Incidence and Survival for Merkel Cell Carcinoma in Queensland, Australia, 1993-2010

Importance  Merkel cell carcinoma (MCC) is an uncommon but highly invasive form of skin cancer. The mechanisms that cause MCC are yet to be fully determined.

Objectives  To compare the incidence and survival rates of MCC in Queensland, Australia, known to be a high-risk area, with MCC incidence and survival elsewhere in the world. We also analyzed incidence trends and differences in survival by key demographic and clinical characteristics.

Design, Setting, and Participants  Retrospective cohort study of population-based administrative data for MCC collected by the Queensland Cancer Registry and supplemented with detailed histopathologic data. Deidentified records were obtained of all Queensland residents diagnosed as having MCC during the period from 1993 to 2010. A subsample of histopathologic records were reviewed by a senior dermatopathologist to determine the potential for misclassification. A total of 879 eligible cases of MCC were included in the study.

Main Outcomes and Measures  Incidence rates were directly age standardized to the 2000 United States Standard Population. Trends were examined using Joinpoint software with results expressed in terms of the annual percentage change. The period method was used to calculate 5-year relative survival, and adjusted hazard ratios were obtained from multivariate Poisson models.

Results  There were 340 cases of MCC diagnosed in Queensland between 2006 and 2010, corresponding to an incidence rate of 1.6 per 100 000 population. Men (2.5 per 100 000) had higher incidence than women (0.9 per 100 000), and rates peaked at 20.7 per 100 000 for persons 80 years or older. The overall incidence of MCC increased by an average of 2.6% per year from 1993 onwards. Relative survival was 41% after 5 years, with significantly better survival found for those younger than 70 years at diagnosis (56%-60%), those with tumors on the face or ears (51%), and those with stage I lesions (49%).

Conclusions and Relevance  Incidence rates for MCC in Queensland are at least double those of any that have been previously published elsewhere in the world. It is likely that Queensland’s combination of a predominantly white population, outdoor lifestyle, and exposure to sunlight has played a role in this unwanted result. Interventions are required to increase awareness of MCC among clinicians and the public.

Merkel cell carcinoma (MCC) is a rare and highly invasive form of neuroendocrine skin cancer that was first described by Toker¹ as trabecular carcinoma in 1972. The neoplasms are composed of small, round, blue cells sharing histomorphologic and immunohistochemical features with various neuroectodermally derived cells including cutaneous Merkel cells.^2,3 It is currently unclear whether MCCs actually derive from cutaneous Merkel cells or share a common precursor.^2,4,5

Merkel cell carcinomas can be difficult for clinicians to identify because of their nondescript features.⁶ However, just recently, Jalilian and colleagues⁷ reported that the 4 most frequent clinical features were cherry-red color, shiny surface, sharp circumscription, and nodular structure. They also outlined significant dermoscopic features including linear, irregular, and polymorphous vessels; poorly focused vessels; and milky pink areas.⁷ Despite the rather characteristic histopathologic features, immunohistochemical staining is required for a definitive diagnosis to differentiate MCC from a broad spectrum of small, round, blue-cell neoplasms (eg, metastasis of an oat cell carcinoma of the lung).^2,5,8 The tumors typically show rapid growth, meaning that many patients develop metastatic disease, and recurrence is common despite surgical removal of the primary lesion.^2,8

Merkel cell carcinoma is most often diagnosed among elderly white patients on sun-exposed body sites such as the head and neck.^2,5,8 Immunosuppression, which occurs when the body’s normal immune responses are inhibited due to medical intervention or disease, is another recognized risk factor,^2,5 which is suggestive of a viral cause.⁸ The discovery of a polyomavirus (MCPyV)⁹ in most MCC tumors has offered new insights into its pathogenesis,^8,10,11 although much remains unknown. For example, while there is an established association between MCC and UV irradiation,⁸ it is not clear whether UV irradiation contributes directly to the development of MCC or if it acts in combination with immunosuppression.¹⁰

The state of Queensland in northeastern Australia has a subtropical climate and high levels of UV radiation year-round combined with a predominantly fair-skinned population and an outdoor lifestyle. Queensland consequently has the highest reported incidence rate of skin cancers worldwide, particularly invasive melanoma.^12-14 The main aim of the present study was to compare the incidence of MCC in Queensland with that found elsewhere in the world. We also examined incidence trends and investigated differences in survival by key demographic and clinical characteristics.

