Association between Lithium Use and Melanoma Risk and Mortality: A Population-Based Study

Accepted Manuscript Association between lithium use and melanoma risk and mortality: A populationbased study Maryam M. Asgari, MD MPH, Andy J. Chien, MD, PhD, Ai Lin Tsai, Bruce Fireman, Charles P. Quesenberry, Jr., PhD PII:

S0022-202X(17)31644-5

DOI:

10.1016/j.jid.2017.06.002

Reference:

JID 902

To appear in:

The Journal of Investigative Dermatology

Received Date: 23 January 2017 Revised Date:

16 May 2017

Accepted Date: 6 June 2017

Please cite this article as: Asgari MM, Chien AJ, Tsai AL, Fireman B, Quesenberry Jr. CP, Association between lithium use and melanoma risk and mortality: A population-based study, The Journal of Investigative Dermatology (2017), doi: 10.1016/j.jid.2017.06.002. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Association between lithium use and melanoma risk and mortality: A population-based study Maryam M. Asgari, MD MPH,1,2 Andy J. Chien, MD, PhD3,4 Ai Lin Tsai,1 Bruce Fireman1,

1

RI PT

Charles P. Quesenberry Jr. PhD,1

Department of Dermatology, Massachusetts General Hospital, and Department of Population

Medicine, Harvard Medical School, Boston, MA.

Division of Research, Kaiser Permanente Northern California, Oakland, California

3

Division of Dermatology, University of Washington Medical Center, Seattle, Washington

4

The Group Health Research Institute, Seattle Washington

M AN U

SC

2

Corresponding Author:

Maryam M. Asgari, MD MPH Department of Dermatology

TE D

Massachusetts General Hospital 50 Staniford Street, Suite 230A Boston, Massachusetts 02114

EP

(617) 643-6812 (phone)

AC C

(617)726-9133 (fax) [email protected]

Short Title: Association between lithium use and melanoma Location of the work: Oakland, CA, United States of America

1

ACCEPTED MANUSCRIPT

ABSTRACT Laboratory studies show lithium, an activator of the Wnt/ß-catenin signaling pathway, slows

RI PT

melanoma progression, but no published epidemiologic studies have explored this association. We conducted a retrospective cohort study of adult white Kaiser Permanente Northern California members (n=2,213,848) from 1997-2012 to examine the association between lithium use and melanoma risk. Lithium exposure (n=11,317) was assessed from pharmacy databases, serum

SC

lithium levels were obtained from electronic laboratory databases, and incident cutaneous

M AN U

melanomas (n= 14,056) were identified from an established cancer registry. In addition to examining melanoma incidence, melanoma hazard ratios (HRs) and 95% confidence intervals (CIs) for lithium exposure were estimated using Cox proportional hazards models, adjusted for potential confounders. Melanoma incidence per 100,000 person-years among lithium-exposed individuals was 67.4, compared to 92.5 in unexposed individuals (p = 0.027). Lithium-exposed

TE D

individuals had a 32% lower risk of melanoma risk (HR=0.68; CI 0.51-0.90) in unadjusted analysis, but the estimate was attenuated and non-significant in adjusted analysis (aHR=0.77; CI 0.58-1.02). No lithium-exposed individuals presented with thick (> 4 mm) or advanced-stage

EP

melanoma at diagnosis. Among melanoma cases, lithium-exposed individuals were less likely to suffer melanoma-associated mortality (rate = 4.68/1,000 person years) as compared to the

AC C

unexposed (rate = 7.21/1,000 person-years). Our findings suggest lithium may reduce melanoma risk and associated mortality.

2

ACCEPTED MANUSCRIPT

INTRODUCTION Melanoma accounts for less than five percent of all skin cancers yet is responsible for

RI PT

80% of all skin cancer deaths, and its incidence is increasing (Linos et al., 2009, Tsao et al., 2004). Despite recent therapeutic advances, the prognosis for patients with metastatic disease remains poor. Developing a clear understanding of the pathways involved in melanoma

formation and progression may lead to improved melanoma prevention and treatment strategies.

