(Psoralen) and UV-A Radiation (PUVA) a Meta-Analysis

STUDY Risk of Squamous Cell Carcinoma and Methoxsalen (Psoralen) and UV-A Radiation (PUVA) A Meta-analysis

Robert S. Stern, MD; Elissa J. Lunder, MD

Objective: To assess the risk of squamous cell carci- high-dose PUVA (Ͼ200 treatments or 2000 J/cm2). Ex- noma (SCC) and the relation of dose to risk among groups act methods were used to calculate the incidence rate of patients with psoriasis exposed to psoralen–UV-A ratios. (PUVA). Data Synthesis: In addition to our study, we identi- Data Sources: Four electronic databases were searched fied and reviewed 8 other studies. Overall, the inci- from 1984 to 1998. dence among patients exposed to high-dose PUVA was 14-fold higher than among patients with low-dose ex- Study Selection: In addition to the PUVA Follow-up posure (95% confidence interval, 8.3-24.1); a greater dose- Study, we included all English-language studies from the dependent increase in risk than that observed in the PUVA United States and Europe with at least 150 patients en- Follow-up Study. rolled, who were followed up for at least 5 years as identified from our bibliographic search. Conclusion: Although the incidence of SCC reported among groups of PUVA-treated patients followed up for Data Extraction: A custom-designed questionnaire was at least 5 years varies greatly, compared with the risk in used to extract data from each of the articles. For each low-dose patients, long-term high-dose exposure to PUVA study, if possible, we determined the incidence of basal was consistently observed to significantly increase the risk cell carcinomas and SCCs and the incidence rate ratio of SCC in all studies reviewed. of SCC among patients exposed to low-dose (we defined as Ͻ100 treatments or 1000 J/cm2) compared with Arch Dermatol. 1998;134:1582-1585

RAL PSORALEN–UV-A ra- posure to PUVA but whose overall expo- diation (PUVA) is an ef- sure to PUVA has been lower.6-13 fective and widely used The published studies of the occur- treatment for psoriasis. rence of skin cancer among patients ex- First used more than 20 posed to PUVA vary substantially in their years ago, PUVA is recognized as one of the methods, population studied, and results. O 1 most valuable treatments for psoriasis. Pso- We undertook a meta-analysis of 8 stud- ralen–UV-A is, however, mutagenic and is ies published in the English language from considered a standard for photocarcino- groups other than the PUVA Follow-up genesis studies in animals.2 In 1975, the first Study5 which provided information on the 1380 patients to be treated with PUVA for occurrence of nonmelanoma skin cancer (in psoriasis in the United States enrolled in a a group of at least 150 patients primarily long-term prospective study of PUVA with psoriasis followed up for at least 5 therapy (The PUVA Follow-up Study).3-5 years.6-13 Based on the data available from Reports from this multicenter study have these articles, we calculated the incidence demonstrated that in this population long- of SCC, the relation of incidence to level term exposure to PUVA increases the risk of exposure to PUVA, and the ratio of SCC of squamous cell carcinoma (SCC) and that to basal cell carcinoma (BCC) overall and this risk increases with greater exposure to according to level of exposure to PUVA. To PUVA.4,5 These findings have been contro- assess the extent to which these findings From the Department of versial. Other groups in the United States concerning SCC risks and PUVA therapy Dermatology, Beth Israel and Europe have published their findings are consistent, we compared the indi- Deaconess Medical Center, based on observations of groups of pa- vidual and aggregate findings of these Harvard Medical School, tients with psoriasis (and in some cases 8 studies6-13 with the published findings Boston, Mass. other diseases) who have had long-term ex- from the PUVA Follow-up Study.5

