at home. The collected data included age, sex, primary site of cancer, day end-of-life care was initiated, day of death, Eastern Cooperative Oncology Group (ECOG) performance status (PS), reduction of oral intake, the presence of edema, dyspnea at rest, delirium, serum CRP and albumin levels, use of percutaneous gastrostomy tube or central venous (CV) port. Survival time was defined from the day end-of-life care was initiated to the day of death. Surviving patients were censored 180 days after the day end-of life care was initiated. These data were obtained from April 2019 to April 2020.

We retrospectively obtained the serum CRP and albumin levels within 6 days of initiating end-of-life care from the electronic medical charts and used them to calculate the GPS. Patients with both elevated CRP levels (>10 mg/L) and hypoalbuminemia (<35 g/L), those with only 1 abnormal biomarker, and those with no abnormal biomarkers were allocated scores of 2, 1, and 0, respectively.

The PPI is the sum of 5 clinical variables. The PPI score ranged between 0 and 15. The PPI was calculated for each patient using their electronic medical chart within 6 days of initiating end-of-life care. We used the ECOG PS instead of the PPS to simplify PPI calculations. The ECOG PS scale scores of 0 to 2, 3, and 4 relate to PPS scores of 100 to 60, 30 to 50, and 10 to 20, respectively.^14,15 Patients receiving total parental nutrition or nutrition fully administrated by an enteral feeding tube or gastrostomy tube were included in the “normal” oral intake category. Delirium was diagnosed using the Diagnostic and Statistical Manual of Mental Disorders (fifth edition) criteria. A doctor assessed the status of patients who had difficulty communicating verbally using a proxy or caregiver response. According to study by Morita et al,¹³ the patients were categorized by PPI as follows: ≤2: group A; >2 and ≤4: group B; and >4: group C. The cases lacking these data were excluded.

To investigate the usefulness of the combination of GPS and PPI, these were divided into 2 groups each (GPS scores 0-1 and 2, PPI groups A-B and C) and combined into 4 categories (PPI group A-B and GPS score 0-1: good; PPI group A-B and GPS score 2: intermediate; PPI group C and GPS score 2: poor; and PPI group C and GPS score 0-1: others).

Statistical Analysis

Overall survival was calculated using the Kaplan-Meier method. A log-rank survival analysis and post hoc test (Holm method) were used to determine significant differences between the survival curves. The GPS and PPI score groups (described in Table 2) were analyzed with Fisher exact test. All statistical assessments were considered significant when P < .05. All statistical analyses were performed using EZR, a graphical user interface for R. More specifically, it is a modified version of R commander designed to add statistical functions frequently used in biostatistics.¹⁶

Research Ethics and Patient Consent

Informed consent was obtained by opt-out, and explanatory documents were posted on the information boards of each institution. Data were entered into a password-protected database and analyzed by descriptive statistics. The study protocol was approved by the ethics and research governance committees of each institution. The Kagawa University Ethical Committee and the National Hospital Organization Shikoku Cancer Center ethical committee approval numbers were 2019-249 (April 6, 2020) and RIN2019-47 (October 4, 2019), respectively. Data anonymity was guaranteed by the use of fully de-identified database records. This study was conducted in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki.

Results

Of 168 eligible patients, 110 provided all the necessary data and were enrolled in this study, including 95 males and 15 females. The median patient age was 67 years (range: 26-93 years). In 58 exclude cases, 6 and 52 cases could not obtain GPS parameters, both GPS and PPI parameters, respectively.

The patient characteristics are presented in Table 1. The most common site was the hypopharynx (40 cases), followed by the oropharynx (24 cases), oral cavity (21 cases), and larynx (13 cases). Thirty (27.3%) and 34 (30.9%) patients had a gastrostomy tube and CV port, respectively.

Table 1.

Patient Characteristics (n = 110).

