Study population

An NPC clinical registry was developed in 2012 at Dr Sardjito General Hospital Yogyakarta Indonesia. The registry includes data on all patients diagnosed and treated at the hospital from January 2007 to December 2016. Data were collected and collated from 2012 till 2018 and extracted from medical records and paired with data from the pathological department. Written informed consent was obtained from participants prior to medical treatment. This retrospective study that utilised data from the registry was approved by the Joint Ethics Committee of Faculty of Medicine, Public Health, and Nursing at Universitas Gadjah Mada and Dr Sardjito General Hospital (reference number: KE/FK/0250/EC/2018).

Details concerning patient characteristics, presentation of the disease, treatment and follow-up were obtained from medical records. The histological diagnosis was made according to the WHO classification of NPC [38]. Pre-treatment evaluations included gathering of sociodemographic characteristics (sex, age, educational status, marital status and insurance type), year of diagnosis, BMI, performance status (by the Eastern Cooperative Oncology Group/ECOG classification), a chest X-ray, computed tomography (CT) of the head-and-neck region, an abdominal ultrasound and a bone X-ray. CT scans of the abdominopelvic region or chest were conducted when clinically indicated. Patients underwent clinical staging according to the American Joint Committee on Cancer staging system. Staging was aligned with the 6^th edition for diagnoses made prior to 2010 [39], and those from 2010 aligned with the 7^th edition [8]. Staging was harmonised across levels of the 7^th edition for the analysis.

Treatment and follow-up

Treatment data was obtained from the clinical registry and ambulatory clinic charts. Generally, for adjuvant treatment intention, a conventional fractionated schedule of radiation (daily fraction of 2 Gy with a total 33–35 fractions) was given. In 2007–2016 2D and 3D conventional radiotherapy techniques were used. Early in 2017, application of IMRT was applied. Different strategies of chemotherapy for locally advanced disease were given due to the waiting time to radiation which included sequential chemotherapy and radiotherapy, CCRT alone, or IC-CCRT. For cases who had been planned to have sequential treatment or IC-CCRT, the TPF regimen (comprising of intravenous docetaxel 75 mg/m2 on day 1, cisplatin 100 mg/m2 on day 1 and 5-fluoro-uracil 1000 mg/m2/day on day 1–4, every 3 weeks) was applied. Alternatively, PF regimen (comprising of intravenous cisplatin 100 mg/m2/day on day 1 and 5-fluoro-uracil 1000 mg/m2/day on day 1–5, every 3 weeks) was applied for a 3–4 courses. In cases that needed to wait for long periods to begin a radiation schedule, chemotherapy courses with PF regimen were extended depending on the clinical conditions. In the CCRT scheme, concurrent chemotherapy consisted of intravenous cisplatin 40 mg/m2 weekly during radiotherapy. For patients with distant metastasis, palliative chemotherapy was the commonest treatment strategy applied using a PF regimen for 3 courses or other regimens depended on the clinical condition. Palliative bisphosphonate and radiotherapy was also applied on bone metastasis as needed.

Following treatment completion (either last cycle of chemotherapy or radiation), patients were followed up routinely with treatment response assessments planned between 8 and 12 weeks after treatment. This generally combined a physical assessment, nasoendoscopy, a nasopharyngeal CT-scan and other imaging for other part of the body. Following an initial follow-up appointment, subsequent follow-up appointments were scheduled between 3 and 6 months during the first 2 years after all treatment completion. Follow-up data for clinical outcomes and patient survival were obtained from the medical records. Survival time is from cancer diagnosis until death or last follow-up. In addition, telephone communication was used for patients whose life status could not be found in the medical records. Family members of cases was considered as a respondent if the patient was deceased. Lost to follow up was determined when individual discontinued any cancer management or follow-up for any reason and life status was thus determined in the last hospital visit.