ARCH DERMATOL / VOL 134, DEC 1998 1582

©1998 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 10/02/2021 low-up Study, studies reported average age at enrollment.6,9-11 These ages were not substantially different from MATERIALS AND METHODS each other or the PUVA Follow-up Study5 (average age range, 43-50 years). Males outnumbered females in 5 of Using the MEDLINE, Healthstar, Aidsline, and Can- the 7 studies with these data provided, including the PUVA Follow-up Study.5 The PUVA Follow-up Study5 and the cerlit databases, we attempted to ascertain all stud- 7 ies published reporting on the nonmelanoma skin can- study by Chuang et al were based in the United States. cer experience for patients primarily with psoriasis The remaining 7 studies were of European popula- exposed to oral PUVA with an average follow-up of tions.6,8-13 Nearly all patients studied were of European at least 5 years. We initially identified 49 studies. For ancestry. cohorts that had been subject to multiple analyses over The average duration of follow-up, study group size time, we relied only on data from the most recent pub- (Table 1), and extent of exposure to PUVA varied greatly lication identified by us in March 1998. We limited among the studies. As a result, both the number of tu- our analysis to studies from the United States and Eu- mors expected and the precision of estimates vary greatly rope that enrolled at least 150 patients followed up Table 2 at least 5 years and published after 1984. In addition among studies. lists the incidence per 1000 per- to the PUVA Follow-up Study, we identified 8 stud- son-years for SCC overall and among patients exposed ies that met these criteria. For each study, we ab- to high and low levels of PUVA. The incidence of SCC stracted the following types of data: number of varied greatly among studies, even within dose groups. patients enrolled, number followed, person years Among high-dose patients, all other studies reported in- of follow-up, average age at enrollment, and per- cidence rates substantially less than the 65 per 1000 percentage of male. We also attempted to determine the son-years calculated for high-dose patients in the PUVA number of individuals and years of exposure among Follow-up Study.5 For 4 of 8 studies reporting results patients exposed to high and low doses of PUVA. among patients with low-exposure to PUVA, no SCCs Whenever it was possible, we defined high dose were noted.6,8,11,12 Every study noted a higher incidence as more than 200 treatments or 2000 J/cm2 and low dose as less than 100 treatments or fewer than 1000 of SCC among members of the study group exposed to 2 higher doses of PUVA. In addition to the PUVA Fol- J/cm . If data in the study were not available in a for- 5 mat that permitted analysis using either of these defi- low-up Study, a point estimate of the IRR for SCC for nitions for dosage, the definition of high- and low- high- vs low-dose groups can be calculated for 4 stud- dose PUVA as defined by the author(s) of the ies. For these 4 studies, the IRRs ranged from 7 to 13 (high- individual study was used. On the basis of the tu- vs low-dose PUVA). In 4 additional studies, the point es- mor experience reported in each article, we at- timate was infinite. For these 4 studies the lower bound tempted to calculate an overall incidence rate for SCC of the 95% confidence interval in every case was more as well as incidence rates per 1000 person-years among than 1 (ie, risk was significantly higher for high-dose patients exposed to high and low doses of PUVA. patients). For the 8 studies other than the PUVA Fol- From these data we then calculated the incidence rate ratios (IRRs) for high- vs low-dose exposure groups low-up Study, the pooled estimate of the IRR for SCC for (ie, the ratio of the incidence of tumors per 1000 per- high- vs low-dose PUVA equals 14.0 (95% confidence in- son-years of follow-up) in the high-dose group com- terval, 8.3-24.1). Therefore, the estimate of the increase pared with that in the low-dose group. To calculate in risk of SCC for high-compared with low-dose expo- summary IRRs and estimate the confidence inter- sure to PUVA is significantly higher for these studies com- vals for individual report results we used exact meth- pared with that calculated by the PUVA Follow-up Study ods.14,15 We also calculated the ratio of BCCs to SCCs (IRR high vs low, 5.9).5 overall, and according to level of exposure to PUVA In addition to the PUVA Follow-up Study data,5 7 for each study and overall, and compared these of the 8 studies6-9,11-13 provided incidence data on BCC. ratios with the ratio we have reported for the PUVA As detailed in Table 3, in 5 of 7 studies the number of Follow-up Study.5 SCCs detected among patients exposed to high doses of PUVA was greater than the number of BCCs.7-13 Overall, among patients exposed to high doses of PUVA the ratio of SCCs to BCCs was 3:1 (Table 3), a ratio almost iden- RESULTS tical to the 2.7:1 ratio reported for the PUVA Follow-up Study.5 Among patients exposed to low doses of PUVA, Table 1 provides a summary of the characteristics of the all studies that documented BCCs detected more BCCs patient populations in the 9 studies analyzed. Except for than SCCs (Table 3). the PUVA Follow up Study,5 all studies were retrospec- tive cohort studies. Six of the 8 studies reported on com- COMMENT pleteness of follow-up.6-9,11,13 The percentage of patients eligible for follow-up who had complete follow-up ranged Using meta-analysis, we have assessed the extent to which from 43% to 78%. The PUVA Follow-up Study followed the published findings from different patient groups with up more than 90% of patients for at least 5 years.5 The psoriasis and exposed to PUVA studied in the United study by Lindleof et al10 was unique, since it used re- States and Europe provide consistent estimates of the risk cord linkage between a registry of PUVA-exposed pa- of SCC associated with high-dose exposure to PUVA. Our tients and the Swedish Cancer Registry. It is likely to have analysis demonstrates both striking areas of agreement complete ascertainment. In addition to the PUVA Fol- as well as some substantial differences among the find-

