Identification of the cohort of patients with chronic lymphocytic leukemia (CLL) and skin cancer from 2 tertiary care centers during a 20-year period. Included were patients with International Classification of Diseases, Ninth Revision CLL diagnosis code 204.1 and the same skin cancer code (172.0-172.9 and 173.0-173.9) diagnosed at least twice between 1990 and 2010. BMT indicates bone marrow transplantation; MCC, Merkel cell carcinoma; RPDR, Research Patient Data Repository; SCC, squamous cell carcinoma.
eTable 1. Clinical Characteristics of Patients With Primary Merkel Cell Carcinoma and Melanoma
eTable 2. Results of Univariate Analysis to Identify Independent Predictors of Poor Skin Cancer Outcomes
Velez NF, Karia PS, Vartanov AR, Davids MS, Brown JR, Schmults CD. Association of Advanced Leukemic Stage and Skin Cancer Tumor Stage With Poor Skin Cancer Outcomes in Patients With Chronic Lymphocytic Leukemia. JAMA Dermatol. 2014;150(3):280-287. doi:10.1001/jamadermatol.2013.6249
Although it has been well established that patients with chronic lymphocytic leukemia (CLL) have an increased risk of developing skin cancer, few studies have investigated the effect of CLL stage on the risk of poor skin cancer outcomes. The present study of CLL staging assesses outcomes of melanoma, squamous cell carcinoma, and Merkel cell carcinoma in this high-risk population.
To determine if progression of CLL measured by advanced Rai stage (III or IV) is associated with worse skin cancer outcomes.
Design, Setting, and Participants
Twenty-year retrospective study at 2 academic centers in Boston, Massachusetts, of adults with CLL and either melanoma, squamous cell carcinoma, or Merkel cell carcinoma.
Main Outcomes and Measures
Hazard ratios (HRs) for the development of poor skin cancer outcomes (local recurrence, nodal metastasis, distant metastasis, or death from skin cancer).
In total, 133 patients with 377 primary skin cancers and a median follow-up of 41 months were included. Squamous cell carcinoma predominated (92.0%). The risk of death from skin cancer was equivalent to the risk of death from CLL (13.5%). On multivariate analysis, advanced Rai stage (III or IV) at the time of the first skin cancer diagnosis (HR, 4.5; 95% CI, 2.3-8.9) and a high skin cancer tumor (T) stage (HR, 4.9; 95% CI, 2.2-10.8) were associated with poor skin cancer outcomes. Those with both a low skin cancer T stage and a low Rai stage (n = 265) had a low risk (5.3%; 95% CI, 3.2%-8.7%) of poor skin cancer outcomes. Those with a low T stage and a high Rai stage (n = 89) had a significantly higher risk of poor skin cancer outcomes (16.9%; 95% CI, 10.9%-26.0%). The 23 patients with a high T stage had high risks of poor outcomes regardless of CLL status (27.3% if a low Rai stage and 50.0% if a high Rai stage, with wide 95% CIs).
Conclusions and Relevance
In patients with CLL and non–basal cell carcinoma skin cancer, mortality is as high from skin cancer as from CLL. The Rai stage and skin cancer T stage should be considered when risk-stratifying patients with skin cancer. Regular communication between dermatologists and oncologists will help facilitate the identification of patients with CLL who are at high risk of having poor skin cancer outcomes.
Chronic lymphocytic leukemia (CLL) is a low-grade lymphoproliferative disorder of B cells and the most common adult leukemia in the United States and Western Europe.1 The overall incidence rate in the United States is 3 to 5 cases per 100 000, with about 16 000 new cases each year.2 The CLL incidence increases with age and is twice as frequent among men as among women.3 While the disease typically has an indolent course, patients with CLL have a 2-fold increased risk of developing a second malignant tumor, with skin cancer being the most likely.4- 7
The association between skin cancer and CLL is well recognized. Nonmelanoma skin cancer (primarily basal cell carcinoma [BCC] and squamous cell carcinoma [SCC]) is 8-fold higher in incidence among patients with CLL and seems to exhibit a heightened aggressive nature, with greater subclinical skin cancer extension1 and higher rates of recurrence after Mohs surgery.8- 10 Cutaneous melanoma and Merkel cell carcinoma (MCC) also develop with increased frequency among patients with CLL.7,11 In a recent analysis of Surveillance, Epidemiology, and End Results population data, Brewer et al12 found worse overall survival associated with melanoma and with MCC in patients who had a history of CLL compared with patients who did not have a history of CLL.
