Changing epidemiology of nasopharyngeal carcinoma in Hong Kong over a 20-year period (1980–99): An encouraging reduction in both incidence and mortality



Epidemiological data from the Hong Kong Cancer Registry for the period 1980–99 were analyzed. Altogether 21,768 new cases of nasopharyngeal carcinoma (NPC) and 8,664 related deaths were registered. In both genders, the peak incidence occurred in the 50–59 years age group, and this age distribution pattern remained similar throughout. The age-standardized incidence rate steadily decreased from 28.5 in 1980–84 to 20.2 in 1995–99 per 100,000 males, and from 11.2–7.8 per 100,000 females, resulting in a total decrease of 29% for males and 30% for females over this 20-year period. The magnitude of total decrease in NPC mortality amounted to 43% and 50%, respectively, as the age-standardized mortality rate steadily decreased from 13.7 in 1980–84 to 7.8 in 1995–99 per 100,000 males, and from 4.5–2.2 per 100,000 females. The age-standardized mortality/incidence ratio also decreased from the peak of 0.48 in 1980–84 to 0.39 in 1995–99 for males, and from 0.40–0.29 for females. Females had significantly lower age-standardized incidence (male/female ratio 2.5–2.6, p < 0.01) and mortality (male/female ratio 3.0–3.5, p< 0.01) throughout the whole period. Furthermore, females had consistently lower mortality/incidence ratio: 0.29 vs. 0.39 in 1995–99. These data are highly suggestive of significant improvement in prevention and control of NPC in Hong Kong. Closer scrutiny of the differences in intrinsic and extrinsic factors between the genders might help to show important factors affecting oncogenesis and prognosis. Possible ways for further reduction of incidence and mortality are discussed. © 2002 Wiley-Liss, Inc.

Nasopharyngeal carcinoma (NPC) is unique for its peculiarly skewed racial and geographic variation in distribution. As reported by the International Agency for Research on Cancer (IARC) under the World Health Organization (WHO), the annual incidence in 1988–92 per 100,000 males ranged from <1 among Caucasians in Western countries to >20 among Southern Chinese.1 Even within a single country, there is wide variation in risk among different races: one good example is Singapore, where the annual incidence per 100,000 males varied from 18.5 for Chinese to 6.5 for Malays and 0.5 for Indians.

Another very interesting observation is substantial variation in risk for Chinese residing in different parts of China: the annual incidence per 100,000 males ranged from 1.6 in Tianjin (a northern city) to 24.3 in Hong Kong (a southern city). Descendants from Chinese who have migrated to western countries show progressively lower risk, but their incidence remains higher than the indigenous populations.2, 3, 4 A study on Chinese immigrants in California3 showed that the risk of NPC for the third generation was less than half of that for the second generation, but this risk was still 10-fold greater than Caucasians.

Furthermore, familial aggregation of NPC has been reported in diverse populations.5, 6, 7, 8, 9, 10 In a study on Southern Chinese by Yu et al.,10 NPC was detected in 6% of first-degree relatives of NPC patients as compared to 1% of first-degree relatives of controls in the same neighborhood.

These epidemiological observations suggest that the etiology of NPC include both inherited genetic predisposition and environmental factors. Early latent infection by the ubiquitous Epstein-Barr virus (EBV) and its reactivation probably plays an important role in the cancer formation process.11, 12, 13 Exposure to carcinogens in traditional southern Chinese food, (volatile nitrosamines in preserved salted fish, in particular), have also been incriminated.4, 14, 15, 16, 17, 18, 19

Dr. John Ho is a key pioneer in the clinical and epidemiological study of NPC. Among his numerous contributions, he established the Hong Kong Anti-Cancer Society to promote public education and the Hong Kong Cancer Registry to study the local epidemiological pattern. Concerted efforts were made to rouse public awareness on the importance of early detection and the potential risk of preserved food. With rapid economic growth and the amazing development of Hong Kong from a fishing village into a metropolitan city, the life-style of most citizens also changed substantially. Shifting from traditional Chinese to western diets, in particular, could theoretically reduce the incidence of NPC. Together with technological advances and increasing resources for improving treatment outcome, the related mortality might also be reduced.

The current study on the epidemiology of NPC in Hong Kong over a 20-year period (1980–99) thus provides an interesting opportunity not only to assess our efforts to combat this cancer, but also to look for possible etiological clues and directions for future improvement.