SC

Technologies such as microarray transcriptional profiling have uncovered changes in several

M AN U

different signal transduction pathways in melanomas, including alterations in the Wnt signaling pathway (Weeraratna, 2005). Wnt genes encode a family of secreted glycoproteins that activate cellular signaling pathways to control cell differentiation, proliferation, and motility(Chien et al., 2009a). The most extensively studied Wnt pathway is the Wnt/β-catenin pathway. Wnt ligands bind to membrane-bound receptors resulting in the cytoplasmic accumulation and nuclear

TE D

translocation of β-catenin. The importance of Wnt/ ß-catenin signaling in melanoma was suggested by initial observations that ß-catenin is more prevalent in benign lesions and early melanomas than metastatic lesions (Omholt et al., 2001, Pollock and Hayward, 2002,

EP

Reifenberger et al., 2002, Rubinfeld et al., 1997, Worm et al., 2004). Absence of both nuclear and cytoplasmic ß-catenin in melanomas is associated with a poorer prognosis(Bachmann et al.,

AC C

2005, Kageshita et al., 2001, Maelandsmo et al., 2003), a conclusion confirmed in patients by automated quantification of immunofluorescence (Chien et al., 2009b). Furthermore, transcriptional profiling of melanoma cell lines has suggested that Wnt/ß-catenin signaling regulates a transcriptional signature predictive of less aggressive melanoma (Hoek et al., 2006). In mouse models, forced activation of Wnt/β-catenin signaling leads to decreased growth of syngenic and xenografted melanoma tumors (Biechele et al., 2012, Chienet al., 2009b). These

3

ACCEPTED MANUSCRIPT

findings have led to the hypothesis that active Wnt/ß-catenin signaling may reflect or promote a particular melanoma phenotype that is less aggressive and portends a better prognosis, and could

RI PT

provide a therapeutic target for melanoma. Lithium, a medication used for treating mood disorders, activates Wnt/ß-catenin signaling by inhibiting glycogen synthase kinase-3, the key enzymatic regulator of β-catenin degradation in cells. Lithium exposure in a mouse melanoma model has been shown to inhibit melanoma

SC

cell proliferation (Ballin et al., 1983, Penso and Beitner, 2003). However, no published

M AN U

epidemiologic studies have examined the association of melanoma risk with lithium exposure. In this study, we examined the association between lithium exposure and incident melanoma risk and melanoma-associated mortality among over two million health plan members of a large integrated health care delivery system.

TE D

RESULTS

Of the 2,213,848 cohort members, 11,317 subjects received lithium therapy, and 14,056 were diagnosed with incident melanoma. The average number of years of follow-up per cohort

EP

member was 6.9 ± 5.4 SD years. The melanoma incidence per 100,000 person-years in the exposed vs. unexposed group was 67.4 and 92.5 respectively (p=0.027). Cohort characteristics

AC C

by case status are outlined in table 1. Exposed individuals tended to be younger (mean age 41.0 ± 14.8 SD vs. 42.5 ± 18.2 for unexposed, p < 0.001), and female (61.1% vs. 54.4%, p<0.001). The majority of exposed cohort members entered the cohort in earlier time-periods as compared to the unexposed cohort members. In an unadjusted analysis (Table 2), lithium use was associated with a 32% reduction in risk of incident melanoma (HR = 0.68, 95% CI 0.51-0.90). However, the aHR was attenuated and was not statistically significant (aHR 0.77, 95% CI 0.58-1.02). Further analysis revealed age 4

ACCEPTED MANUSCRIPT

adjustment resulted in the most attenuation of the HR for ever use (data not shown). In examining cumulative duration of exposure, the point estimate for risk dropped with increasing durations of exposure (adjusted HR = 0.81 for short-term exposure, 0.78 for moderate-term, and

RI PT

0.65 for long-term exposure as compared to unexposed individuals), and the test for trend

bordered on being statistically significant (p = 0.06). There was no association with current dose of lithium use, or serum lithium levels.

SC

We examined the impact of lithium exposure on tumor characteristics at disease

M AN U

presentation (Table 3). Lithium-exposed individuals did not present with thick tumors (defined as Breslow depth > 4 mm) at diagnosis. Moreover, all lithium exposed individuals were diagnosed with local disease and none had melanoma which involved regional or distal disease at diagnosis. In examining mortality among exposed and unexposed melanoma cases, exposed subjects were less likely to have melanoma specific mortality (melanoma mortality rate of

TE D

4.68/1,000 person-years), as compared to unexposed individuals (melanoma mortality rate of 7.21/1,000 person-years. Given the small sample size of lithium exposed subjects who developed melanoma-associated mortality (n=1), test of statistical significance were not

AC C

DISCUSSION

EP

performed because the tests were deemed underpowered.