The staging system by Rai et al,13 developed more than 30 years ago, has been the most consistently used prognostic tool in the CLL population. Based on the presence of cytopenias, lymphadenopathy, and organomegaly (spleen and liver), the Rai staging system stratifies patients into low-risk or intermediate-risk (stages 0 through II) and high-risk (stages III and IV) categories, correlating with survival outcomes.14 In recent years, the development of new biologic tools, including fluorescence in situ hybridization, and the degree of somatic hypermutation of the immunoglobulin heavy-chain variable-region gene, have provided additional prognostic variables that are highly accurate. However, because the Rai staging system has been the most consistently available tool for CLL staging during the past 2 decades, it was used in the present study.
This study sought to identify potential risk factors for poor skin cancer outcomes in patients with CLL to improve dermatologic screening and management of cutaneous malignancies in this population. Considering the potential effect of CLL on the immune system, it was hypothesized that progression of CLL, as determined by advanced Rai stage, would be associated with poor skin cancer outcomes (local recurrence, nodal metastasis, distant metastasis, or death from skin cancer).
After approval from Partners Human Research Committee, all medical records from January 1, 1990, through December 31, 2010, at Brigham and Women’s Hospital and at Massachusetts General Hospital were searched via the Research Patient Data Repository (RPDR) for a diagnosis of CLL and skin cancer. Diagnoses were based on International Classification of Diseases, Ninth Revision (ICD-9) codes.
The RPDR is a centralized clinical data registry that combines information from more than 1.8 million patients seen at the above hospitals. All patient visits, operative notes, laboratory data, and pathology reports documented in the electronic medical record system are automatically included in the RPDR. The RPDR was searched for all patients with at least 2 records of ICD-9 code 204.1, corresponding to CLL, to obtain patients with CLL seen between 1990 and 2010. This group was then searched for all patients who also had at least 2 records of the same skin cancer–related ICD-9 code during the same 20-year period. The skin cancer ICD-9 codes included 172.0 through 172.9 and 173.0 through 173.9, where code 172 refers to malignant melanoma, code 173 refers to BCC or SCC, and codes 0.1 through 0.9 refer to various anatomic locations.
Included cases were those occurring in patients 18 years or older with both a confirmed diagnosis of CLL (based on clinical oncology notes) and non-BCC (confirmed via pathology reports). Basal cell carcinoma was excluded from this study because it has a low risk of poor outcomes, which were the primary focus of this investigation (see the confirmatory subanalysis in the Patient Characteristics subsection of the Results section). For patients who developed skin cancers both before and after CLL diagnosis, only those skin cancers that developed after CLL diagnosis were included in the analysis. Skin cancers that developed after bone marrow transplantation for CLL were excluded because the effect of immunosuppressive medication use and a reconstituted immune system could not be distinguished from the effect of CLL in such cases. Three or more years of follow-up observation from the time of the CLL diagnosis were required for study inclusion.
