Cancer Risk Among Patients With Myotonic Muscular Dystrophy FREE

Myotonic muscular dystrophy (MMD) is an autosomal-dominant, multisystem disorder comprising 2 subtypes.¹ Type 1 MMD (Online Mendelian Inheritance in Man [OMIM] 160900) is caused by unstable trinucleotide (CTG) repeat expansion in the 3′ untranslated region of the dystrophia myotonica-protein kinase (DMPK) gene^2- 4; type 2 MMD (OMIM 602668) is a tetra-nucleotide (CCTG) repeat expansion in intron 1 of the zinc finger 9 (ZNF9) gene.^5- 6 It is the most common adult muscle dystrophy, with an estimated prevalence of 1 in 8000.⁷ Type 1 MMD displays a more severe phenotype that can present at any age (median, 20-30 years)¹ and result in premature death (at 50-65 years of age).^8- 9 Both subtypes share myotonia and progressive skeletal muscle weakness and wasting. Other manifestations include cardiac conduction defects, insulin resistance, testicular atrophy, respiratory insufficiency, cognitive impairment, and premature cataract.¹ Both MMD subtypes result from interactions between CUG or CCUG repeat RNA and regulatory binding proteins, mainly the muscle-blind-like (MBNL) protein family,¹⁰ which lead to abnormal regulation of pre–messenger RNA alternative splicing.¹¹ Depletion of MBNL1 in MMD has been implicated in the development of myotonia, premature cataract,¹² and skin cancer.^13- 14

Case reports have suggested that MMD patients may be at increased risk of benign and malignant tumors. Pilomatricoma, a rare benign calcifying cutaneous neoplasm derived from hair matrix cells, is the most commonly reported. Additionally, multiple skin basal cell carcinomas have been suggested as an MDD phenotypic variant.^14- 17 We have previously reviewed this literature, described neoplasms reported by participants enrolled in the National Registry of Myotonic Dystrophy and Facioscapulohumeral Muscular Dystrophy and Family Members, and discussed possible molecular reasons for a hypothetical cancer predisposition.¹³

To further explore whether the MMD phenotype includes cancer risk, we conducted a population-based linkage study of MMD patients using the nationwide Swedish and Danish registries. Our data provide the first objective, quantitative evidence to suggest that the MMD syndrome includes cancer susceptibility.

Within the Swedish Hospital Discharge Register, which began in 1964 and reached 100% nationwide hospitalization coverage in 1987, we identified all patients with an MMD discharge diagnosis (ICD-9 code 359C, ICD-10 code G711) between January 1987 and December 2004 (n = 768) (Figure). Diagnoses were coded using the International Classification of Diseases (ICD) revision 7 (1964-1968), revision 8 (1969-1986), revision 9 (1987-1995), and revision 10 (after 1995).^18- 19 We excluded 99 patients from analysis: 59 had cancer before or during the first MMD hospitalization, 36 died during the first hospitalization, and 4 had incomplete data. Using the Swedish national identification numbers, we linked MMD patients to the Cancer Registry. The Swedish MMD cohort did not capture data for patients who were managed exclusively as outpatients. When analyses in Swedish MMD patients suggested a possible excess cancer risk, we sought to replicate our findings in a separate, independent population (the Danish cohort).

The Danish National Patient Registry covers all hospital admissions and outpatient visits since 1977 and 1995, respectively. All Danish registry diagnoses are coded according to the ICD revision 8 (1969-1993) and revision 10 (from 1994).²⁰ Each individual's unique civil registration number was used to link the patients to the Danish Cancer Registry. After excluding registrants with prior cancer and those who died prior to follow-up initiation, we identified 989 inpatients or outpatients discharged with an MMD diagnosis between 1977 and 2008 (ICD-8 codes 330.90, 330.91; ICD-10 code G711) (Figure).

Access to the Swedish and Danish registry data was approved by the Karolinska Institutional Review Board and the Danish Data Protection Agency, respectively. Informed consent was waived because we had no contact with study participants. An institutional review board waiver was obtained from the National Institutes of Health Office of Human Subjects Research because all analyses were performed using deidentified data.