Approval for this study was granted by the Griffith University Human Research Ethics Committee (PBH/34/11/HREC). Because the study involved a retrospective review of deidentified records, participant informed consent was not required.

Population-based data on all cases of MCC (defined as International Classification of Diseases for Oncology, Third Edition (ICD-O-3) code C44 and morphology code M8247-3) diagnosed in residents of Queensland between 1993 and 2010, inclusive, were obtained from the Queensland Cancer Registry. We excluded cases diagnosed prior to 1993 owing to the possibility of underreporting of MCC in this period.¹⁵ Cases of MCC were further restricted to those occurring on the skin.

The data available from the Queensland Cancer Registry included demographic information (sex, age at diagnosis, remoteness of residence¹⁶), diagnostic details (year and anatomic site), full history of any other cancer diagnoses, and, where relevant, items relating to mortality (time from diagnosis to death and cause of death). Additional clinical information was extracted from histopathology reports where available, including the size of the lesion, lymph node involvement, recurrence, metastases, and immunohistochemical results. Immunosuppression status was categorized as “known to be immunosuppressed” for persons who were a transplant recipient, tested positive for human immunodeficiency virus, or who were reported to have another cause of immune suppression. Cases were staged according to the criteria set out by the American Joint Committee on Cancer.¹⁷

A positive test result for cytokeratin 20 (CK20) was considered confirmation of a diagnosis of MCC, but other variants exist that are CK20 negative.^4,18 To estimate the potential for misclassification of cases included in the study, a subsample of approximately 1 in 10 of the histopathology reports for tumors coded to MCC where a negative result was returned for CK20 or where CK20 testing was not recorded were reviewed by a senior dermatopathologist (H.P.S.).

Annual incidence rates for the period 1993 through 2010 and average incidence rates for the latest 5 years (2006-2010) were generated by sex, age group, site, and stage at diagnosis. Estimated resident population information used for the rate denominators was obtained from the Australian Bureau of Statistics.¹⁹ To allow for consistency with other similar studies,^15,20 rates were directly age standardized to the 2000 United States Standard Population.²¹ Incidence rate ratios were calculated by simple division, with confidence intervals derived using the method specified by Kegler.²²

Incidence trends were analyzed using Joinpoint regression models (software developed by the National Cancer Institute, version 4.0.4). This approach quantifies the annual percentage change (APC) and specifies any significant changes to the magnitude or direction of the trend (known as a “joinpoint”) based on Monte Carlo permutation tests.²³ A maximum of 2 joinpoints were specified in each model, with a minimum of 5 years allowed between joinpoints or between a joinpoint and either end of the data series. Two-sided t tests were used to determine the statistical significance of the trends (P < .05).

Median survival time, defined as the time from diagnosis to censoring or death, was calculated using the reverse Kaplan-Meier method.²⁴ Follow-up was censored for subjects who were alive at the end of the study period and/or at the date of death for those who died from causes other than MCC for cause-specific survival analysis. One case was excluded from the survival analysis owing to diagnosis on the basis of death certificate only.

Five-year survival was evaluated using the period method,²⁵ which follows rolling groups of patients within a recent “at risk” window of time and thus has the advantage of producing more up-to-date survival estimates than the more traditional cohort method. In the current study, persons diagnosed as having MCC contributed to the survival calculations if theirs was a prevalent case at some time between January 2006 and December 2010. Estimates for both relative survival and MCC-specific survival were produced. Relative survival was calculated by dividing the observed survival probability for the study group by the expected survival within the Queensland population,²⁶ matched by age group, sex, and year. The Ederer II technique²⁷ was used to compute expected survival.

Differences in survival were assessed by applying multivariate Poisson models to examine excess mortality up to 5 years after diagnosis,²⁸ with results expressed in terms of adjusted hazard ratios (HRs). The dependent variable was the number of deaths, with an offset term for the log of the person-years at risk. A range of key demographic and clinical variables was included in the initial model to determine which of these characteristics had independent prognostic value. Variables were excluded if there was no evidence (P > .25) of their overall effect on survival (remoteness of residence¹⁶ and level of invasion). Sex, age group, body site, multiple primary cancers, stage at diagnosis, and surgical margins were retained in the final model. If the overall effect for a variable was not statistically significant (P ≥ .05), then there was deemed to be no difference in survival, even if there appeared to be individual differences in the adjusted HRs between some of the categories.