Variables	Number of patients	%
Gender
Male	95	86.4
Female	15	13.6
Age (years)
<50	6	5.4
50-75	82	74.5
>75	22	20.0
Site of primary tumor
Nasopharynx	4	3.6
Oropharynx	24	21.8
Hypopharynx	40	36.4
Larynx	13	11.8
Oral cavity	21	19.1
Paranasal	4	3.6
Others	4	3.6
Components for scoring PPI
ECOG PS 3 (score 2.5)	26	23.6
ECOG PS 4 (score 4)	7	6.3
Dyspnea at rest (score 3.5)	31	28.2
Delirium present (score 4)	12	10.9
Edema present (score 1)	22	20.0
Reduced oral intake >1 mouthful (score 1)	51	46.4
Reduced oral intake <1 mouthful (score 2.5)	17	15.5
Total PPI score, median (IQR)	1.5 (0-5)
Components for scoring GPS
CRP > 5 g/L (mGPS score 1)	100	90.9
CRP > 10 g/L (GPS score 1)	98	89.1
Albumin <35 g/L (score 1)	84	76.4
Gastrostomy tube
With	30	27.3
Without	80	72.7
Central venous catheter
With	34	30.9
Without	76	69.1

Abbreviations: CRP, C-reactive protein; ECOG PS, Eastern Cooperative Oncology Group performance status; GPS, Glasgow prognostic score; mGPS, modified Glasgow prognostic score; PPI, palliative prognostic index.

The GPS values of 0, 1, and 2 were reported in 8, 22, and 80 cases, respectively. Figure 1 displays the survival curves obtained from the GPS. The median survival and 95% CI for scores of 0, 1, and 2 were 148.5 (73 to none applied), 102 (49-136), and 39 (25-52) days, respectively. There was no significant difference between scores of 0 and 1 (P = .12), but there was between scores of 0 and 2 and between 1 and 2 (P < .01, respectively). Thus, GPS values of 0 and 1 were merged. The median survival and 95% CI of the score 0-1 group was 114 (72-148) days. There were significant differences between 0-1 and 2 (P < .01).

Figure 1.

Survival curves for Glasgow prognostic score (GPS) There were significant differences between GPS scores 0-1 and 2 (P < .01).

The PPI groups A, B, and C included 58, 14, and 38 cases, respectively. Figure 2 displays the survival curves obtained from the PPI. The median survival and 95% CIs of groups A, B, and C were 81.5 (64-112), 72 (22-105), and 16 (9-29) days, respectively. Groups A and B displayed similar survival curves with no statistical difference (P = .14). Therefore, we merged groups A and B. The median survival and 95% CIs of group A-B was 79 (64-99) days. There were statistical differences between groups A and C and groups B and C (P < .01, respectively) and between groups A-B and C (P < .01).

Figure 2.

Survival curves for palliative prognostic index. There were statistical differences between groups A and C and groups B and C (P < .01, respectively) and between groups A-B and C (P < .01).

The correlation between GPS and PPI was examined, and the results are displayed in Table 2. We divided the patients into 4 categories (PPI group A-B and GPS score 0-1: good [27 cases]; PPI group A-B and GPS score 2: intermediate [45 cases]; PPI group C and GPS score 2: poor [35 cases]; and PPI group C and GPS score 0-1: others [3 cases]). The survival curves are presented in Figure 3. There was a significant difference in the survival curves between the former 3 groups (P = .03 between good and intermediate, P < .01 between intermediate and poor and between good and poor). The median survival and 95% CIs of the good, intermediate, and poor categories were 127 (73-149), 64 (44-80), and 15 (9-27) days, respectively (Ps < .01). The median survival and 14- and 30-day survival probabilities of GPS scores 0-1 and 2, PPI groups A-B and C, and the former 3 categories in their combination are provided in Table 3.

Table 2.

Proportions Classified by GPS and PPI.^a

	PPI group A-B	PPI group C
GPS score 0-1	27	3
GPS score 2	45	35

Abbreviations: GPS, Glasgow prognostic score; PPI, palliative prognostic index.