The data were obtained from the Hong Kong Cancer Registry, which is a population-based registry for cancer statistics. In addition to voluntary notification, the Registry has established access enabling direct extraction of data on cancer incidence and mortality from all related government/ public departments and major private institutes.

Sources of data on clinical and histopathological diagnoses of cancers include:

  • All clinical oncology centers under the Hospital Authority,

  • All pathology departments under the Hospital Authority and the Department of Health,

  • All regional hospitals under the Hospital Authority,

  • All departments of radiotherapy and oncology in the private sector,

  • Major pathology departments/institutes in the private sector,

  • Births, Deaths and Marriages Registry of the Hong Kong Government,

  • Voluntary notification by private hospitals/medical practitioners.

The raw data were coded and checked for eligibility by a series of comprehensive cross-checking programs before registration. First, they were meticulously checked to eliminate duplications by using the Hong Kong Identity Card Number (unique for each citizen), together with each patient's name and date of birth. Validity of the data was further tested for compatibility of site-gender, site-age and site-pathology combinations. The source of data was documented and histological confirmation traced as far as possible. Original records were checked in case of doubt. The incidence rates calculated were then compared to reports from previous years, with further verification sought if obvious anomalies or inconsistency was detected.

Data on cancer deaths were collected from the Births, Deaths and Marriages Registry, which is the sole government department responsible for registering all death events that occur in Hong Kong. The information was used not only for calculating mortality rates, but also for cross-checking and supplementing the incidence data, as cause(s) of death, both primary and contributory, were clearly recorded on the death certificate issued.

The denominators for calculation of various rates were based on the population demographic data obtained through regular census by the Census and Statistics Department of the Hong Kong Government.

Age-specific incidence per 100,000 population (ASp-I) =

equation image

Age-specific mortality per 100,000 population (ASp-M) =

equation image

The respective age-standardized incidence rate (WSt-I) and mortality rate (WSt-M) were the corresponding rates adjusted to the World Standard Population defined by WHO.20 Differences between the genders were studied by comparing the standardized rates in the male population with corresponding rates in the female population using the χ2 test. Changes over time were estimated by fitting a linear regression line to the natural logarithm of standardized rates as a regressor variable. Thus the calculations for testing were most powerful for the hypothesis that these rates changed at a constant rate over the time interval.

Analyses were carried out using both the annual rates and 5-year group rates. Although some random variation in rates did exist across the years, the conclusions were similar. Hence, the 5-year group rates are used in the current paper for easy description and illustration of the overall trends.



During this 20-year period (1980–99), a total of 21,768 new cases of NPC had been registered: 15,801 males and 5,967 females. The proportion of cases with documented histological confirmation increased from <80% in 1980–84 to >90% in 1995–99 (Table I). In addition, the proportion of incidence unreported until death certification progressively decreased from 5% to 0.4% in the corresponding period.

Table I. Proportion (%) of Diagnosis with Histological Confirmation and Incidence Reported by Death Certificate Alone
PeriodDiagnosis with histological verificationReport by death certificate only1
  • 1

    Diagnosis not registered in sources other than death certificate.


Table II summarizes the incidence rates for different gender and age groups during different periods. The ASp-I ranged widely from 1.4 for females aged <30 years during 1995–99 to 83.1 for males aged 50–59 years during 1980–84. When standardized to the World Standard Population, the most recent WSt-I (1995–99) was 20.2 for males and 7.8 for females.

Table II. Incidence Rates Per 100,000 During Different Periods
PeriodAge-specific incidence rate for different age groupsAge-standardized incidence rate
Male population       
 Change (%)−16−27−34−29−21−26−29
Female population       
 Change (%)−10−27−36−33−16−37−30

Figure 1 shows the ASp-I by gender for the whole cohort. NPC was very rare among population younger than 30 years: the ASp-I was ≤3.0 for males and ≤1.6 for females. The incidence then rose sharply to reach its peak in the 50–59 years age group, with ASp-I ≥59.3 for males and ≥17.8 for females. This pattern was observed in the male cohort throughout the whole period. A minor fluctuation was noted in the female cohort with ASp-I slightly higher in the 40–49 years group during 1980–84 (29.6 vs. 26.4) and 1995–99 (19.1 vs. 17.8), but there was no consistent trend of shifting to younger population.

