| Relative values of different outcomes | Incidence rates for malignancy
Melanoma is the major cause of death due to skin cancer as a whole so any increase in risk of melanoma is considered of greater significance when compared to risk of SCC or BCC. Non-melanoma skin cancers (SCC and BCC), whilst undesirable, are generally curable; SCC has greater implications than BCC in terms of impact on health as it can be aggressive and metastasise, especially at genital and lip sites, whereas this is rare with BCC. Skin cancers as a whole are common in the UK and therefore any increase in skin cancer incidence is potentially significant. |
| Trade off between clinical benefits and harms | PUVA is associated with an increased risk of skin cancer, both non-melanoma and melanoma. The risk is most marked for squamous cell carcinoma, is consistent across different studies and populations, is related to number of UV exposures, does not reduce on stopping PUVA and persists for a lifetime. There is no absolute safe number of exposures. The current belief that fewer than 200 treatments is safe practice is not supported by the data. This led the GDG to recommend that cumulative number of exposures to PUVA should be documented.
There is a particular risk of genital SCC, which has been addressed by a change in clinical practice with the introduction of genital shielding in the 1990s.
People with skin types 1 and 2 are at a greater risk of SCC than people with skin types 3 and 4, but there is a risk for all skin types. Subsequent treatment with ciclosporin further increases the risk and long-term treatment with methotrexate also increases the risk, although it was unclear whether the risk was associated with methotrexate exposure before or after PUVA. However, it is likely that methotrexate use after PUVA, as with ciclosporin, is also a greater risk than before PUVA because the mechanism is widely thought to involve immunosuppressive treatments after PUVA inducing the emergence of skin cancer.
Regarding the exposure to both PUVA and UVB the data were limited and mainly focused on broadband UVB, so the GDG agreed not to include UVB as a known additional risk factor for skin cancer in people receiving PUVA.
The GDG noted that the relative and absolute risk of SCC compared with the general population increased markedly once more than 160 PUVA exposures had been received, so it was agreed that it is unreasonable to expose people to greater than 160 treatments.
When considering the place of PUVA for the treatment of psoriasis, the GDG considered the efficacy and adverse effects of UVB as those patients who are suitable for PUVA are also likely to be suitable for UVB. In relation to efficacy, clearance rates are probably equivalent; the 2–3 week improved time to clearance, and 1.55 relative risk of relapse with oral PUVA were not felt to offset the increased inconvenience, risks (both short-term in relation to taking an oral psoralen and long-term risk in relation to skin cancer) and cost when compared to NBUVB. Bath PUVA was less effective that NBUVB in terms of time to clearance and relapse rates. The GDG concluded that it would be difficult to justify the use of PUVA in patients who had not already failed UVB.
There were no studies investigating the efficacy of PUVA in people who had failed UVB to be confident that PUVA would be effective in these individuals. The GDG noted that the efficacy rates of oral PUVA were high in terms of clearance (and may be better than methotrexate or ciclosporin or some of the biological drugs). However, PUVA is not an intervention that can be used to maintain remission (relapse rate 45% by 6–12 months) and the risks of skin cancer are clinically relevant, lifelong and compounded by future use of other treatments used to treat psoriasis, even accepting that the morbidity and mortality rates from skin cancer are low, that some of the data relate to very high numbers of exposures to PUVA over prolonged periods of time and that the risks in relation to skin cancer or other risks of alternative treatment options such as methotrexate or biological therapy are poorly documented. The GDG concluded that for most people who had failed or relapsed rapidly with NBUVB, use of PUVA may not be justified if other treatments could be used.
The GDG did not wish to limit treatment options by making a recommendation not to use PUVA at all, but felt it important to highlight the risks of PUVA and groups at particular risk, and offer PUVA only when other options had been actively considered and rejected.
Healthcare professionals should fully explain the risks of PUVA treatment including the absolute risk, and the potential implications of PUVA in relation to future treatment options. Fully informed written consent should be obtained.
The GDG wished to ensure that the risk of significant PUVA-related harm was minimised by recommending that those already in high risk groups are offered annual surveillance for skin cancer.
When considering the role of local PUVA for palmoplantar pustulosis, there are very few effective interventions for this condition and the area of skin exposed to UVA is very limited; hence the clinical benefit of local PUVA, if the impact of palmoplantar pustulosis is high, may be justified.