Data from this large cohort of over 2 million KPNC Health Plan members suggests that

lithium exposure may reduce melanoma incidence, lower risk of very thick tumors, and lower risk of melanoma-associated mortality. The protective effect of lithium exposure on melanoma progression has been suggested in preclinical studies involving mice, and may be associated with lithium’s activation of the Wnt/ß-catenin signaling pathway, which has been associated with melanoma progression. Whereas in unadjusted models, the association of lithium exposure on

5

ACCEPTED MANUSCRIPT

incident melanomas appeared protective, the association was no longer significant in adjusted models, although the point estimates suggested a protective effect. There was no impact of current dose on incident melanoma risk, although there was a suggestion of a protective effect

RI PT

with cumulative duration with a monotonic decrease in risk across increasing duration

categories, though the p value for trend did not reach statistical significance. No prior studies have examined the association of lithium exposure on melanoma in human subjects, although

SC

evidence to suggest that lithium inhibits melanoma tumor growth in mouse models was first reported over thirty years ago (Ballinet al., 1983). More recent findings suggest that

M AN U

lithium alters several immune- and signal transduction-related cellular functions (Anand et al, 2016). Recent findings show that activation of Wnt/β-catenin signaling predicts a poorer clinical response to targeted BRAF inhibitors (Chien et al., 2014) suggesting that lithium may impact clinical response.

TE D

This study utilizes one of the largest cohorts with detailed pharmacoepidemiology data and outcome data from a SEER-reporting cancer registry to explore the association between lithium and melanoma risk. Yet this study also has several limitations. We relied on information

EP

from filled prescriptions rather than self-reported use to characterize lithium exposure, though we supplemented our pharmacy data with a biomarker, namely, serum lithium level

AC C

measurements. Cohort members may have filled their prescriptions for lithium outside a health plan pharmacy, although use of non-Kaiser pharmacies by health plan members is estimated to be rare (Moffet et al., 2009). Unmeasured confounding, such as concomitant use of photosensitizing medication by unexposed individuals or differential sun exposure patterns by exposure status, may have influenced our findings. Information about sun exposure history and other melanoma risk factors was unavailable, although it is unlikely that melanoma risk factors

6

ACCEPTED MANUSCRIPT

would be much different in those exposed to lithium vs. unexposed individuals. Nevertheless, as an observational study of clinical practice, we cannot completely exclude residual confounding, such as sun exposure behaviors, or other treatment selection biases as an alternative explanation

RI PT

of our findings. The risk of confounding by indication is low because bipolar disease, which is the most common indication for lithium use, has not been associated with melanoma risk. The difference in melanoma incidence between the exposed and unexposed groups is largely driven

SC

largely by differences in in-situ and very thick tumors, and given the possibility of diagnostic drift, we included calendar time into our adjusted models to help mitigate against time trends.

M AN U

Given the low melanoma rates among the exposed group, our study was not powered to examine stage or progression as independent outcomes. Finally, as our study was conducted among insured adults in Northern California, our results may not be generalized to uninsured persons and other health care or geographic settings, although a recently published study comparing the

TE D

characteristics of KPNC melanomas to those in from State and SEER program show that our melanoma patients are comparable in patient, tumor and care characteristics to state and national cases (Asgari et al., 2014).

EP

In conclusion, findings from this large cohort study revealed that lithium-exposed individuals had a reduced incidence of melanoma, did not develop very thick tumors (> 4 mm

AC C

Breslow depth) or extensive disease at presentation, and had decreased melanoma-specific mortality compared to unexposed individuals suggesting a possible role for lithium in altering melanoma risk. Our conclusions provide evidence that lithium, a relatively inexpensive and readily available drug, warrants further study in melanoma. MATERIALS & METHODS Study Setting

7

ACCEPTED MANUSCRIPT

Kaiser Permanente Northern California (KPNC) is a pre-paid healthcare delivery system providing comprehensive health care and pharmaceutical benefits to a large and diverse community-based population that grew from 2.9 to 3.3 million persons in northern California