Figure 1.

Age-specific incidence rates by gender for the whole cohort.

A progressive decrease in incidence rate was noted in both genders (Fig. 2 and Table II) : WSt-I decreased steadily from 28.5 in 1980–84 to 20.2 in 1995–99 for males, and from 11.2–7.8 for females in corresponding periods. With an average reduction of 11.1% for males (p < 0.01) and 11.5% for females (p = 0.01) per 5-year period, the total decrease in WSt-I during this 20-year period amounted to 29% and 30%, respectively. Reduction occurred in all age groups: the greatest magnitude was 34% for males aged 40–49 years and 37% for females aged ≥70 years.

Figure 2.

Age-standardized incidence and mortality rates during different periods for (a) male and (b) female population.


Altogether 8,664 patients had died of NPC during this period: 6,619 males and 2,045 females. Table III summarizes the mortality rates for different gender and age groups during different periods. The ASp-M ranged widely from 0.2 for females aged <30 years during 1985–99 to 44 for males aged 50–59 years during 1980–84. The peak mortality occurred mostly in patients aged 50–69 years in both genders. When standardized to the World Standard Population, the most recent WSt-M (1995–99) was 7.8 for males and 2.2 for females.

Table III. Mortality Rates Per 100,000 During Different Periods
PeriodAge-specific mortality rate for different age groupsAge-standardized mortality rate
 Male population       
 Change (%)−60−58−53−42−27−30−43
Female population       
 Change (%)−24−56−64−55−33−28−50

As with incidence, a progressive decrease in mortality was noted in both genders (Table III and Fig. 2): WSt-M decreased steadily from 13.7 in 1980–84 to 7.8 in 1995–99 for males, and from 4.5–2.2 for females in corresponding periods. With an average reduction of 16.6% for males (p < 0.01) and 20.1% for females (p = 0.02) per 5-year period, the total decrease in WSt-M during this 20-year period amounted to 43% and 50%, respectively. Reduction occurred in all age groups: the greatest magnitude was 60% for males aged <30 years and 64% for females aged 40–49 years.

Table IV shows the age-standardized mortality/incidence ratio (WSt-M/WSt-I) during different periods. In both genders, the worst ratio was recorded in 1980–84: 0.48 for male and 0.40 for female. A steady decrease then occurred in subsequent years with mortality/incidence ratio falling to 0.39 for males and 0.29 for females in 1995–99.

Table IV. Mortality to Incidence Ratio for Each Gender During Different Periods
PeriodAge-standardized mortality/incidence ratio

Difference by gender

Throughout the whole period, the WSt-I in females was significantly lower than that in males (p < 0.01): the age-standardized male/female ratio in incidence varied within a narrow range of 2.5–2.6 (Table V). Similarly, the WSt-M in females was significantly lower than that in males (p < 0.01): the age-standardized male/female ratio in mortality varied from 3.0–3.5 (Table V). Furthermore, the age-standardized mortality/incidence ratio was consistently lower in females throughout (Table IV).

Table V. Male/Female Ratio in Incidence and Mortality Rates During Different Periods1
PeriodAge-standardized incidence rateAge-standardized mortality rate
M/F ratio95% CIp-valueM/F ratio95% CIp-value
  • 1

    CI, confidence interval; M/F, male/female ratio.



The Hong Kong Cancer Registry is a full member of the International Association of Cancer Registries under the WHO. The quality of data and the statistical analyses fully comply with international standards. Reports on cancer incidence in Hong Kong for years 1968–92 have been published in the previous IARC Scientific Publications ‘Cancer Incidence in Five Continents.’1, 21, 22, 23, 24 Although notification of new cancer is not a statutory requirement in Hong Kong, the systems established by the Hong Kong Cancer Registry have ensured direct access to data on cancers diagnosed/treated in all public departments and major private institutions. The coverage of population data for NPC is particularly comprehensive, as all centers with radiotherapy facilities (the primary treatment modality for NPC) are included in the existing network.