^a P = .0007.

Figure 3.

Survival curves for the combination of Glasgow prognostic score and palliative prognostic index. There was a significant difference in the survival curves between all 3 groups (P = .03 between good and intermediate, P < .01 between intermediate and poor and between good and poor).

Table 3.

Overall Survival Probability of GPS and PPI and Their Combination.

	Number of. patients	Median survival (days) (95% CI)	14-Day survival probability (%) (95% CI)	30-Day survival probability (%) (95% CI)
GPS score 0-1	30	114 (72-148)	96.7 (78.6-99.5)	86.7 (68.3-95.8)
GPS score 2	80	39 (25-52)	76.2 (65.3-84.1)	53.7 (42.3-63.9)
PPI group A-B	72	79 (64-99)	97.2 (89.4-99.3)	80.6 (69.4-88.0
PPI group C	38	16 (9-29)	55.3 (38.3-69.3)	29.0 (15.7-43.6)
Good	27	127 (73-149)	100	92.6 (73.5-98.1)
Intermediate	45	64 (44-80)	95.6 (83.4-98.9)	75.6 (60.2-85.6)
Poor	35	15 (9-27)	51.4 (34.0-66.4)	31.4 (17.1-46.8)

Abbreviations: GPS, Glasgow prognostic score; PPI, palliative prognostic index.

The “others” category included only 3 cases and was excluded from the statistical analysis. One case had anterior thoracic skin metastasis and a large fistula reaching pleura. His ECOG PS was 3, and reduced oral intake was documented. He died 15 days after the introduction of end-of-life care. The other 2 cases had cervical lymph node metastasis and head and neck carcinoma–specific local problems (upper airway obstruction, reduced oral intake, hemorrhage, and dysphagia). They died 46 and 49 days after the introduction of end-of-life care, respectively.

Discussion

Primary Findings and Implications of the Results

To our knowledge, our report is the first to confirm the efficacy of GPS and PPI as prognostic tools of life expectancy when performed on the day of initiation of end-of-life care in patients with HNSCC receiving palliative care. Furthermore, we demonstrated a more detailed prognostic tool from the combination of GPS and PPI to predict survival. The most critical finding of this study was that different statuses of GPS, PPI, and their combination could reveal significantly different survival prognoses.

The combination of GPS and PPI can further subdivide the cases and predict the prognosis of each category. Table 2 summarizes that almost all (35/38, 92.1%) patients in PPI group C had a GPS score of 2. Conversely, less than half of the GPS score 2 cases (35/80, 43.8%) were categorized in PPI group C. These data suggest that the GPS worsens before PPI in many cases. Systemic inflammation in patients with advanced-stage HNSCC is always high,¹⁷ which may be why inflammation-based scores become worse early. On the other hand, poorer PPI parameters (reduced oral intake, dyspnea at rest, worsening of PS/delirium/edema) were often observed in terminal cases of HNSCC. Thus, the GPS worsens before the PPI.

By combining GPS and PPI, we were able to identify cases with GPS score 2 and PPI group 1 as the “intermediate” group. In our result, their combination could evaluate the prognosis in detail. For our 3 cases classified as “others” (GPS score of 1 and PPI group C), the PPI became worse earlier than the GPS. These findings suggest that a combination of GPS and PPI is sometimes inappropriate when patients have major local problems caused by primary or neck tumors or a worse PS caused by distant metastasis.

In our district, estimating the survival of patients with HNSCC under palliative care is crucial for determining the place of care. Based on this study, patients classified as good have enough time to consider where to receive care because their 30-day survival was estimated at 92.6%. Patients classified as intermediate have enough time to introduce home care because their 14-day survival was estimated at 95.6%. However, approximately one-fourth of intermediate patients could not transfer to other medical facilities for palliative care because their 30-day survival was estimated at 75.6%. Half of the patients classified as poor will fail to introduce home care because their 14-day survival was estimated at 51.4%. Therefore, the combination of GPS and PPI may be useful for determining the place of care for patients with HNSCC receiving palliative care.