Medical records were obtained, and the following information was extracted for each primary skin cancer: body location, skin cancer diameter and depth, and dates and methods of all treatments of the primary skin cancer, as well as any recurrence or metastasis and the date and type of recurrence (local, lymph node, or distant organ). Melanoma and MCC were staged according to the American Joint Committee on Cancer’s Cancer Staging Manual, Seventh Edition staging criteria.15,16 Categories T1 and T2 were considered low risk and categories T3 and T4 were considered high risk. Because recent data indicate that American Joint Committee on Cancer staging may not provide optimal prognostic segregation for cutaneous SCC,17 a new alternative SCC staging system was used to stage SCC tumors. In this system, categories T1 and T2a are considered low risk and categories T2b and T3 are considered high risk.17
Patient-specific information was also extracted, including sex, race/ethnicity, skin cancer history, and the presence of noncutaneous malignancies in addition to CLL; for deceased patients, the cause of death was recorded as noted in the medical records. The CLL-related information that was collected included the date of diagnosis, exposure to chemotherapy, and outcomes of CLL at the end of the study, as well as the Rai stage of CLL at the time of the first skin cancer diagnosis. The Rai stage of CLL was obtained from clinical oncology notes or was calculated from an evaluation of laboratory and clinical examinations recorded within 6 months of the first skin cancer diagnosis. Rai staging is defined as follows: stage 0 of absolute lymphocytosis (>5000 B lymphocytes/µL), stage I of absolute lymphocytosis and palpable lymphadenopathy, stage II of absolute lymphocytosis and palpable hepatomegaly or splenomegaly, stage III of absolute lymphocytosis and anemia (hemoglobin level <11.0 g/dL), and stage IV of absolute lymphocytosis and thrombocytopenia (platelet count <100 × 103/µL).13,18 [To convert lymphocyte count to ×109/L, multiply by 0.001; to convert hemoglobin level to grams per liter, multiply by 10.0; and to convert platelet count to ×109/L, multiply by 1.0.]
Patients were considered in remission at the end of the study if they had received chemotherapy for CLL and their blood cell counts had subsequently normalized (≤5000 B lymphocytes/µL, hemoglobin level ≥11.0 g/dL, and platelet count ≥100 × 103/µL) and if there was no evidence of lymphadenopathy after treatment. Skin cancer outcomes of local recurrence, nodal metastasis, distant metastasis, or death from skin cancer were determined to have occurred if a treating physician had documented the occurrence of the outcome in the medical record. If any of these outcomes had occurred, the skin cancer was considered to have a poor outcome.
Statistical analysis and model building were performed on primary skin cancers. Baseline demographic variables and skin cancer clinical data were analyzed using descriptive statistics and frequency tabulation. Survival time for poor outcomes was defined as the time between the primary skin cancer diagnosis and the time of a poor outcome occurrence. For tumors that did not result in a poor outcome, survival time was censored on the date of death or the date of medical record review if the patient was alive. The primary comparison groups were patients with a high Rai stage of CLL at the time of the first skin cancer diagnosis vs those with a low Rai stage. Cox proportional hazards regression modeling was used to determine univariate and multivariate associations of risk factors with the development of any poor skin cancer outcome.
Multivariate models were built through forward stepwise variable addition, followed by backward elimination. In this form of model building, modeling begins with the variable having the largest effect estimate on univariate modeling. Other variables are added based on the next variable with the largest effect estimate and are retained in the model if the P value of the Wald test comparing the smaller model with the larger model was significant at P ≤ .05 or if the P value of the χ2 test comparing the 2 models was borderline (range, >.05 to .10) and the addition of the variable changed the hazard ratio (HR) by at least 10%.
The primary analysis was skin cancer based because each skin cancer was considered to have its own inherent risk of developing the outcomes of interest. The resulting multivariate models were corrected for intrapatient correlation via the robust variance estimate. A patient-level sensitivity analysis was also conducted. For this analysis, 1 skin cancer per patient was selected. If a patient developed a poor outcome, the skin cancer that resulted in the poor outcome was selected. For patients with multiple skin cancers that resulted in multiple poor outcomes, the skin cancer that resulted in the worst poor outcome was selected. If no poor outcomes occurred in a given patient with multiple skin cancers, 1 skin cancer was selected at random for the sensitivity analysis. All statistical tests were performed using a 2-sided 5% type I error rate. Analyses were conducted using commercially available software (STATA Statistical Software, Release 12; StataCorp LP).