There is little doubt that the quality of data is improving progressively. As the nasopharynx is anatomically inaccessible to direct inspection/palpation, the diagnosis could have been missed easily in the past. The advent of modern imaging technologies and flexible endoscopes have contributed substantially not only to reveal the tumor, but also to locate the most suspicious sites for biopsy. The proportion of patients with histologically confirmed ante-mortem diagnosis progressively increased from ≤80% in 1980–84 to ≥00% in 1995–99 (Table I). Together with increasing access to all possible sources of data and meticulous documentation, the cancer statistics recorded in latter periods are increasingly accurate and complete. As potential biases due to past inadequacies would more likely lead to under-diagnosis and under-reporting, the progressive decreases in incidence and mortality observed in the current analyses are likely to be genuine, and indeed the actual magnitude might even be greater than that observed currently.

The consistent falling trend in WSt-I (Table II, Fig. 2) resulted in a total decrease of 29% for males and 30% for females during this 20-year period. As there is no substantial change in the proportion of Chinese in the community according to the Census and Statistics Department of the Hong Kong Government, the genetic background was relatively stable. On the other hand, the life-style for most citizens changed progressively from traditional Southern Chinese style to a more western style, particularly in terms of diets. Preserved salted fish is no longer a common food for most households. Although it remains difficult to single out the exact etiological factor(s), as the sum total of life-style includes many different aspects, we can at least confidently postulate that the declining incidence of NPC in Hong Kong is mainly attributed to changing environmental risk factors.

Interestingly, review of the incidence in other representative communities over a similar period shows that Hong Kong is the only place where such significant reduction was achieved (Table VI). Although Singapore had also shown rapid economic growth during this period, the annual WSt-I only decreased from 19.4 in 1973–77 to 18.5 in 1988–92 for Chinese males, and from 7.5–7.3 for Chinese females. The low incidence in non-endemic communities also remained static.

Table VI. Age-Standardized Incidence Rate Per 100,000 in Different Communities During Different Periods
Location and raceMaleFemale
  1. 1NA, not available; LA, Los Angeles; NSW, New South Wales.–2Reference 22, Waterhouse et al.–3Reference 23, Muir et al.–4Reference 24, Parkin et al.–5Reference 1, Parkin et al.

 Hong Kong32.930.028.524.314.412.911.29.5
United States        
 SEER: WhiteNANA0.50.5NANA0.20.2
 SEER: BlackNANA0.80.9NANA0.30.2
 LA: Chinese7.
England and WalesNA0.40.40.4NA0.20.20.2
Australia: NSW0.

Similar to the pattern observed among Chinese immigrants in Western countries,3 our incidence rates remained the highest reported despite the remarkable decrease over the years: the WSt-I in 1995–99 was 20.2 for males and 7.8 for females. In addition to environmental factors, it is highly likely that inherited genetic predisposition also plays an important role in the oncogenesis.

Specific haplotypes in human leucocyte antigen (HLA), situated on the short arm of chromosome 6, had been found to be associated with increased risk of NPC.25, 26, 27, 28 A study in the United States by Burt et al.28 showed that the associated haplotypes were different between Chinese and Caucasians; the HLA-A2 allele found specifically in Non-Chinese might be associated with a protective effect.

An extensive review of the genetic changes by Huang and Lo29 suggested that involvement of HLA, genetic susceptibility factor(s), and loss of heterozygosity (LOH) at chromosomes 3p and 9p might all play an important part in the early stage of cancer development. They postulated that genetic alterations might be induced by chemical carcinogens in the environment that lead to subsequent transformation of normal epithelium to low-grade dysplasia. Latent EBV infection and over-expression of bcl-2 protein (that inhibits apoptosis) might then affect the progression of low-grade precursors to high-grade lesions. Other possible changes in the pathogenesis include activation of telomerase, inactivation of p16 (tumor suppressor) gene, LOH of Chromosomes 11q and 14q, and over-expression of c-myc and ras oncogenes. Moreover, it is thought that overexpression of p53 protein is associated with metastatic processes.

Thus far, there are little data to suggest that the sex chromosomes are involved in the oncogenesis. The epidemiological patterns in all communities clearly show, however, that female population has significantly lower risk of developing NPC (Tables II,V,VI). The male/female ratio in age-standardized rates in Hong Kong in 1995–99 was 2.6 (95% CI = 1.6–4.0, p < 0.01) in terms of incidence (Table II) and 3.5 (95% CI = 1.5–6.9, p < 0.01) in terms of mortality (Table III). This male excess was observed in all age groups throughout the whole period (Table V). Furthermore, the age-standardized mortality/incidence ratio was consistently lower in the female population (Table IV): 0.29 vs. 0.39 in 1995–99, suggesting that females achieved a higher survival rate after treatment than male patients.