RI PT

during the study period. The membership represents 33% of the insured population and 28% of the total population in the service area, and is similar to the general population in Northern

California with regard to socio-demographic and health characteristics (NP, 2012). KPNC’s

SC

computerized record system, which has been fully operational since 1996, contains

administrative and clinical electronic databases linked by a unique patient medical record

M AN U

number, providing a detailed and comprehensive record of members’ demographic characteristics, clinical status, laboratory, pathology and radiology results, healthcare and pharmacy utilization, and benefits status. The pathology database contains information on all pathology specimens received for examination including date and type of tissue, tumor location,

TE D

tumor subtype, and gross and microscopic diagnoses in text format. Data from electronic medical records are used as the basis for the KPNC Cancer Registry which collects, codes, and reports all cancer data (except non-melanoma skin cancer and in-situ cervical cancer) to the

EP

Surveillance, Epidemiology, and End Results (SEER) Program. Numerous quality control processes and audits help verify data accuracy and reporting completeness under standards set

AC C

forth by the California Cancer Registry and the SEER program. Study Population

Self-identified non-Hispanic white KPNC members were eligible to enter the cohort at

the earliest date of meeting the age criteria (≥ 18 years) during January 1, 1997 through June 30, 2012, and must have had pharmacy coverage to enter the cohort. Subjects were excluded if they had any SEER reportable melanoma diagnoses prior to cohort entry (n=5,856), or had missing or

8

ACCEPTED MANUSCRIPT

other gender (n=208), leaving 2,213,848 cohort members. Censoring occurred at the earliest of: 1) the first day of any KPNC membership gap > 3 months, 2) the first day in any KPNC pharmacy benefits gap > 3 months, 3) date of death, 4) incident melanoma diagnosis date, or (5)

Foundation Research Institute Institutional Review Board.

SC

Case Ascertainment

RI PT

the end of the observation period on June 30, 2012. This study was approved by the Kaiser

All cohort members diagnosed with an incident cutaneous melanoma between January 1,

Assessment of Lithium Exposure

M AN U

1997 and June 30, 2012 as reported by the KPNC Cancer Registry were assigned case status.

Pharmacy records: Exposure to lithium therapy was ascertained for each cohort member using information found in automated KPNC pharmacy databases on filled prescriptions during the

TE D

observation period. Cohort members were defined as exposed to lithium on the first date a lithium prescription was filled, assuming at least two or more lithium prescriptions were filled during the observation period. If the subject filled only a single prescription for lithium with no

EP

further refills during the observation period (n = 3,084), the subjects was categorized as unexposed. Using the cohort entry and censoring dates, periods of exposed and unexposed

AC C

person-time were established in the cohort. We assigned periods of continuous lithium exposure during follow-up for all cohort members using data from serial prescriptions based on each prescription’s fill date and the number of days supplied. We considered an individual as continuously receiving lithium if the gap between the expected end date of a prescription and fill date of the next prescription was ≤30 days. If the gap was >30 days, the individual was considered not taking lithium until the fill date of the next prescription. Cumulative lithium exposure was categorized as: short-term (< 2 year), moderate-term (2-5 years), and long-term 9

ACCEPTED MANUSCRIPT

(>5 years). Dose of lithium exposure was categorized as: low (≤300 mg/day), moderate (>300600 mg/day), and high (>600 mg/day).

RI PT

Biomarker: We also examined serum lithium levels as an alternate measure of exposure. Serum lithium was treated as both a continuous variable, and also categorized, based on clinical

recommendations for target serum lithium (Wijeratne and Draper, 2011), measured in mmol/L as

SC

follows: low: < 0.5, moderate: ≥ 0.5– <0.8, and high: ≥ 0.8. Assessment of Covariates

M AN U

Age at cohort entry, and sex, and calendar year of cohort entry, and healthcare utilization, including all ambulatory and emergency department encounter types, were ascertained from administrative databases. Outpatient healthcare utilization was chosen to account for possible detection bias in the Kaiser Permanente setting, given that skin lesions can

TE D

be detected at any point of care encounter, including visits to primary care providers. We adjusted for calendar year to address potential bias from secular trends in cancer incidence and prescribing patterns. Mortality, both overall, and disease-associated, was ascertained from