ARCH DERMATOL / VOL 134, DEC 1998 1583

©1998 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 10/02/2021 Table 1. Characteristics of the PUVA Follow-up Study5 and the 8 Studies of the Meta-analysis*

No. Patients No. (%) Patients Person-years Years of Country of Source, y Enrolled at Follow-up Follow-up Follow-up Male, % Origin High Dose† Stern and Laird,5 1994 1380 1049 (91) 13 637 13 65 United States Ͼ299 Rx Brunzeel et al,6 1991 334 260 (78) 3328 12.8 49 Netherlands Ͼ200 Rx Chuang et al,7 1992 492 310 (63) 1674 5.4 51 United States Ͼ1000 J/cm2 Cockayne and August,8 1997 150 150 1500 10 NR United Kingdom Ͼ1000 J/cm2 Henseler et al,9 1987 3175 1643 (52) 13 144 8 59 Germany Ͼ2000 J/cm2 Lindelof et al,10 1991 4799 NR ‡ 7 43 Sweden Ͼ200 Rx Maier et al,11 1996 1149 496 (43) 3372 6.8 55 Austria Ͼ2000 J/cm2 McKenna et al,12 1995 245 NR ‡ 9.5 59 United Kingdom Ͼ200 Rx Tanew et al,13 1986 418 297 (71) 1485 5 NR Austria Ͼ2000 J/cm2

*PUVA indicates psoralen–UV-A; Rx, treatment; and NR, not recorded. †Definition of high-dose exposure for this study. ‡Data to calculate not provided.