It is indeed a common finding in both endemic and non-endemic communities that female patients have a more favorable prognosis.30, 31, 32, 33, 34, 35 Detailed analyses by Teo et al.35 showed that gender was a significant factor (independent of age, histology and stage) for disease-free survival and overall survival: risk ratio for male vs. female = 1.4, p = 0.01. The local failure rates achieved by both genders were similar, but female patients showed significantly lower risk of distant metastases.

The reason why females of all races have a lower risk of developing NPC and dying from it remains an interesting puzzle. Close scrutiny of genetic factors together with other intrinsic and extrinsic differences between the genders might help to reveal important factor(s) in oncogenesis and the metastatic process, and hence lead to more effective prevention and disease control.

Although the male population had a higher risk than female in all corresponding periods, it is encouraging to note that both genders showed a progressive decrease in their mortality rates (Fig. 2) resulting in a total decrease in WSt-M of 43% for males and 50% for females during the study period (Table III). The age-standardized mortality/incidence ratio steadily decreased from the peak of 0.48 in 1980–84 to 0.39 for males in 1995–99, and from 0.40–0.29 for females in the corresponding periods (Table IV). Hence the decrease in mortality is not caused by the fall in incidence rate alone, but also by other factors (possibly earlier presentation or improved treatment).

Although there is yet no screening program for the general public by the Hong Kong government, intensive screening by annual physical examination, EBV serology test and nasopharyngoscopy is provided for the high-risk group (the first-degree relatives of NPC patients). Studies are being conducted to assess the cost-effectiveness of such a program. With recent advances in diagnostic tools, improvement in public awareness and streamlining of referral system, delay in diagnosis is gradually decreasing. Previous studies on patients referred to public cancer centers in Hong Kong showed that the mean duration from onset of symptom(s) to establishment of histological diagnosis decreased from 9 months in 1976–80 to 7 months in 1981–8536 and 5 months in 1996.37 This duration significantly affected the stage distribution pattern (the OR of presenting with Stage I–II vs. III–IV disease was 0.98 per month delay, p < 0.01), and this in turn affected the disease-specific survival.36

Using the cohort of NPC patients treated in our public centers in 1996, detailed analyses of the mean time taken for the various steps from onset of symptom(s) to commencement of radiotherapy showed that 45% of the mean delay was attributed to patients, 20% to initial attending doctor(s), 14% to waiting for diagnostic investigations, and 21% to preparation and waiting for radiotherapy. Obviously, much has yet to be done to minimize all aspects of delay. Besides further efforts in public education, we also have to educate family physicians and to increase availability of both diagnostic and therapeutic facilities.

One of the major treatment improvements in Hong Kong is the increase in the number of linear accelerators (and corresponding manpower) from a total of 2 in 1975 to 24 in 1999 (that is, an increase from 0.4–3.6 accelerators per million population). This expansion not only enables oncologists to shorten waiting time for radiotherapy, but also to provide better quality services with optimal technique and dose fractionation. Statistics from the Hospital Authority in 2001 showed that the median waiting time for radiotherapy in public cancer centers was 3.3 weeks. Additional linear accelerators are still needed to further shorten the waiting time, and to cope with the increasing demand by the growing and aging population.

Other improvements during this period include increasing accuracy in staging, addition of more potent chemotherapy and improvements in supportive care. Retrospective analyses of treatment outcomes from the largest center in Hong Kong did show significant improvement in disease-specific survival over time.38

Given the known limitations in interpretation of changing epidemiology over an extended period, the current data are indeed most encouraging. Further improvements, however, are obviously still needed. Besides vigorous scientific research to identify the key etiological factors, public education to promote primary/secondary prevention and early presentation, management provisions to minimize delay, and clinical research to enhance treatment efficacy, the Hong Kong Cancer Registry should also be expanded (to include survival analyses and information on grading, staging, and treatment) to facilitate better planning of health care strategies and monitoring of treatment outcome. In addition, in-depth reviews of screening options are needed to identify the most cost-effective screening program affordable by the community. The combat against NPC demands concerted efforts by all.