All results were presented with corresponding 95% confidence intervals (CIs).

A total of 903 cases of MCC were diagnosed in Queensland between 1993 and 2010. Of these, 24 were excluded because the lesion occurred on sites other than the skin (mainly on the lips), leaving 879 eligible cases in our cohort.

The distributions for some of the main characteristics of the cohort are listed in Table 1. Men accounted for around two-thirds of all cases (68%), with a median age at diagnosis of 75.5 years compared with 78.0 years for women. Half of the cases were diagnosed at stage I, while 12% already had lymph node metastases (stage III), and a further 15% had distant metastases (stage IV) at the time of diagnosis. Only 1% of patients in the cohort were known to be immune suppressed. More than a third (36%) were diagnosed as having another primary cancer apart from MCC, including 10 patients (1%) with chronic lymphocytic leukemia and 18 patients (2%) with a lymphoma.

The most common site was the face or ears (35%), although the distribution of MCC across the body sites differed by sex (P < .001). In particular, MCCs on the face or ears occurred more frequently for men (38%) than women (29%), while women were far more likely than men to have an MCC diagnosed on their lower limbs (26% and 11%, respectively). Women also had a higher proportion of MCCs that had not yet penetrated beyond the dermal layers of the skin (31% compared with 23% of their male counterparts). No statistically significant differences by sex were found among MCC patients by age group at diagnosis, remoteness of residence,¹⁶ presence of any multiple primary cancers (including second primary MCCs), stage at diagnosis, surgical margins, or immunosuppression status.

The average annual age-standardized incidence rate of MCCs in Queensland between 2006 and 2010 was 1.6 per 100 000 population (95% CI, 1.5-1.8; see the eTable in the Supplement). Incidence was almost 3 times higher for men (2.5 per 100 000) than for women (0.9 per 100 000). Rates increased rapidly by age, peaking at 20.7 per 100 000 for persons 80 years or older.

Incidence rate trends for MCC in Queensland increased by an average of 2.6% per year (95% CI, 1.1%-4.2%) between 1993 and 2010, equating to a total rise of 54% over that time interval (Table 2 and Figure 1). Significant and ongoing annual increases were found for men (+2.5% per year), people in the older age groups (70-79 years, +2.3%; 80 years or older, +3.7%), and tumors that occurred on the head (+3.1%). Significant increases were also observed for tumors that were either stage I (+3.6%) or stage II or higher (+2.9%).

There were 281 deaths due to MCC within the study group (32%) between 1993 and 2010. A further 82 people in the cohort (9%) died from other types of cancers; 213 deaths (24%) were from noncancer causes; and the remaining 303 persons (34%) were still alive at the end of 2010. Of the other cancer-related deaths, the main causes were other types of skin cancer and lung cancer (14 deaths each).

The median follow-up time for cases that were prevalent between 2006 and 2010 was 2.8 years (interquartile range, 1.2-6.1 years). Cause-specific survival was 88% (95% CI, 84%-91%) after 1 year and 66% (95% CI, 60%-71%) after 5 years. The corresponding estimates for 1- and 5-year relative survival were 77% (95% CI, 71%-81%) and 41% (95% CI, 34%-48%), respectively.

After adjustment for confounding variables, significant differences in survival remained for age group, site, and stage at diagnosis (Table 3 and Figure 2). Specifically, persons 80 years or older were 3 times more likely to die from MCC within 5 years of diagnosis compared with those aged 60 to 69 years (adjusted HR, 3.0; 95% CI, 1.8-5.1). Persons with an MCC on either the trunk (adjusted HR, 2.2; 95% CI, 1.2-4.1) or lower limbs (adjusted HR, 1.8; 95% CI, 1.1-3.0) had around double the risk of MCC-related mortality compared with MCC on the face or ears; however, there was no significant difference for lesions on the scalp or neck and upper limbs or shoulders in relation to the face or ears. The risk of MCC-related mortality was also around double for either stage II (adjusted HR, 1.8; 95% CI, 1.1-3.2) or stage IV tumors (adjusted HR, 2.3; 95% CI, 1.2-4.4) compared with stage I. No significant survival differences were found for sex, remoteness of residence, level of invasion, surgical margins, or the presence of multiple primary cancers.