Table 2. Incidence and Incidence Rate Ratios (IRRs) and Table 3. Number of SCCs and BCCs by High- and Low-Dose 95% Confidence Interval (CI) of Squamous Cell Carcinoma PUVA Level* by Dose of PUVA* High Dose Low Dose All Doses All High Low Source, y Doses Dose Dose IRR† 95% CI Source, y SCC BCC SCC BCC SCC BCC 5 Stern and Laird,5 20 65 10 5.9‡ 4.0 to 8.7 Stern and Laird, 1994† 195 75 80 114 275 189 6 1994 Brunzeel et al, 1991 13‡ 19‡ 0 4 13 23 7 Brunzeel et al,6 1991 3 4§ 0 Infinity 1.8 to infinity Chuang et al, 1992 3 1 1 4 4 5 8 Chuang et al,7 1992 ࿣ 5.3 0.47 11.3 1.2 to 298 Cockayne and August, 500 050 Cockayne and August,8 3.3 13.8 0 Infinity 3.9 to infinity 1997 9 1997 Henseler et al, 1987§ 10 0 8 18 18 18 10 Henseler et al,9 1987 ࿣ 8.5 0.66 12.7 4.9 to 33.6 Lindelof et al, 1991࿣ 11 ... 13 ... 24 ... 11 Lindelof et al,10 1991 1.7 5.5 0.47 11.8 5.2 to 26.7 Maier et al, 1996 4 0 0 3 4 3 12 Maier et al,11 1996 1 10.8 0 Infinity 4.3 to infinity McKenna et al, 1995 3 4 0 2 3 6 13 McKenna et al,12 1995 1.7 9 0 Infinity 2.5 to infinity Tanew et al, 1986 5 0 4 18 9 18 Tanew et al,13 1986 4.3 20 3.2 6.3 1.6 to 27.1 All tumors except data 72 24 38 66 110 90 from Stern and Laird5 *Per 1000 person-years of follow-up. PUVA indicates psoralen–UV-A. †Incidence rate ratio for cohort members exposed to high vs low doses of *SCCs indicates squamous cell carcinomas; BCCs, basal cell carcinomas; PUVA. PUVA, psoralen–UV-A; and ellipses, not applicable. See the “Materials and ‡The IRR is adjusted for confounding variables. Methods” section for definitions of high and low doses. §High- and mid-dose PUVA levels combined (high dose defined as Ͼ200 †Population rate count 1 tumor in each calendar year even if a patient treatments or 2000 J/cm 2; mid dose, between 101 and 199 treatments or developed multiple tumors in 1 year. 1001 and 1999 J/cm 2. ‡Includes 3 patients with inflammatory conditions other than psoriasis, ࿣Insufficient data for calculation. and includes oral and PUVA bath. §Number of patients, not number of tumors. ࿣High- and mid-dose PUVA levels combined. ings of these studies. The greatest difference in the findings among the studies we reviewed was the overall incidence of SCC. Reported incidence rates varied from 1 those for Swedish women. Rates in the Netherlands18 and to 4 per 1000 person-years. These incidence rates are dra- Switzerland19 are also significantly lower than estimated matically lower than the incidence of 20 per 1000 person- rates for northern areas of the United States (Table 4). years for the PUVA Follow-up Study.5 However, once SCC For patients exposed to high-dose PUVA, adjusting for incidence was stratified according to exposure to PUVA, differences in baseline rates of incidence rates of SCC be- the rates for patients exposed to high doses of PUVA were tween the United States and Europe results in compa- more similar, ranging from 4 to 20 SCC per 1000 person- rable incidence estimates for SCC in all European stud- years for the 8 other studies compared with an inci- ies and the PUVA Follow-up study.5 The average dose dence of 65 per 1000 person-years in the PUVA Fol- among high-dose patients is higher for the PUVA Fol- low-up Study.5 low-up Study cohort5 than for the other study groups. A number of factors probably account at least in part Since risk is likely to increase with dose in a continuous for differences in SCC risk among studies within dose fashion, the US cohort,5 who had the greatest exposure groups. First, studies varied in the characteristics of the to PUVA and the highest baseline incidence of SCC, would populations studied they were from many different coun- be expected to have the highest incidence of tumors. In tries and the incidence of SCC varied greatly among the addition, the PUVA Follow-up Study was the only pro- regions studied. As detailed in Table 4, even northern spective study and had the highest follow-up rate. There- areas of the United States have incidence rates of SCC fore, more complete ascertainment of tumors would be for men that are 5 to 8 times higher than in Sweden.16,17 expected than for other studies, except perhaps for the Rates for women living in the northern US are 3 times study of Lindelof et al.10