The Swedish and Danish cancer registries have identified all incident cancers detected in Sweden and Denmark since 1958 and 1943, respectively. Registry completeness and diagnostic accuracy exceeded 95% in several validation studies.^21- 24 For this study, cancer sites were identified using ICD-7 and ICD-10 codes from Sweden and Denmark, respectively. To ensure comparability between reported cancer sites, we used a slightly modified version of the NORDCAN cancer dictionary,²⁵ which was designed by the Association of the Nordic Cancer Registries and the International Agency for Research on Cancer to formalize Nordic countries' data harmonization. Amendments to the cancer codes were made for cancers of the brain (only malignant tumors were included), as well as lymphoma and leukemia, to accommodate differing coding schemes over time in Denmark and Sweden, respectively. Furthermore, we included subanalyses of rectal and anal cancers in recognition of their etiologic differences. Nonmelanoma skin cancers were not considered in the current analyses because of differing registration practices between the 2 countries.²⁴

We calculated standardized incidence ratios (SIRs) for all cancers combined and by anatomic site, stratified by sex and age (<50 and ≥50 years). Each SIR was calculated by dividing the number of observed cancers in MMD patients by the expected number of cancers. Expected cancer numbers were calculated by applying country-, age-, sex-, and calendar time–specific population incidence rates from each cancer registry to the person-years observed among its subset of MMD cases. To prevent survival bias from affecting cancer risk estimates, MMD patients were followed up from date of first MMD-related inpatient discharge diagnosis, or date of first outpatient contact, to the first cancer diagnosis, death, emigration, or the end of complete cancer registration (Sweden: December 31, 2004; Denmark: December 31, 2008). The observed and expected cancer rates for MMD patients were calculated by dividing the observed and expected numbers of cancers by the person-years of follow-up.

In a subgroup analysis using the Danish database, we calculated SIRs for all cancers combined, stratified by type of hospital contact (inpatient vs outpatient) as a proxy for disease severity, hypothesizing that patients who were hospitalized represented a more severe MMD phenotype. Patients who were first identified from the outpatient registry and later hospitalized (n = 125) contributed person-years of follow-up to the outpatient group until the first hospitalization date and subsequently contributed person-years to the inpatient group. Of note, 331 of 456 Danish MMD patients (72.6%) first ascertained in the outpatient setting were never hospitalized during follow-up.

We evaluated the risk of each cancer site and considered an association to be statistically significant at P = .05. In the “Results” and “Comment” sections, we focused primarily on associations with P values less than .01 to minimize testing-related false discovery. Mid P tests and confidence intervals were used throughout, defined by the mean value of the Poisson distribution that makes the probability the test statistic is less than its observed value plus half the probability of its observed value equal to 0.975 (upper limit) and 0.025 (lower limit).²⁶ Subgroup interactions were tested using conditional exact tests with mid P values. Because the site-specific SIRs did not differ statistically in the 2 national cohorts, the data were combined for presentation. (See eTable 1, for country-specific data).

The study included 1658 MMD patients (Sweden = 669; Denmark = 989), contributing 4724 and 9446 person-years of observation, respectively. In Sweden, the median age at first MMD discharge diagnosis (inpatient only) was 46 years vs 38 years in Denmark (41 years for inpatients and 37 years for outpatients). In both countries, approximately half were men, 40% died during follow-up, and 6% developed cancer (Table 1).

During 14 170 person-years of follow-up, 104 MMD patients developed cancer compared with 52.3 expected cases, corresponding to an observed cancer rate of 73.4 per 10 000 person-years in MMD patients vs an expected rate of 36.9 per 10 000 person-years. Compared with expected case numbers based on cancer rates in the general population, MMD patients had an increased overall cancer risk (SIR, 2.0; 95% CI, 1.6-2.4). Most notably, we observed significant excesses of endometrial cancer (n = 11; observed rate, 16.1/10 000 person-years; SIR, 7.6; 95% CI, 4.0-13.2), brain cancer (n = 7; observed rate, 4.9/10 000 person-years; SIR, 5.3; 95% CI, 2.3-10.4), ovarian cancer (n = 7; observed rate = 10.3/10 000 person-years; SIR, 5.2; 95% CI, 2.3-10.2), and colon cancer (n = 10; observed rate, 7.1/10 000 person-years; SIR, 2.9; 95% CI, 1.5-5.1). Our data also suggested possible excesses of eye cancer (n = 2; observed rate, 1.4/10 000 person-years; SIR, 12.0; 95% CI, 2.0-39.6), other female genital organ cancer (n = 2; observed rate, 2.9/10 000 person-years; SIR, 9.6; 95% CI, 1.6-31.8), thyroid cancer (n = 3; observed rate, 2.1/10 000 person-years; SIR, 7.1; 95% CI, 1.8-19.3), and pancreatic cancer (n = 4; observed rate, 2.8/10 000 person-years; SIR, 3.2; 95% CI, 1.0-7.6) (Table 2). Close similarity in overall and site-specific cancer excess was observed in both Swedish and Danish MMD patients (eTable 1).