To our knowledge, this report provides the first description of the epidemiology for MCC in the state of Queensland, Australia. Several recent population-based studies have been published on the incidence of MCC.^15,20,29-32 Rates in these other studies vary from 0.2 per 100 000 in Denmark²⁹ and Finland³² to 0.8 per 100 000 in Western Australia.¹⁵ Our results therefore reveal that the incidence rate of MCC in Queensland is at least double that previously reported anywhere else in the world.

Queensland has a predominantly white population living in an area with high levels of ambient UV radiation throughout the year.³³ The higher incidence reported here compared with other countries and/or regions therefore appears to support a link between exposure to UV radiation and MCC and is consistent with the findings of Agelli and Clegg,³⁴ who noted a correlation between the UV index and rates of MCC in the United States. In addition, most cases of MCC in Queensland were found on sun-exposed sites such as the face and ears, scalp and neck, and upper limbs and shoulders, whereas in a less sunny country such as Sweden, a relatively larger percentage of MCCs were observed on the trunk.³⁰ Similar to Girschik and colleagues¹⁵ working in Western Australia, which also has a high UV index, we found that MCCs occurred more frequently for men than in women. This contrasts with what has been reported in several places in Europe.^29-32

Average age at diagnosis in other studies ranged from 75 to 78 years, consistent with the median of 76 years of age in the Queensland cohort. However, a higher proportion of MCC tumors in Queensland were found to have distant metastases at diagnosis (15%) as opposed to MCC tumors studied elsewhere (4%-8%).^{15,20,29,31,32}

There was an annual increase of 8% in the incidence of MCC in the United States between 1986 and 2001,³⁵ while rates doubled in the Netherlands between 1993 and 2007.³¹ Although the magnitude of the increase in both of these countries has been considerably greater than that observed in Queensland, their incidence rates still remain much lower. The incidence of MCC has stayed fairly stable in Nordic countries since the mid-1990s.³² At least part of the rapid increase in incidence rates in the United States during the last 2 to 3 decades has been attributed to advances in diagnostic techniques, such as immunohistochemical analysis, along with greater awareness on the part of medical practitioners.^4,20 This means that a larger proportion of MCC cases are now correctly identified. Changes in the proportion of the population who are immunosuppressed could also have contributed.³¹ It is unknown to what extent these factors might have influenced incidence rate trends in Australia or why there has been no corresponding change in incidence rates of MCC in countries such as Finland, Sweden, Denmark, Iceland, and Norway.³²

Although the incidence rate of MCC is about 40 times lower than that of invasive melanoma in Queensland,³⁶ these 2 types of skin cancer share some epidemiologic characteristics. For example, they are more common among men and the elderly. The incidence of both is continuing to increase sharply among older people, while rates have stabilized among those who are younger.¹³ If the role of UV radiation in the development of MCC is found to be similar to that in melanoma development, then this stabilizing of rates among younger people may provide support for the effectiveness of prevention campaigns.³⁷

An analysis of data from Finland reported an estimate similar to ours for 5-year relative survival among men (36%) but a much higher rate for women (69%).³² Lemos et al¹⁷ calculated a 5-year relative survival rate of 54% for a large series of patients with MCC from the United States, whereas other studies have estimated relative and cause-specific survival rates of 62% and 64% after 5 years in the Netherlands³¹ and Western Australia,¹⁵ respectively. While these results are similar to our estimate for cause-specific survival, they are substantially higher than the relative survival rate in Queensland.

In most instances, relative survival closely resembles cause-specific survival estimates because these metrics are designed to measure the same outcome of net survival. An exception is when the underlying mortality in the cohort is substantially different from that of the general population after accounting for the disease of interest.³⁸ This could explain why there was a large disparity between the 2 measures in our study. Even so, it is still not clear why our estimate of relative survival was substantially lower than some of the other published results, although the higher proportion of cases with metastases at diagnosis in the Queensland cohort may have contributed to these differences to some extent.

Several prognostic factors for MCCs have been identified by other researchers, including sex, site, immune suppression, histopathologic type, growth pattern, and most importantly, the extent of disease at presentation (variously characterized by stage, tumor size or depth, and lymphovascular invasion).^5,15,20,39 The influence of MCPyV on survival is controversial,^10,11 but we were not able to measure this in the current study. We found significant differences in the risk of mortality by age, body site, and stage at diagnosis after multivariate analysis. One notable difference was that MCCs on the head and neck have been previously associated with a poorer prognosis,⁵ which is the opposite of what occurred in Queensland. The reasons for this are not evident.