The breakdown of the 14 cases in group B was 4 cases of decreased PS (ECOG PS 4), 3 cases of decreased oral intake less than 1 mouthful, 5 cases of decreased PS (ECOG PS 4) and decreased oral intake more than 1 mouthful, and 2 cases of respiratory distress. Nine cases were associated with a decrease in PS. Although these patients were able to take oral intake reasonably well, long-term survival was expected. This may be the reason why there was no statistical difference in the survival between PPI groups A and B.

Of the 22 patients with a GPS score of 1, 19 had a CRP of 0.5 or higher. Inflammatory response increases early in head and neck cancer. Nevertheless, the general condition is often maintained. This may be the reason for the lack of significant difference between GPS scores 0 and 1.

Comparison With Literature

Systemic inflammation-based scores (GPS and modified Glasgow prognostic score [mGPS]¹⁸) have been assessed as prognostic tools in terminally ill patients with cancer in several studies.^19,17 Laird et al¹⁹ reported that systemic inflammation-based scores for survival at 3 months varied between 82% (mGPS 0) and 39% (mGPS 2) and had a prognostic value independent of ECOG PS. The GPS requires only 2 objective parameters that can be obtained from blood samples. A blood sample test is necessary when determining whether patients with advanced-stage HNSCC can continue antineoplastic treatment. Therefore, GPS is an easy and useful prognostic tool for head and neck oncologists when patients decide not to continue antineoplastic treatment.

Palliative prognostic index has been a useful prognostic tool for several patients with cancer under palliative care.^14,20 Chou et al¹⁴ reported that the median survival of the good, intermediate, and poor prognostic groups based on the PPI were 49, 14, and 7 days, respectively, and that the PPI is a useful prognostic tool for assessing life expectancy in terminally ill patients under palliative care for hematologic malignancy. Our study suggests that PPI is also associated with patients with HNSCC under palliative care.

There are several representative prognostic tools for patients under palliative care,⁸ including the PPI, palliative prognostic score (PaP), delirium-PaP (D-PaP), prognosis in palliative care study (PiPS), PPS, and GPS. The other prognostic tools did not appear feasible for retrospective analyses and daily use in real-world settings for head and neck oncologists. The PaP, D-PaP, and PPS require CPS data, but some physicians experience difficulty in formulating prognoses. Therefore, CPS was not often described in medical charts. The PiPS consists of PiPS-A and PiPS-B. The PiPS-A requires 13 clinical markers and PiPS-B requires 9 clinical markers and 9 biomarkers to calculate. We could not obtain these data from chart reviews in many cases. The GPS only requires 2 objective parameters, and PPI only requires 5 subjective parameters. Blood samples are easy to obtain, and the PPI parameters are often considered when the physician and cancer treatment team determine whether the patients should receive antineoplastic treatment. Therefore, GPS and PPI can be analyzed retrospectively. This combination is a better prognostic tool because both subjective and objective factors can be used to estimate the patient’s prognosis. Moreover, prognostic tools must be calculated quickly. Thus, the GPS, PPI, and the combined index are feasible for head and neck oncologists.

Limitations

This study has several limitations. First, only Japanese patients were included. Thus, additional study is necessary to apply our results to other ethnicities. Second, the PPI score may be biased because the evaluation of each PPI parameter depends on the clinicians in charge. Third, it was difficult to estimate survival in some cases because patients with HNSCC under palliative care sometimes have specific problems (eg, upper airway stenosis, hemorrhage due to primary site or neck recurrence).

What This Study Adds

In this study, the survival of terminally ill patients with HNSCC can be predicted by the GPS, PPI, and combined index of the PPI and GPS with sufficient probability. The method of scoring used for GPS and PPI is easy and reproducible. Moreover, GPS and PPI do not require any additional invasive examinations, except for a blood sample test. These tools might enable head and neck oncologists, palliative care physicians, health workers, and families to plan an adequate place of care for each patient and provide more appropriate end-of-life care for patients with HNSCC.