The initial RPDR query identified 388 patients with CLL and skin cancer based on ICD-9 code inclusion as detailed in the Methods section (Figure). Seventy-eight patients were excluded because the diagnosis of CLL was ruled out during their diagnostic workup. One-hundred four patients were excluded because of insufficient clinical information or follow-up data or skin cancer occurrence outside of the inclusion variables. Seventy-three patients had only BCC skin cancers. A subanalysis of these 73 patients revealed outcomes similar to those of the general population of patients with BCC (2% risk of local recurrence and no metastases or deaths from disease), confirming that the risk of poor outcomes was low for BCC. These patients were excluded from further analysis because the focus of the study was the risk of poor skin cancer outcomes. Therefore, the final cohort consisted of 133 patients with a confirmed history of CLL and at least 1 primary SCC, MCC, or melanoma. A total of 347 SCCs among 115 patients, 22 melanomas among 19 patients, and 8 MCCs among 8 patients were included. Although most patients had 1 form of skin cancer, 8 patients had an SCC and a melanoma, and 1 patient had an SCC and an MCC. Two patients developed an SCC at a site that had been previously treated with radiation therapy for a prior skin cancer. Both patients had poor outcomes from these SCCs (local recurrence and nodal metastasis in one patient and nodal metastasis and death in the other patient).
There were 94 men and 39 women in the study. The median follow-up periods were 120 months from the time of the CLL diagnosis and 41 months from the time of the skin cancer diagnosis. Patients were divided into those with a high Rai stage of III or IV at the time of the first skin cancer diagnosis (41 of 133 [30.8%]) vs those with a low Rai stage of 0 through II (92 of 133 [69.2%]) (Table 1). Among patients with a CLL diagnosis, their mean (range) ages were 64 (35-84) years for those with a high Rai stage and 64 (44-84) years for those with a low Rai stage (P = .77). Among patients with a skin cancer diagnosis, their mean (range) ages were 73 (44-89) years for those with a high Rai stage and 70 (53-98) years for those with a low Rai stage (P = .10). The cohort was representative of the general population having CLL, being approximately 70% male in both groups and with a mean age at CLL diagnosis of 64 years. Most patients had no history of skin cancer before CLL diagnosis. In total, 27.8% of the cohort developed a secondary noncutaneous malignancy as well (15 prostate, 3 breast, 3 lung, 2 colon, and 14 other), consistent with prior evidence showing a higher rate of solid tumors in patients having CLL.5
Patients with a high Rai stage had a significantly higher risk of overall death (71.1% [27 of 38] vs 30.0% [27 of 90]), and specifically of death from CLL (29.3% vs 6.5%) (P < .001 for both). Most skin cancers were of low tumor (T) stage. Significantly more high-T stage skin cancers were observed in patients with a high Rai stage (12% vs 4%, P = .005), which often required wide local excision (42.2% vs 23.7%, P < .004) and adjuvant radiation therapy (11.0% vs 2.3%, P = .001).
Poor skin cancer outcomes (local recurrence, nodal metastasis, distant metastasis, or death from skin cancer) are given by skin cancer type and are subdivided by low or high T stage and by Rai stage in Table 2. Although poor outcomes were uncommon in patients with a low skin cancer T stage and a low Rai stage SCC (2.8%, 2.0%, 0.8%, and 0.8%, respectively), risks were markedly higher in patients with a high skin cancer T stage and a high Rai stage SCC (25.0%, 37.5%, 12.5% and 12.5%, respectively). There was no difference in χ2 analyses or multivariate models in the risk of poor skin cancer outcomes between those who received chemotherapy vs those who did not receive chemotherapy for CLL. Forty-five of 78 patients who received chemotherapy were exposed to fludarabine phosphate. Similarly, χ2 analysis of those with vs without fludarabine exposure showed no significant difference in the risk of poor skin cancer outcomes (data not shown). Overall, death is as likely to occur from skin cancer (16 of 133 [12.0%]) as from CLL (18 of 133 [13.5%]). Details of the 27 patients with MCC and melanoma are given in eTable 1 in the Supplement.
Results of the univariate analysis to identify independent predictors of poor skin cancer outcomes are summarized in eTable 2 in the Supplement. Significant predictors of poor skin cancer outcomes included the following: 5 or more skin cancers (P = .02), a high skin cancer T stage (P < .001), a skin cancer diameter of at least 2.0 cm (P = .004), treatment with adjuvant radiation therapy (P = .005), and a high Rai stage at the time of the first skin cancer diagnosis (P < .001).