After excluding genital organ cancers (uterus, cervix, ovary and fallopian tubes, vulva, vagina, prostate, testis, penis, scrotum, and unspecified parts), no statistically significant sex difference was observed in overall cancer risk (SIR, 1.9; 95% CI, 1.4-2.5 in women vs SIR, 1.8; 95% CI, 1.3-2.5 in men; P = .81 for heterogeneity; observed rates, 64.5 and 47.7/10 000 person-years in women and men, respectively). However, the data suggested possible sex-specific differences for cancers of the rectum and lung (Table 3), findings that did not reach statistical significance.

In age-stratified analyses (<50 vs ≥50 years), no statistically significant difference was observed in overall cancer risk (SIR, 2.2; 95% CI, 1.4-3.2 and SIR, 1.9; 95% CI, 1.6-2.4, respectively; P = .58 for heterogeneity; observed rates, 25.7 and 165.6/10 000 person-years in the younger and older groups, respectively). However, we did observe a significantly higher excess risk of endometrial cancer among women younger than 50 years (n = 5; observed rate, 11.1/10 000 person-years; SIR, 35.6; 95% CI, 13.0-78.9) vs women 50 years and older (n = 6; observed rate, 25.8/10 000 person-years; SIR, 4.6; 95% CI, 1.9-9.5; P = .003 for heterogeneity) (Table 4).

In a subgroup analysis among Danish patients by type of hospital contact, our data suggested that inpatients were at higher relative risk of developing cancers than outpatients. The difference, however, did not reach statistical significance (SIR, 2.0; 95% CI, 1.5-2.6, vs SIR, 1.1; 0.5-2.0, respectively; P = .08 for heterogeneity).

In 3 additional analyses based on the Danish cohort, we found the overall results to remain the same when starting follow-up 5 years after the first MMD discharge diagnosis (SIR, 2.04; 95% CI, 1.49-2.72), when restricting the analysis to those with MMD as their main diagnosis (n = 702; SIR, 1.8; 95% CI, 1.4-2.4), and when considering those inpatients with no diagnoses other than MMD at the first MMD admittance (n = 255; SIR, 1.9; 95% CI, 1.3-2.8). Among 287 MMD patients who had a main discharge diagnosis other than MMD, a family history of congenital or specified condition (n = 34) was the most commonly reported diagnosis, followed by respiratory problems (n = 33, 33% of whom had respiratory failure), cardiovascular problems (n = 27), and eye diseases (n = 22, 55% of whom had cataract).

Our study is the first to our knowledge to quantify cancer risk in patients with MMD. In this large population-based study, we observed an excess cancer risk compared with the general population, first among Swedish MMD patients and then among the replication cohort of Danish patients. Because of the close similarity in the results, we combined the findings from these 2 cohorts to improve statistical power. The elevated overall cancer risk was primarily due to malignancies of the endometrium, brain, ovary, and colon.

Case reports have suggested a strong association between MMD and pilomatricoma, a rare benign skin neoplasm (which is not registered in the Swedish Cancer Registry and is incompletely registered in the Danish Cancer Registry) and also included reports of a number of rare malignancies.¹³ Published case reports tend to present unusual cases and therefore cannot provide conclusive evidence of a genuine association. This methodological shortcoming led to the current study, which provides strong evidence that MMD may in fact be a cancer susceptibility disorder.

Several biological mechanisms for the apparent increased cancer risk have been proposed, including possible RNA-mediated alterations in tumor suppressor genes or oncogene expression, modification of the coding features of proteins,²⁷ up-regulation of the Wnt/β-catenin signaling pathway, or all of these.¹³ Of note, alterations in RNA binding proteins, suggested as a key player in MMD pathogenesis, have been observed in human carcinogenesis.²⁸ On the other hand, it is worth noting that the type 1 MMD gene product—DMPK (OMIM 605377)—is a protein kinase, ie, a member of a large gene family that contains numerous examples of cancer susceptibility genes, such as RET (multiple endocrine neoplasia type 2), STK11 (Peutz-Jeghers syndrome), PRKAR1A (Carney complex), RAF1 (Noonan syndrome), ALK (neuroblastoma), and PDGFRA (gastrointestinal stromal tumor).²⁹

The absence of an excess cancer risk in other repeat disorders, such as fragile X syndrome³⁰ and Huntington disease,³¹ is noteworthy. It is possible that repeat expansion size may be a key determinant of cancer risk in this context, because nucleotide repeat expansions are longer in MMD patients compared with Huntington disease or fragile X patients,³² and case reports have demonstrated longer nucleotide repeat expansion in tumor tissue from MMD patients compared with their normal tissue.^33- 34 If proven true, we would expect that patients with type 2 MMD, who are known to have the longest repeat sizes, would have higher risk of cancer, a hypothesis that needs further investigation. The observed cancer risk differences between various repeat disorders might also be related to the precise repeat expansion location within the affected gene. Myotonic muscular dystrophy differs from Huntington disease by having expansion in a noncoding region, which is more likely to produce a toxic RNA-mediated gain-of-function that can affect downstream effector genes,^5,35 some of which may be tumor suppressor genes, eg, mismatch repair genes. Additional analyses of paired normal tissue and tumor samples from well-characterized MMD patients could shed further light on the relationship between expansion repeat length and cancer risk in MMD.