Footnotes

Acknowledgments

The authors thank Enago () for the English language review.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Takehito Kishino

References

Jemal A, Bray

Center

, et al. Global cancer statistics. CA Cancer J Clin. 2011;61(2):69–90.

Jemal

Siegel

Ward

, et al. Cancer statistics, 2007. CA Cancer J Clin. 2007;57(1):43–66.

Boyle

Levin

. World Cancer Report 2008. International Agency for Research on Cancer: IARC Press; 2008.

Glare

Virik

Jones

, et al. A systematic review of physicians’ survival predictions in terminally ill cancer patients. BMJ. 2003;327(7408):195–198.

Barry

Kasl

Prigerson

: Psychiatric disorders among bereaved persons: the role of perceived circumstances of death and preparedness for death. Am J Geriatr Psychiatry. 2002;10(4):447–457.

Kishino

Mori

Samukawa

, et al. Current status of end-of-life care for head and neck patients: single institution experience in Kagawa, Japan. Nippon Jibiinkoka Gakkai Kaiho (Jpn). 2019;122:1127–1133.

Maltoni

Caraceni

Brunelli

, et al. Prognostic factors in advanced cancer patients: evidence based clinical recommendations—a study by the Steering Committee of the European Association for Palliative Care. J Clin Oncol. 2005;23(25):6240–6248.

Simmons

CPL

McMillan

McWilliams

, et al. Tools in patients with advanced cancer: a systematic review. J Pain Symptom Manage. 2017;53(5):962–970.

Hui

Prognostication of survival in patients with advanced cancer: predicting the unpredictable?

Cancer Control. 2015;22(4):489–497.

10.

Hui

Paiva

Del Fabbro

. Prognostication in advanced cancer: update and directions for future research. Support Care Cancer. 2019;27(6):1973–1984.

11.

Zhou

Cao

, et al. Validation of the palliative prognostic index, performance status–based palliative prognostic index and Chinese prognostic scale in a home palliative care setting for patients with advanced cancer in China. BMC Palliat Care. 2020;19(1):167.

12.

McMillan

Crozier

JEM

Canna

Angerson

McArdle

. Evaluation of an inflammation-based prognostic score (GPS) in patients undergoing resection for colon and rectal cancer. Int J Colorectal Dis 2007;22(8):881–886.

13.

Morita

Tsunoda

Inoue

Chihara

. The Palliative prognostic index: a scoring system for survival prediction of terminally ill cancer patients. Support Care Cancer 1999;7(3):128–133.

14.

Chou

Kao

Wang

, et al. The application of the palliative prognostic index, Charlson Comorbidity Index, and Glasgow prognostic score in predicting the life expectancy of patients with hematologic malignancies under palliative care. BMC Palliative Care. 2015;1:14–18.

15.

Bandukwala

Burman

, et al. Interconversion of three measures of performance status: an empirical analysis. Eur J Cancer. 2010;46(18):3175–3183.

16.

Kanda

. Investigation of the freely available easy-to-use software ‘EZR’ for medical statistics. Bone Marrow Transplant. 2013;48(3):452–458.

17.

Nakayama

Tabuchi

Hara

. Clinical utility of the modified Glasgow prognostic score in patients with advanced head and neck cancer. Head Neck. 2015;37(12):1745–1749.

18.

McMillan

. An inflammation-based prognostic score and its role in the nutrition-based management of patients with cancer. Proc Nutr Soc. 2008;67(3):257–262.

19.

Laierd

Kaasa

McMillan

, et al. Prognostic factors in patients with advanced cancer: a comparison of clinicopathological factors and the development of an inflammation-based prognostic system. Clin Cancer Res. 2013;19(19):5456–5464.

20.

Maltoni

Scarpi

Pittureri

, et al. Prospective comparison of prognostic scores in palliative care cancer populations. Oncologist. 2012;17(3):446–454.