Statistical analysis

We carried out statistical analysis using multivariable semiparametric Cox regression and multivariable parametric survival analysis. We estimated models both for a sample of the complete cases (i.e., where all variables are non-missing) and further imputed for the remaining cases with missing variables, giving a final imputed estimation sample of all cases with valid survival times. Cox regression imposes no assumptions about the underlying hazard function and estimates the mean multiplicative impact variable on the log-hazard of failure at any given time. This gives rise to estimated hazard ratios (HRs) which, if the proportional hazards (PH) assumption is not violated, can be interpreted as the impact of the variable on the instantaneous failure rate at any given time. If the PH assumption is violated, these HRs can still be interpreted as the geometric mean of such a relationship within the sample and, further, stratification for variables where such an assumption is violated can obviate any such problems. In our base case, we stratify by stage of NPC at entry to the study and further stratify in sensitivity analysis.

Fully parametric regression assumes an underlying baseline hazard of failure over time that can be specified as a function of time. Such models are estimated under either the PH assumption or the accelerated failure time (AFT) assumption. We estimate models for baseline hazards parameterised as exponential, lognormal, loglogistic, Gompertz, Weibull, and generalised gamma. We inform the choice of such models by comparing estimated Akaike Information Criteria (AIC) and Bayesian Information Criteria (BIC) metrics, and by examination of existing literature. Where missing values are exhibited for certain variables, we impute values under the assumption that these are missing at random. We further assume that right-censoring of our survival times due to loss to follow-up is non-informative.

Summary of key findings

This study presents the first comprehensive and most extensive report on the survival of NPC patients for Indonesian cases. Despite being an endemic region, data on survival of NPC cases in Indonesia has been underreported to date. In this cohort, patients’ median OS of 31 months was relatively low compared to other reports from both endemic and non-endemic areas. When extrapolated to 5-year survival time, the OS rate presented in this data only reached 35%. There were significant survival differences between cases with younger age, longer duration of education, better ECOG performance status, earlier clinical stage, and those receiving combination of chemotherapy and radiation, and their counterparts.

Comparison with other studies in endemic and non-endemic areas

Low survival status of patients with NPC has been previously been reported from Indonesia. An earlier study by the team reported a 21-month median survival from 78 selected cases diagnosed from 2008–2011 which had a shorter follow-up time due to focusing on factors influencing the duration of radiation waiting lists [42]. Data from these cases were also included in the larger dataset reported in this manuscript. A further study in Indonesia demonstrated a 2-year OS of 39–71% in young, non-metastatic cases [37]. Combining this study with earlier reports, survival of Indonesian NPC patients is lower than those reported from surrounding endemic Asian countries which ranged from 54.7%-83.2 [12, 16, 19, 21, 43–45], and from non-endemic areas (ranging from 49.7%-69.1) for 5-year OS [17, 20, 46]. Studies that recruited only cases with early disease demonstrated much better survival (94.5%-96.5%) [47, 48]. The predicted rate for 5-year survival for our present cohort (35%) matched only the survival rate of patients in the US four decades ago, between 1973–1979 [21].

Prognostic factors and clinical and population health applicability

Prognostic factors favourable to survival in the present study included young age, education > 9 years, a good performance index, early stage of disease and undergoing a combination of chemotherapy and radiation treatment. Some studies reported various cut-off values (40–55 years) for univariable analyses regarding age [18–21, 28]. Our findings are consistent demonstrating a survival benefit in young age. Education has generally drawn attention as a crucial factor with worse survival in the lower educational groups [16, 49]. Large epidemiological studies have highlighted socioeconomic differences in cancer survival that are more pronounced than those in cancer incidence, with lower survival for individuals in disadvantaged socioeconomic conditions [50, 51]. However, teasing apart the causal argument from selection persists in making it difficult to determine the underlying mechanisms influencing this trend. The Indonesian government made it obligatory for an education duration of 9 years since 1994, 13 years prior to the intake of NPC cases in our current panel, but lower educational attainment persists. Alongside identifying and addressing barriers to completing minimum education years, there may be scope to explore health promotion approaches to encourage healthy lifestyles particularly in rural areas where education levels are reduced (see S1 Table).