The prognosis for people diagnosed with MCC is clearly worse than that for those with melanoma: 5-year relative survival for melanoma was 93% in Queensland for the period 2006 through 2010,³⁶ in contrast to 41% for MCC. It seems unlikely that this difference is due to melanomas being diagnosed earlier than MCCs: Girschik et al¹⁵ reported significantly worse stage-specific survival for MCC compared with melanoma for localized and regional disease but not for distant metastases. Another possible explanation is that differences exist in the general health of those diagnosed as having MCC vs melanoma. More than half of the deaths within the MCC cohort were due to other causes. It is therefore feasible that serious comorbidities were also common among those whose deaths were attributed to MCC, which would thus have had a negative impact in addition to that of the MCC tumor on patient survival.

Most previous studies on MCC have involved small samples of fewer than 200 patients. Therefore, a main strength of our work is the larger size of the cohort. The inclusion of all cases diagnosed in Queensland means that the study was not subject to some of the selection biases that can result from a hospital-based series of patients. Another advantage is the supplementation of standard registry data items with information from histopathology reports.

Only 1% of patients included in our data were documented as being immunosuppressed, compared with 6% in a population-based MCC cohort from Western Australia.¹⁵ Information about immunosuppression status is not routinely reported in Queensland, so our result was probably an underestimate of the true proportion. Unfortunately, the small number of immunosuppressed cases also meant that we were unable to assess survival for this subgroup.

While every effort was made to verify legitimate MCC cases, we did not have access to stored biological material (such as specimens and tissue blocks) to assist with this process. While immunohistochemical results other than CK20⁴⁰ were available to confirm the diagnosis in approximately 80% of the audited histopathology records, the audit revealed that up to 5% of the patients included in this study may have been overdiagnosed. However, this is a conservatively high estimate and is considered acceptable when taking into account the specific histopathologic diagnostic challenges when dealing with small, round, blue-cell tumors of the skin. Conversely, it is also possible that some cases of MCC were inadvertently excluded owing to incorrect classification as another type of cancer, but it was not possible to confirm this.

Our results establish that the incidence rate of MCC is much higher in Queensland than anywhere else in the world. The central factor behind this undesirable statistic appears to be a high level of exposure to UV radiation by a mainly fair-skinned population. While the greater attention placed on melanoma may be warranted given its higher incidence, people diagnosed as having MCC have greatly reduced survival expectations. In light of these findings, it is imperative that clinical practice guidelines for the diagnosis and management of MCC be developed and implemented within Australia as soon as possible,⁷ similar to those that already exist for other countries.^41,42 Public awareness campaigns are also required to alert people that melanoma is not the only lethal form of skin cancer. In particular, timely medical evaluation should be encouraged for the rapid appearance of a new lesion, with emphasis on the importance of people becoming familiar with what is normal for their own skin. It is hoped that these steps will lead to better outcomes for patients with MCC.

Corresponding Author: Danny R. Youlden, BSc, Cancer Council Queensland, PO Box 201, Spring Hill, Brisbane, Queensland 4004, Australia (dannyyoulden@cancerqld.org.au).

Accepted for Publication: January 24, 2014.

Published Online: June 18, 2014. doi:10.1001/jamadermatol.2014.124.

Author Contributions: Mr Youlden and Dr Baade had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Youlden, Youl, Fritschi, Baade.

Acquisition, analysis, or interpretation of data: Youlden, Soyer, Youl, Fritschi, Baade.

Drafting of the manuscript: Youlden, Baade.

Critical revision of the manuscript for important intellectual content: Soyer, Youl, Fritschi, Baade.

Statistical analysis: Youlden, Baade.

Administrative, technical, or material support: Youlden, Youl.

Questionnaire: Fritschi.

Study supervision: Fritschi, Baade.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported in part by National Health and Medical Research Council (NHMRC) Practitioner Fellowship APP1020145 (Dr Soyer); NHMRC Early Career Fellowship ID1054038 (Dr Youl); NHMRC Fellowship ID513706 (Dr Fritschi); a Cancer Council of Western Australia fellowship (Dr Fritschi); and NHMRC Career Development Fellowship ID1005334 (Dr Baade).

Role of the Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank the staff of the Queensland Cancer Registry who provided the data extract used in this study. The staff members were compensated for their contributions only in the normal course of their employment.