The GDG noted that the long-term risks of PUVA were relatively well documented compared to those associated with the alternative options, including systemic biological and non-biological therapies; the GDG were aware of long-term registries comparing the risks of these different interventions and agreed that participation should be encouraged.
Only limited data were available for UVB. It was noted that data up to five years are now available for NBUVB and no significant increase in skin cancer risk is reported, whereas risks associated with PUVA were evident by this time point. The GDG discussed the evidence that after NBUVB the risk of BCC was more increased than of SCC, in contrast to PUVA. The GDG considered that in light of experience with PUVA where there may be a prolonged lag period between use of PUVA and development of skin cancer, and that the risk is related to the number of exposures, it is important that all patients receiving phototherapy of any kind should have the cumulative amounts of phototherapy recorded carefully.
From GDG knowledge, people with a personal history of skin cancer or predisposition to skin cancer (for example, xeroderma pigmentosum) should not be offered PUVA. It was also noted that risk rates reported in more recent studies are likely to exclude groups of people already at risk of skin cancer (both non-melanoma and melanoma). The GDG agreed that alternative treatment strategies to PUVA should be sought in younger people due to the lifetime risk of skin cancer and impact on potential future treatment options. Whilst the GDG did not review data pertaining to genetic predisposition as it was outside of the remit of the scope, the GDG agreed an important consensus safety recommendation. People with a personal history of skin cancer or predisposition to skin cancer (for example, xeroderma pigmentosum) should not be offered PUVA. |
| Economic considerations | No economic evidence was available to inform the GDG on how the risk of skin cancer may impact the relative cost-effectiveness of different interventions including systemic and photo therapies used in the treatment of psoriasis. In the absence of such information, the GDG considered the balance between short-term gains in the form of disease improvement and increased long-term risks of different skin cancers. For most patients, the GDG did not consider the increased long-term risks of psoriasis treatments (in terms of associated morbidity, mortality or costs) to outweigh the benefits in the short-term, but did highlight the importance of carefully communicating a treatment’s potential benefits and harms to patients. However, the evidence showed that some patients may be at even higher risk given a personal history of skin cancer, skin type, previous and future treatments. In particular they also discussed the synergistic effect certain treatments have when combined or used in immediate succession (e.g. PUVA immediately preceded or followed closely by ciclosporin) and felt that this should be avoided because the risks far outweighed the potential benefits.
The GDG considered that different skin cancers have different prognoses and treatment costs. BCC and SCC rarely metastasise or lead to death, but they can cause considerable morbidity. The estimated cost of removing BCC and SCC is £132 as an outpatient procedure (HRG JC07Z)74.
In order to ensure patients are not exceeding reasonably safe levels of exposure to phototherapy, the GDG considered it important to document cumulative number of treatments. They believed that benefit of documentation, arising from cancers and associated morbidity and mortality avoided, was likely to represent good value for NHS resources. |
| Quality of evidence | There was a lack of data for a number of interventions and subgroups:
No subgroup data for disease severity, age at first exposure, smoking and alcohol. Nor were there data on oral versus bath PUVA. No studies designed specifically to investigate the risk associated with methotrexate, UVB or tar. There were insufficient data to assess the risk of skin cancer associated with exposure to NBUVB or biologics as the available studies had a relatively short follow-up time and were not controlled for confounding factors such as prior treatments and in one 304 the reference cohort was not from the same geographic location so different natural UV exposure could confound the findings. Future reports on the NBUVB cohort are awaited. The GDG noted that there is a suggestion, mainly from animal studies, that biologics may have a carcinogenic effect.
The ideal study design to address this question would have been a cohort study designed specifically to compare people with psoriasis not treated with an intervention with people with psoriasis treated with an intervention. This would help to determine the specific risk associated with the intervention independent of any risk associated with psoriasis per se. However, this is not a feasible design. Therefore, for all studies the unexposed cohort was a general population sample and so would have included a proportion with psoriasis and potentially with exposure to the interventions beings assessed as risk factors (e.g., PUVA or ciclosporin).
All of the studies also had a high level of outcome surveillance bias as there is likely to be more complete ascertainment of skin cancer cases among the exposed cohort who were actively followed-up and examined compared with the general population where diagnoses may be missed.