EP

clinical mortality information from the KPNC membership data, supplemented by California mortality files, which contain death certificate data including cause of death, and by the United

AC C

States Social Security Administration Death Master Database. These linkage files are updated annually approximately one year after the close of the calendar year in which deaths occur. Statistical Analysis

We used Cox proportional hazards regression models to estimate unadjusted melanoma hazard ratios (HRs) and HRs adjusted for age, gender, and year of cohort entry (aHRs). Fixed covariates included age, sex and calendar year of cohort entry. Time varying covariates included

10

ACCEPTED MANUSCRIPT

the primary lithium exposure metrics, which were measured from cohort entry and captured during follow-up. Four different measures of lithium exposure were examined in our analysis: 1) ever/never use; 2) cumulative duration; 3) current dose; and 4) most recent serum lithium

RI PT

level. We investigated cumulative duration of exposure by categories: unexposed, short-term (1 day-<2 year), moderate (> 2-5 years), and long-term (>5 years) use, with never use as the

referent category. Next, we examined dose (low, moderate and high) among current users

SC

defined as the most recent dose filled by the subject. Finally, we examined most recent serum lithium level among ever users as an additional measure of exposure, treated as a categorical

CONFLICT OF INTEREST

M AN U

variable with never use as the referent category.

Dr. Maryam Asgari and Dr. Charles Quesenberry have each served as an investigator for studies funded by Valeant Pharmaceuticals and Pfizer Inc, but this association has not influenced their

TE D

work on this paper. The remaining authors state no conflict of interest. ACKNOWLEDGEMENTS

This research was supported by the National Cancer Institute (R01CA166672). The sponsors had

EP

no role in the design and conduct of the study; in the collection, analysis, and interpretation of

AC C

data; or the preparation, review or approval of the manuscript.

11

ACCEPTED MANUSCRIPT

REFERENCES:

AC C

EP

TE D

M AN U

SC

RI PT

Asgari MM, Eide MJ, Warton M, Fletcher SW. Comparing characteristics of melanoma cases arising in health maintenance organizations with state and national registries. Melanoma Res. 2014;24(4):381-7. Anand A, McClintick JN, Murrell J, Karne H, Nurnberger JI, Edenberg HJ. Effects of Lithium Monotherapy for Bipolar Disorder on Gene Expression in Peripheral Lymphocytes. Mol Neuropsychiatry. 2016;2(3):115-123 Bachmann IM, Straume O, Puntervoll HE, Kalvenes MB, Akslen LA. Importance of P-cadherin, beta-catenin, and Wnt5a/frizzled for progression of melanocytic tumors and prognosis in cutaneous melanoma. Clin Cancer Res 2005;11(24 Pt 1):8606-14. Ballin A, Aladjem M, Banyash M, Boichis H, Barzilay Z, Gal R, et al. The effect of lithium chloride on tumour appearance and survival of melanoma-bearing mice. Br J Cancer 1983;48(1):83-7. Biechele TL, Kulikauskas RM, Toroni RA, Lucero OM, Swift RD, James RG, et al. Wnt/betacatenin signaling and AXIN1 regulate apoptosis triggered by inhibition of the mutant kinase BRAFV600E in human melanoma. Sci Signal 2012;5(206):ra3. Chien AJ, Conrad WH, Moon RT. A Wnt survival guide: from flies to human disease. J Invest Dermatol 2009a;129(7):1614-27. Chien AJ, Haydu LE, Biechele TL, Kulikauskas RM, Rizos H, Kefford RF, et al. Targeted BRAF inhibition impacts survival in melanoma patients with high levels of Wnt/betacatenin signaling. PLoS One 2014;9(4):e94748. Chien AJ, Moore EC, Lonsdorf AS, Kulikauskas RM, Rothberg BG, Berger AJ, et al. Activated Wnt/beta-catenin signaling in melanoma is associated with decreased proliferation in patient tumors and a murine melanoma model. Proc Natl Acad Sci U S A 2009b;106(4):1193-8. Hoek KS, Schlegel NC, Brafford P, Sucker A, Ugurel S, Kumar R, et al. Metastatic potential of melanomas defined by specific gene expression profiles with no BRAF signature. Pigment Cell Res 2006;19(4):290-302. Kageshita T, Hamby CV, Ishihara T, Matsumoto K, Saida T, Ono T. Loss of beta-catenin expression associated with disease progression in malignant melanoma. Br J Dermatol 2001;145(2):210-6. Linos E, Swetter SM, Cockburn MG, Colditz GA, Clarke CA. Increasing burden of melanoma in the United States. J Invest Dermatol 2009;129(7):1666-74. Maelandsmo GM, Holm R, Nesland JM, Fodstad O, Florenes VA. Reduced beta-catenin expression in the cytoplasm of advanced-stage superficial spreading malignant melanoma. Clin Cancer Res 2003;9(9):3383-8. Moffet HH, Adler N, Schillinger D, Ahmed AT, Laraia B, Selby JV, et al. Cohort Profile: The Diabetes Study of Northern California (DISTANCE)--objectives and design of a survey follow-up study of social health disparities in a managed care population. Int J Epidemiol. 2009 ;38(1):38-47. NP G. Similarity of the Adult Kaiser Permanente Membership in Northern California to the Insured and General Population in Northern California: Statistics from the 2009 California Health Interview Survey. Internal Division of Research report Oakland, CA: Kaiser Permanente Division of ResearchKaiser Permanente Division of Research2012.