On multivariate analysis, an advanced Rai stage at the time of the first skin cancer diagnosis (HR, 4.5; 95% CI, 2.3-8.9) and a high skin cancer T stage (HR, 4.9; 95% CI, 2.2-10.8) were significantly associated with poor skin cancer outcomes. Patients with both a low skin cancer T stage and a low Rai stage (n = 265) had only a 5.3% (95% CI, 3%-9%) risk of poor skin cancer outcomes. Those with a low skin cancer T stage and a high Rai stage (n = 89) had a significantly higher 16.9% (95% CI, 11%-26%) risk of poor skin cancer outcomes. Patients with high skin cancer T stages were comparatively rare but had very high risks of poor skin cancer outcomes, with wide 95% CIs: these values were 27.3% (95% CI, 10%-57%) for a low Rai stage (n = 11) and 50.0% (95% CI, 25%-75%) for a high Rai stage (n = 12).
In the patient-level sensitivity analysis, the development of 5 or more skin cancers (predominantly SCCs) was found to be a significant risk factor for poor skin cancer outcomes in addition to a high skin cancer T stage and a high Rai stage. Otherwise, the results of the patient-level sensitivity analysis did not differ from the results of the primary skin cancer tumor–level analysis.
To our knowledge, this is the first study in the English-language literature to evaluate the effect of Rai stage on the outcomes of SCC, MCC, and melanoma. Skin cancer is a significant source of morbidity and mortality in patients with CLL. Patients had an equivalent 12% to 13% risk of death from skin cancer as from CLL. Patients having a high Rai stage at the time of their skin cancer diagnosis were 4.5 times more likely to develop poor skin cancer outcomes compared with patients having a low Rai stage, even after adjusting for skin cancer T stage. Furthermore, skin cancer T stage was associated with a 5-fold increased risk of poor skin cancer outcomes when adjusting for CLL stage. Therefore, these 2 factors seem to work in concert to increase the risk of poor skin cancer outcomes. Patients with low-stage skin cancers and early CLL had a low risk (5.3%; 95% CI, 3.2%-8.7%) of poor skin cancer outcomes. The risk was significantly higher (16.9%; 95% CI, 10.9%-26%) in patients with low-risk skin cancers but with more advanced (high Rai stage) CLL. Those with advanced (high T stage) skin cancers had poor outcomes even with early CLL and did still worse if their CLL was advanced, although the number of high-stage skin cancers was too small to precisely estimate these risks. Forty-nine percent of the cohort received chemotherapy during the study period, including a high percentage of patients with a low Rai stage, who may have been treated for extensive adenopathy or signs and symptoms of recurrent disease after prior chemotherapy. Exposure to chemotherapy did not affect skin cancer outcomes in this study.
Squamous cell carcinoma was the most common skin cancer, occurring in 92.0% (347 of 377 primary skin cancers). Patients having SCCs with a low skin cancer T stage herein had outcomes similar to those reported for SCC in the general population (3%-5% risk of local recurrence, 4% risk of metastases, and 1%-2% risk of death).19,20 However, SCCs with a high T stage resulted in much worse outcomes, with a 21% risk of local recurrence, a 29% risk of nodal metastasis, and a 7% risk of death (although deaths from SCC were still rare). The numbers of melanoma and MCC cases were small, so outcomes were difficult to compare with the rates in the general population. However, it is striking that of 133 patients with CLL and skin cancer, 8 patients developed MCC and 19 patients developed melanoma. It is also notable that 5 of 8 patients with MCC had a high Rai stage at the time of the diagnosis. This apparent increased incidence supports the findings of other large investigations that document elevated standardized incidence rates of MCC in patients with CLL.21 Adverse outcomes were common in patients with melanoma (6 of 19 [31.6%] developed nodal metastases and 5 of 19 [26.3%] died of melanoma), which parallels the results of a large Surveillance, Epidemiology, and End Results population study12 that documented a standardized mortality ratio of 2.8 for patients with CLL and melanoma.