ARCH DERMATOL / VOL 134, DEC 1998 1584

©1998 American Medical Association. All rights reserved. Downloaded From: https://jamanetwork.com/ on 10/02/2021 size of the cohorts exposed to high doses of PUVA and Table 4. Age-Adjusted Incidence of SCC by Country, Sex, limited duration and completeness of follow-up, the num- and Ratio of BCC to SCC* ber of patients developing SCCs is sufficiently small that these studies do not provide substantial additional data SCC Incidence† Ratio of to address the morbidity of SCCs that arise in associa- Country Males Females BCC to SCC tion with PUVA treatment. These studies also provide Eindhoven, the Netherlands18 14.7 3.4 5.3:1 some reassuring information. At least within an average 5 Vaud, Switzerland19 29.1 18.0 2.9:1 of 8.6 years and a maximum of 13 years of follow-up, Sweden17 22.0 10.0 ‡ the pooled data suggest that low levels of exposure to Oregon, United States16 106.1 29.8 ‡ PUVA are unlikely to greatly increase the risk of SCC. British Columbia, Canada20 31.2 16.9 4.4:1

*SCC indicates squamous cell carcinoma; BCC, basal cell carcinoma. Accepted for publication July 24, 1998. †Tumors per 100 000 person-years. This study was supported by grant NO1-AR4-2214 from ‡Basal cell carcinoma data not available. the National Institute of Arthritis, Musculoskeletal and Skin Diseases, Bethesda, Md. In the 8 studies6-13 we reviewed, among patients ex- A complete list of the articles excluded from our meta- posed to low doses of PUVA, the results are consistent analysis is available on request from us. with those expected from published population-based in- Reprints: Robert S. Stern, MD, Department of Der- cidence rates for Europe and the United States. The in- matology, Beth Israel Deaconess Medical Center, Harvard cidence of SCC in the PUVA Follow-up Study5 among Medical School, 330 Brookline Ave, Boston, MA 02215 low-dose patients is significantly higher than that esti- (e-mail): [email protected]. mated from population data. This finding may well re- flect the substantial proportion of patients in this cohort with exposure to other carcinogens that are likely REFERENCES to increase SCC risk.5 1. Parrish JA, Fitzpatrick TB, Tahenbaum L, Pathak MA. Photochemotherapy of pso- There are at least 3 findings that support a dose- riasis with oral methoxsalen and longwave ultraviolet light. N Engl J Med. 1974; dependent increased risk of SCC in association with PUVA 291:1207-1211. use that are consistently noted among all studies we re- 2. Dunnick JK, Forbes PD, Eustis SL, Hardisty JF, Goodman DG. Tumors of the skin in the HRA/Skh mouse after treatment with 8-methoxypsoralen and UVA radia- viewed. Within all studies with adequate data, the inci- tion. Fundam Appl Toxicol. 1991;16:92-102. dence of SCC is significantly higher among patients ex- 3. Stern RS, Laird N, Melski J, Parrish JA, Fitzpatrick TB, Bleich HL. Cutaneous squa- posed to higher doses of PUVA compared with those with mous-cell carcinoma in patients treated with PUVA. N Engl J Med. 1984;310: 1156-1161. lower-dose exposure. None of the studies reviewed pro- 4. Stern R, Lange R. Non–melanoma skin cancer occurring in patients treated with vide data to suggest that differences in the distribution PUVA 5 to 10 years after first treatment. J Invest Dermatol. 1988;91:120-124. 5. Stern R, Laird N. The carcinogenic risk of treatments for severe psoriasis. Can- of other risk factors for squamous cell cancer between cer. 1994;73:2759-2764. high- and low-dose exposure groups within a study would 6. Brunzeel L, Bergman W, Hartevelt HM, et al. High single dose European PUVA explain the observed dose-related increase in risk. Also, regimen also causes an excess of non-melanoma skin cancer. Br J Dermatol. 