12

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

Omholt K, Platz A, Ringborg U, Hansson J. Cytoplasmic and nuclear accumulation of betacatenin is rarely caused by CTNNB1 exon 3 mutations in cutaneous malignant melanoma. Int J Cancer 2001;92(6):839-42. Penso J, Beitner R. Lithium detaches hexokinase from mitochondria and inhibits proliferation of B16 melanoma cells. Mol Genet Metab 2003;78(1):74-8. Pollock PM, Hayward N. Mutations in exon 3 of the beta-catenin gene are rare in melanoma cell lines. Melanoma Res 2002;12(2):183-6. Reifenberger J, Knobbe CB, Wolter M, Blaschke B, Schulte KW, Pietsch T, et al. Molecular genetic analysis of malignant melanomas for aberrations of the WNT signaling pathway genes CTNNB1, APC, ICAT and BTRC. Int J Cancer 2002;100(5):549-56. Rubinfeld B, Robbins P, El-Gamil M, Albert I, Porfiri E, Polakis P. Stabilization of beta-catenin by genetic defects in melanoma cell lines. Science 1997;275(5307):1790-2. Tsao H, Atkins MB, Sober AJ. Management of cutaneous melanoma. N Engl J Med 2004;351(10):998-1012. Weeraratna AT. A Wnt-er wonderland--the complexity of Wnt signaling in melanoma. Cancer Metastasis Rev 2005;24(2):237-50. Wijeratne C, Draper B. Reformulation of current recommendations for target serum lithium concentration according to clinical indication, age and physical comorbidity. Aust N Z J Psychiatry 2011;45(12):1026-32. Worm J, Christensen C, Gronbaek K, Tulchinsky E, Guldberg P. Genetic and epigenetic alterations of the APC gene in malignant melanoma. Oncogene 2004;23(30):5215-26.

13

ACCEPTED MANUSCRIPT

Table 1: Cohort characteristics by lithium exposure status Jan 1997- June 2012 (n=2,213,848) Characteristic

Unexposed (n=2,202,528)

Ever Exposed (n=11,317)

41 (18-111)

41 (18-92)

41.0 (14.8) 8,247 (72.9%)4 2,400 (21.2%)5 670 (5.9%)6 4,408 (38.9%) 6,909 (61.1%) 6,336 (56.0%) 2,493 (22.0%) 1,633 (14.4%) 855 (7.6%)

TE D

M AN U

Mean + SD 42.5 (18.2) <50 1,442,031 (65.5%)1 50-65 481,073 (21.8%)2 65+ 279,424 (12.7%)3 Sex Male 1,004,840 (45.6%) Female 1,197,688 (54.4%) Year of cohort entry 1997 1,015,155 (46.1%) 1998-2002 486,387 (22.1%) 2003-2007 371,588 (16.9%) 2008-2012 329,398 (15.0%) 7 Healthcare Utilization 170 (1.5%) None 3,066 (27.1%) Low (< 10 visit per year) 8,081 (71.4%) High (>=10 visit per year) Melanoma rate8 92.5 (per 100,000 person-years) 1 Mean length of follow-up 6.19 (± 5.27) person-years 2 Mean length of follow-up 8.31 (± 5.60) person-years