Surprisingly, the cancer spectrum we observed in MMD patients included many of the same excess cancers observed in patients with hereditary nonpolyposis colorectal cancer (HNPCC),³⁶ eg, cancers of the colon, brain, ovary, and endometrium. In that context, MMD-related pilomatricoma may be analogous to the increased risk of sebaceous adenoma in the Muir-Torre variant of HNPCC. Inherited DNA mismatch repair abnormalities are the genetic basis for HNPCC, a prototypic cancer susceptibility disorder.³⁶ Defective mismatch repair may have a role in the formation of unstable nucleotide repeats, perhaps through a disease-specific mechanism.^37- 38 The nexus between the nucleotide repeat expansion pathway and mismatch repair warrants further investigation in this context, because in vitro studies³⁸ and mouse models^39- 41 suggest that abnormal mismatch repair has a major role in mediating the biological effects of MMD-related nucleotide repeat instability. eTable 2 summarizes the observed cancer profile in MMD patients vs HNPCC patients. The occurrence of a similar spectrum of malignancies in both raises the possibility of shared causal pathways.

Our study has several strengths. We used population-based registries, minimizing selection bias and maximizing complete cancer ascertainment. Both MMD and cancer diagnoses were derived from registry-based records rather than self-report, minimizing recall bias. The study included MMD patients identified from both inpatient and outpatient registries, broadening the generalizability of our results. However, the severely affected MMD subset is still likely to be overrepresented because the majority of patients in the study were identified from inpatient hospitalization records.

The remarkable similarity of findings obtained from the Swedish and Danish components of the study provides substantial reassurance that our observations are genuine. The absence of excess screening-related cancers such as breast, cervical, and prostate in our analysis argues against a possible influence of surveillance bias on our results. Most of the excess cancers observed in the present study were lethal cancers that would be diagnosed regardless of whether a person had prior contact with the health care system. Thus, surveillance bias did not appear to influence our results. Furthermore, we found MMD patients with a similarly increased risk of cancer when restricting the analyses to more than 5 years after the first MMD discharge diagnosis, to those with MMD as main diagnosis, and to those with no other diagnoses besides MMD at the first MMD admittance, arguing against the possibility that our results may be confounded by cause of hospitalization or increased surveillance.

Because of the underreporting of nonmelanoma skin cancer in the Swedish Cancer Registry, we were not able to fully evaluate its risk in MMD. However, data available from the Danish registry only suggested a possible excess risk of nonmelanoma skin cancer (SIR, 2.08; 95% CI, 1.2-3.4), an association that needs further confirmation. Of note, our combined data suggested an excess risk of cutaneous melanoma, although it was not statistically significant.

The lack of information regarding known cancer risk factors, eg, smoking, which prevented evaluating them as possible confounders, represents 1 important study limitation. In addition, our data did not permit identifying which specific MMD subtype each participant had, so we could not determine whether the increased cancer risk observed in MMD was common to all patients or confined to a specific subtype. We expect that most of the cases in this study were type 1 MMD, because it is more prevalent, and it was identified and molecularly characterized before type 2 MMD. Furthermore, we did not have data on gene repeat length. Thus, it remains to be evaluated in future studies to what extent gene repeat length modifies the cancer risk. Finally, we acknowledge that the point estimates for some of the cancer sites had wide confidence intervals, which prevent drawing firm conclusions about these sites.

Our study provides quantitative epidemiologic evidence of an increased risk of cancer in MMD patients. The specific cancer patterns observed in our study raise the possibility of a role for aberrant mismatch repair in the etiology of MMD-related cancer. Further research is needed to explore whether the observed associations are similar in both type 1 and type 2 MMD; to determine whether cancer risk correlates with disease severity, repeat length, or both; and to understand the biological mechanisms that might explain the associations we have reported. Our findings have implications for the clinical management of MMD patients, including at a minimum the implementation of appropriate validated routine population-based cancer screening strategies, particularly for colon cancer, and careful assessment of therapy-related risks and benefits. The incidence rates for a number of the excess cancers are relatively low, despite their large relative risks. Screening for these uncommon cancers should not be implemented in the absence of demonstrated clinical utility. The evaluation of persistent central nervous system and abdominopelvic symptoms or dysfunctional uterine bleeding warrants clinical consideration with a higher prior probability of neoplasm, in light of our new findings.