1991;124:49-55. the higher ratio of SCC to BCC among patients with 7. Chuang T, Heinrich L, Schultz M, Reizner G, Kumm R, Cripps D. PUVA and skin higher-dose exposure to PUVA than patients with lower- cancer: a historical cohort study on 492 patients. J Am Acad Dermatol. 1992; dose exposure supports high-dose PUVA exposure as a 26:173-177. 8. Cockayne SE, August PJ. PUVA photocarcinogenesis in Cheshire. Clin Exp Der- risk factor for SCC. In fact, the ratios of SCC to BCC for matol. 1997;22:300-304. the other studies are nearly identical to those noted for 9. Henseler T, Christophers E, Honigsmann H, Wolff K. Skin tumors in the Euro- 5 pean PUVA study: eight-year follow-up of 1643 patients treated with PUVA for the PUVA Follow up Study in each PUVA dosage group. psoriasis. J Am Acad Dermatol. 1987;16:108-116. As with all meta-analyses and secondary analyses of 10. Lindelof B, Sigurgeirsson B, Tegner E, et al. PUVA and cancer: a large scale epi- data, our analysis has a number of important limita- demiological study. Lancet. 1991;338:91-93. 11. Maier H, Schemper M, Ortel B, Binder M, Tanew A, Honigsmann H. Skin tumors tions. First, there may well be factors not reported or rec- in PUVA for psoriasis: a single center follow-up of 496 patients. Dermatology. ognized by us that might explain the higher risk of SCC 1996;193:185-191. among high- compared with low-dose PUVA patients in 12. McKenna K, Patterson C, Handley J, McGinn S, Allen G. Cutaneous neoplasia following PUVA therapy for psoriasis. Br J Dermatol. 1995;134:639-642. any study we reviewed. Such factors might include older 13. Tanew A, Honigsmann H, Ortel B, Zussner C, Wolff K. Nonmelanoma skin tu- age or a higher proportion of males in the high-dose group, mors in long-term photochemotherapy treatment of psoriasis: an 8 year follow-up study. J Am Acad Dermatol. 1986;15:960-965. factors that were only adjusted for in the PUVA Fol- 14. Rothman KJ. Modern Epidemiology. Boston, Mass: Little Brown & Co Inc; 1986. 5 10 low-up Study and the analysis of Lindelof et al. The 15. Guess HA, Lydick EG, Small RD, Miller LP. A general estimator for the variance proportion of patients with exposure to carcinogens other of the Mantel-Haenszel odds ratio. Am J Epidemiol. 1987;125:340-347. 16. Glass AG, Hoover RN. The emerging epidemic of melanoma and squamous cell than PUVA is also likely to be higher among high- than skin cancer. JAMA. 1989;262:2097-2100. low-dose patients. If one assumes a similar effect for other 17. National Board of Health and Welfare, The Cancer Registry. Cancer Incidence in studies attributable to these exposures to that which was Sweden. Stockholm, Sweden: National Board of Health and Welfare, The Cancer 5 Registry; 1987:24. observed in the PUVA Follow-up Study, the pooled es- 18. Coebergh JW, Newmann HA, Vrints LW, van der Heijen L, Meijer WJ, Verhagen- timate of increase in the risk of SCC for high- vs low- Teulings MT. Trends in the incidence of non-melanoma skin cancer in the SE Netherlands 1975-1988: a registry-based study. Br J Dermatol. 1991;125: dose PUVA patients would be nearly identical in the 8 353-359. other studies and in the PUVA Follow-up Study. 19. Levi F, Franceschi S, Te VC, Randimbison L, La Vecchia C. Trends of skin cancer Taken together, all the studies we analyzed pro- in the Canton of Vaud, 1976-53. Br J Cancer. 1995;72:1047-1053. 20. Gallagher RP, Ma B, McLean D, et al. Trends in basal cell carcinoma, squamous vide evidence that high-dose exposure to PUVA in- cell carcinoma, and melanoma of the skin from 1973 through 1987. J Am Acad creases the risk of SCC. However, because of the limited Dermatol. 1990;23:413-421.

ARCH DERMATOL / VOL 134, DEC 1998 1585