Although patients with high-risk (high-T stage) skin cancers had poor outcomes regardless of Rai stage, the Rai stage may be of particular importance in stratifying patients with low T stage skin cancer. Four of 6 patients with a low T stage SCC who died and 2 of 3 patients with a low-T stage melanoma who died had a high Rai stage at the time of their skin cancer diagnosis. Similarly, 7 of 10 patients with SCC nodal metastases from low T stage skin cancers had a high Rai stage at the time of their skin cancer diagnosis.
The reason for the increased incidence of and mortality from skin cancer among patients with CLL is unknown but may be related to an impaired immune system. The CLL B cells produce immunosuppressive factors and downregulate the expression of CD40 ligand on activated T cells, preventing T-cell interactions with antigen-presenting cells.22,23 These immunologic alterations may prevent cancer recognition and antitumor immune activity. Immunosuppressed states (eg, in patients following organ transplantation) have been associated with poor outcomes in SCC, MCC, and melanoma.24- 26 In addition, immunostimulatory therapies have been used effectively to treat skin cancer.27,28 Nevertheless, advanced CLL, as measured by the Rai stage, does not account for all the poor outcomes in low-T stage skin cancers, and other factors are probably involved. One hypothesis is that these patients simply develop more skin cancers; while they are mostly of low stage, the higher number of skin cancers increases their likelihood of an adverse outcome. A trend was observed toward more adverse outcomes in patients who had a history of more than 1 skin cancer (P = .07), although multiple skin cancer formations could indicate more marked immune dysfunction as well. Further studies are needed to investigate the interplay between immunity and skin cancer outcomes.
This study has several limitations. The cohort was limited to patients diagnosed as having CLL and skin cancer at 2 academic tertiary care centers, where the patient population may differ from that seen in the community. However, the skin cancer and CLL staging systems used in this study should enhance the generalizability to patients with similarly staged disease. The ICD-9 coding of CLL was used to identify the study cohort, which has inherent limitations because of the inaccuracy of coding; however, such error was minimized by requiring at least 2 visits with the CLL code for study inclusion and verification of the diagnosis via medical record review. The study population was predominantly of white race/ethnicity because both skin cancer and CLL are much more common in this group, so conclusions about skin cancer incidence and outcomes in nonwhite or Hispanic populations cannot be drawn. Finally, although several new prognostic tools for CLL have been developed during the past few years (eg, fluorescence in situ hybridization and immunoglobulin heavy-chain variable-region gene mutation status), these data were unavailable in most cases, and their usefulness could not be evaluated. Inclusion of these measurements in future studies may help to further risk-stratify patients having CLL with regard to both CLL and skin cancer prognosis.
This study confirms the significant morbidity and mortality of skin cancer in patients with CLL and reinforces the need for these patients to undergo routine dermatologic surveillance. Even low-T stage skin cancers carry an elevated risk of local recurrence, metastases, and death in patients with advanced CLL. Understanding patients’ CLL status, through the Rai stage and close communication with their hematologist or oncologist, is important in predicting which patients may have poor outcomes from their skin cancer. Further studies are needed to determine the optimal screening interval and skin cancer treatment modalities for patients with CLL.
Accepted for Publication: June 17, 2013.
Corresponding Author: Chrysalyne D. Schmults, MD, MSCE, Department of Dermatology, Brigham and Women’s Hospital, Harvard Medical School, 1153 Centre St, Boston, MA 02130 (firstname.lastname@example.org).
Published Online: January 15, 2014. doi:10.1001/jamadermatol.2013.6249.
Author Contributions: Drs Velez and Schmults had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the analysis.
Study concept and design: Velez, Davids, Brown, Schmults.
Acquisition of data: Velez, Karia, Vartanov, Brown.
Analysis and interpretation of data: Velez, Karia, Schmults.
Drafting of the manuscript: Velez, Karia, Schmults.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Velez, Karia, Schmults.
Administrative, technical, or material support: Karia.
Study supervision: Schmults.
Conflict of Interest Disclosures: None reported.
Additional Contributions: Ye Guo, MSCS, provided assistance with our database. Victor Neel, MD, PhD, served as our Massachusetts General Hospital collaborator and provided access to data.