Aligned with previous findings, performance index and nutritional status that was determined by BMI were found to be closely associated with stage of disease at presentation (see S2 Table), with both generally being low in the presence of advanced disease [31]. Tumour site and stage may directly impair the oral intake of patients with cancer in the head and neck area and a low performance status may enlarge the risk of nutritional worsening [52]. Bozetti et al also confirmed an association of primary site and ECOG status with nutritional risk score [53]. However, nutritional status of our patients did not have a significant impact on their survival status.

The late presentation of NPC in the local setting of Yogyakarta needs to be addressed. At both the health service and patient level, known barriers to earlier presentation exist. At the health service level, a poor diagnosis by general practitioners in the local primary health centres has previously been found to lead to a delay of patient referral [54]. Similarly, findings in Malaysia highlighted the low knowledge and awareness of NPC by health professionals may have an impact on delayed diagnosis [55] and patient outcomes [56]. For patients, low awareness may be linked to lack of visibility of the nasopharyngeal region, so tumour development may not be evident initially. Additionally, NPC-associated symptoms often mimic manifestation of other chronic diseases in head and neck area, such as chronic rhinitis, hearing impairments and headache [36]. Denial of illness and economic constraints may become further factors influencing delayed diagnosis and treatment [57]. Once identified, national universal health coverage ensures access to necessary treatments for NPC patients in Indonesia following diagnosis. This study highlights increased survival when presenting and accessing such treatments at an earlier stage of NPC. Wider benefits to patients of increased access to medical treatments through national universal health coverage are starting to be identified [58]. However, efforts to increase early presentation must be met by simultaneous efforts to reduce convoluted referral processes that may produce long waiting times between presentation and diagnosis.

Treatment strategy was found to have an impact on the survival status in NPC cases in this study. This aligns with findings from previous literature. When appropriate to the stage and in the absence of problematic comorbidity, various treatment strategy for NPC can provide good clinical outcomes and favourable survival [12, 34, 59]. For all stage I NPC without very bulky tumours, receipt of radiation alone is recommended with best option using IMRT [10, 34]. The addition of platinum-based chemotherapy improves clinical outcomes in physically fit patients with stage II-IVB either in the form of IC-CCRT,CCRT-AC, or CCRT alone [14, 16, 34, 59–62]. Sequential chemotherapy and radiation can be another strategy, especially in stage II NPC [11]. In our local panel, patients received radiation from 2D and 3D machines except those treated from early 2017 when IMRT was firstly introduced. There was variation in clinical outcome and the best result was achieved through CCRT and CCRT-AC approaches. Sequential chemo-radiation was widely applied especially before 2014. This strategy often became an option considering the long waiting lists for radiation treatment as shown by two previous studies from our local cases [42, 63]. and to accommodate a lower toxicity in patients with low performance status. The findings of the study suggest that an optimal strategy for NPC patients is a combination of radiotherapy and chemotherapy in locally advanced patients. Combination can be in a form of CCRT and CCRT-AC, although sequential treatment could be considered for those with low ECOG and old age. Whilst recommendations of treatment strategies for NPC in Indonesia can be guided by this study, they need to account for a context with limited equipment for delivering radiation therapies and complex referral systems.

All prognostic parameters obtained from this study may serve as targets for improvement. While the Indonesian government pursues efforts to provide optimal equipment in all oncology centres, optimization and refinement of the existing health service delivery could be effective to give NPC patients better access to optimal treatments and which may positively influence survival. A key area to direct investment will be in radiotherapy equipment which is necessary to support increases in capacity to deliver optimal treatments for NPC, alongside support wider capacity in supporting treatment delivery for other cancer types.

Strength and study limitation

A strength of this study is the inclusion of statistical analyses using fully parametric tests, supporting the first comprehensive report on Indonesian NPC survival. This also contributes to gaps in literature about NPC treatment in low- and middle-income countries. The study also has limitations. Our study design is retrospective, where it relies greatly on the availability of data from existing medical records, and may be subject to caveats when using routine data, such as potential coding errors by those creating and maintaining clinical records. A further limitation is the high number of patients lost to follow up in the life status (46%) which may introduce bias in the calculation of survival duration.