SC

Median (range)

RI PT

Age at cohort entry, y

3

195,657 (8.9%) 1,526,392 (69.0%) 489,647 (22.1%) 67.4

p= 0.027

AC C

8

EP

Mean length of follow-up 7.78 (± 5.26) person-years Mean length of follow-up 5.89 (± 4.84) person-years 5 Mean length of follow-up 7.06 (± 5.04) person-years 6 Mean length of follow-up 7.06 (± 4.70) person-years 7 Measured as number of ambulatory care and emergency care visits per year. Note that utilization is time-dependent, and most people contribute some time in the unexposed category before they move into the exposed. 4

14

ACCEPTED MANUSCRIPT

Table 2: Lithium use and melanoma risk (n=2,213,848)

Serum Lithium Never use Exposed, no serum lithium Low ≤ 0.5 mmol/L Moderate > 0.5 - <0.8 mmol/L High ≥ 0.8 mmol/L

1.00 (referent) 0.77 (0.58 – 1.02)

1.00 (referent) 0.68 (0.47-0.99) 0.73 (0.40-1.32) 0.62 (0.33-1.20) 0.0148

1.00 (referent) 0.81(0.56-1.17) 0.78(0.43-1.41) 0.65(0.34-1.24) 0.06

1.00 (referent) 0.78 (0.43-1.42) 0.62 (0.30-1.30) 1.08 (0.41-2.89)

1.00 (referent) 0.84 (0.46-1.51) 0.66 (0.31-1.38) 1.14 (0.43-3.03)

1.0 (referent) 0.69 (0.33-1.44) 0.64 (0.39-1.05) 0.74 (0.42-1.30) 0.67 (0.39-1.51)

1.0 (referent) 0.81 (0.38-1.69) 0.72 (0.42-1.24) 0.73 (0.45-1.20) 0.84 (0.48-1.48)

EP

TE D

Adjusted for age, gender, year of cohort entry, and healthcare utilization.

AC C

1

1.00 (referent) 0.68 (0.51-0.90)

RI PT

Cumulative duration Never use (n=2,211,695) Short-term (< 2 year) ) (N=11,317) Moderate (2-5 years) (N=4,314) Long-term (>5 years) (N=2,080) Ptrend Current Dose Not current user Low (≤300 mg/day) Moderate (>300-600 mg/day) High (>600 mg/day)

HR Adjusted1 (95% CI)

SC

Exposure Never use (n =2,211,695) Ever use (n=11,317)

HR unadjusted (95% CI)

M AN U

Lithium exposure

ACCEPTED MANUSCRIPT

Table 3: Tumor characteristics and melanoma-associated mortality among exposed and unexposed melanoma cases (n=14,056)

Staging Local Regional Distant Unknown Mortality (per 1,000 person-years) Overall1 Melanoma-specific2

Ever Exposed (n = 48) n (%)

5,855 (41.8%) 5,469 (39.0%) 1,254 (9.0%) 623 (4.5%) 388 (2.8%) 419 (3.0%)

14 (29.2%) 22 (45.8%) 7 (14.6%) 3 (6.3%) 0 (0.0%) 2 (4.2%)

48 (100.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%)

M AN U

12,974 (92.6%) 615 (4.4%) 263 (1.9%) 156 (1.1%)

RI PT

Breslow Depth 0 (in situ) >0 to < 1 ≥1 to 2 >2 to 4 >4 Unknown

Unexposed (n =14,008) n (%)

SC

Melanoma Characteristics

29.19 7.21

1

32.78 4.68

AC C

EP

TE D

Overall n=2,165 deaths among those diagnosed with melanoma with 74,147 years of personyears of follow up. Among unexposed, n=2,158 deaths among 73,933 years of follow-up. Among exposed group, n= 7 deaths per 214 person-years of follow-up. 2 Overall n=534 deaths among those diagnosed with melanoma with 74,147 years of personyears of follow up. Among unexposed, n=533 deaths among 73,933 years of follow-up. Among exposed group, n= 1 deaths per 214 person-years of follow-up.