In addition, the majority of the data were derived from the Stern cohort from 16 centres in the USA, collected since the 1970s and followed-up for many years. The GDG discussed that the standard PUVA regimen in the USA differs from the UK and that the baseline SCC incidence is higher in the USA. There is a higher proportion of people with skin type 3 and above in this cohort. Whilst the GDG agreed that data from a UK cohort would be more relevant they agreed that the Stern data set was a very large study with a long follow up period. The GDG were aware of data from a retrospective European PUVA study (Lindelof 1991) with approximately 7 year follow-up that did not meet the inclusion criteria (because the population was only 50% psoriasis and it was a retrospective cohort). It was noted that the Lindelof study also demonstrated an exposure-dependent increase in the risk of squamous cell skin cancer and a greater risk in those with fairer skin but the magnitude of the risk was lower than that in the Stern cohort.
It was noted that the stratification of PUVA dose or number of exposures varied between the studies in the Stern cohort and it was unclear whether the thresholds for stratification were pre-specified or had been chosen based on the data, which could lead to bias.
The GDG also noted that the results from the Stern cohort may be biased by the fact that 39 patients out of the 1380 had a history of skin cancer before PUVA (so the reported rates may not be related to true incidence) and this was not controlled for in all analyses. According to current practice these individuals would not have been offered PUVA.
Due to the long-term nature of this study, less than 80% of the original cohort remained after 1984. The authors report that most of the loss was due to death and consistent with the expected rate. Withdrawal and loss to follow up were acceptable, but reasons for loss were unclear. Therefore we do not know if the characteristics of those lost are the same as those who remained in the study and whether this could have biased the results.
Studies differed in their method of recording tumour incidence. Some used a total count where each tumour is counted; others used person counts, whereby the first tumour of a specific type is counted. The latter tends to be a conservative estimate of risk. Other studies report population counts, including reporting only the first tumour in a year in an individual. This approach may limit the influence of cohort members who may be outliers (i.e. those rare individuals who develop a large number of tumours per year) by restricting to annual incidence. This last method was also in accordance with the method of recording in the national registries that were used to estimate the expected incidences in the unexposed cohort in the Stern studies. Some studies included pre-malignant skin cancers, and so the risk of skin cancer would potentially have been over-estimated in these studies compared with studies that did not include pre-malignant skin cancers. Additionally it was apparent that genital sites are especially vulnerable and current practice is to shield the genital area during exposure to PUVA. The early use of PUVA in the Stern cohort will have been prior to the practice change to use genital shielding, and therefore an overestimate of the current risk associated with PUVA. There are no data on the risk when genital tumours are excluded, although the studies looking at genital tumours specifically did adjust for variation in genital shielding between enrolled centres.
The studies also varied in the statistical analysis, with many of the earlier studies not performing a regression analysis to control for confounders, instead matching the exposed and unexposed cohorts for age, sex and geographic location. Only one study used Cox proportional hazards to take account of time in the analysis, although other studies did control for time in the analysis by different methods. Even when regression analysis was performed the number of confounders that were adjusted for varied between the studies and was not complete in any: use of UVB and history of skin malignancy were rarely controlled, although age and geographic residence were used as surrogate markers of cumulative sun exposure
The GDG noted specific biases in the following studies:
Stern 1997 study on melanoma: the threshold for the different time periods appeared to have been selected based on the data and the observed increase in incidence, which introduces bias. Marcil 2001: there were very few people receiving ciclosporin. Paul 2003: this study was primarily designed to assess the risk of ciclosporin and had a high attrition rate. The duration of follow up for PUVA is unclear. 34% of the cohort received their systemic treatment during the follow up period, and this did not seem to be taken into account in the analysis. Due to these major limitations the GDG gave little weight to this study, apart from the ciclosporin findings.
|
| Other considerations | One of the later follow-up studies in the Stern PUVA cohort demonstrated no independent carcinogenic effect of UVB, topical tar or ionising radiation, which conflicted with earlier findings. This may be because PUVA is the main carcinogen and as more is received it outweighs the impact of other factors.
In light of the absence of data noted above, the GDG believed that future research was warranted in this area and made a research recommendation. |
