Drug-Induced Malignancy

General Principles and a Drug Causation Determination Algorithm

Q64.1 The determination of drug causation for cutaneous hypersensitivity reactions is quite difficult. Even more challenging, is the determination of causation for possible drug-induced malignancies. The reasons for this difficulty are detailed subsequently. In general, causation of a ‘drug reaction’ of any nature can best be determined to a reasonable degree (virtually “never” does the level of certainty reach ‘100%’) of certainty, with a combination of the following steps:

• 1.

  Challenge—original drug course circumstances;

• 2.

  Dechallenge—stopping the drug in question and observing if the ‘reaction’ subsides;

• 3.

  Rechallenge—very selectively, observing for recurrence of the ‘reaction’;

• 4.

  Exclusion—of nondrug precipitators of the same outcome/condition.

As a supplement to the earlier algorithm, determination of biologic plausibility is of significant interest. Details of this brief algorithm are listed in Table 64.2 .

The difficulty of determining causation of drug-induced malignancy has at least three different realities. First is that the ‘dechallenge’ step is generally not of value, given that the malignancy will not reliably resolve with drug discontinuation alone. Second, all malignancies potentially induced by drugs occur at a given rate in various populations (‘background noise’), totally independent of drug therapy. Third, it is unwise from either an ethical or a medicolegal viewpoint to rechallenge the drug in question, if the malignancy does indeed resolve with the dechallenge step.

It is most important to determine whether the incidence of a given malignancy is statistically significantly increased compared with this ‘background noise.’ Thus, reliable incidence data for the general population and for specific diseases are of tremendous importance.

Entire Population Versus Disease-Specific Databases

Comparisons with Drug-Specific Databases

Upon determining the background incidence of malignancy for RA and psoriasis, the final comparison requires incidence data in ‘person-years’ from pharmaceutical company databases, or published data for a specific drug to compare with the disease-specific incidence rates. It is essential to have reliable denominator data, which are most easily obtained from Phase II and III clinical trials. In most settings, the pharmaceutical companies continue to track the patients from the Phase II and III trials indefinitely, to determine long-term risk for lymphoma and other malignancies of importance. Given the typical long-term time period of drug-induced carcinogenesis, the data from long-term follow-up studies is essential.

Given the tremendous importance of establishing certainty as to whether drugs prescribed in dermatology increase the risk of any malignancy, large pharmaceutical company databases, compared with disease-specific databases, are currently the optimal way to determine the risk of drug-induced malignancy. The statistical analyses used for tables concerning individual drug risk, subsequently in the chapter, are performed in this manner.

Confidence intervals (95% CI) must be reasonably ‘narrow’ to draw the most precise conclusions. Databases generally need to contain 5000 to 10,000 person-years of data to allow reasonably narrow CI. Data with this large number of person-years are much more readily available in the rheumatology literature. It stands to reason that more recently released drugs (or older drugs studied for newer indications) will not have the above number of person-years to attain the most reliable statistical analyses.

Limitations of Voluntary Reporting Systems and Case Reports/Series

With these concepts in mind, case reports and the experience of individual practitioners are of very limited value. Even case series, collating all reports from the literature, are of limited value. Case series obtained from spontaneous reporting systems do not have denominator data and do not contain a control group to make any valid statistical analyses. Thus, case reports and case series just bring up ‘possibilities’ and do not reliably establish ‘causation’. Realistically, the dechallenge and rechallenge steps of the drug causation algorithm are not possible or reliable, leaving a very low level of certainty concerning causation with individual case reports of ‘possible’ drug-induced malignancy.

Multistep Model of Carcinogenesis

For a number of decades, the multistep model of carcinogenesis has been generally accepted. Q64.3 In this model, there are three distinct ‘steps’ for a cell to develop into cancer, generally requiring many years or decades to proceed through the process. The multistep carcinogenesis model includes the: (1) initiation, (2) promotion, and (3) progression steps. The progression step is generally subdivided into: (a) malignant conversion and (b) metastasis subcomponents. The initiation and progression steps in this multistep model are ‘genetic’ (mutations involved) and the promotion step is ‘epigenetic’ (no mutations involved).

At least two mutations are generally required for a cell to develop a malignant genotype and biologic behavior. More than chance location near vascular or lymphatic supply is necessary for metastasis to occur. A malignant keratinocyte must develop subsequent mutations that allow penetration of the basement membrane zone, invasion of vascular structures, and localization/growth, at a distant anatomic site. With these multiple steps required for a given cell to develop malignant features and subsequently metastasize, it is logical that many years (or even decades) are necessary for the process to evolve.

The initiation step requires at least one somatic mutation, whereas the tumor promoters (promotion step) induce a clonal expansion of the mutated cell into a much larger population of atypical cells. In studies of chemical carcinogenesis, benz(a)pyrene is an example of an ‘initiator,’ whereas phorbol esters, such as tetraphorbol acetate, have been established as ‘tumor promoters.’ In studies of ultraviolet photocarcinogenesis, ultraviolet B can serve as both initiator and promoter.

A third area of study that perhaps is most relevant to drug-induced malignancy, is viral carcinogenesis. Oncogenic viruses include various subtypes of human papilloma virus (HPV, inducing squamous cell carcinoma [SCC] of skin and cervix) and Epstein-Barr virus (EBV, inducing various lymphoma subsets). Q64.4 Recent evidence has demonstrated viral causation of Merkel cell carcinoma (‘Merkel cell polyomavirus’) and has suggested a possible viral role in melanoma causation (‘melanoma-associated retrovirus’). This concept will be expanded in the following section on review of malignancy risk with organ transplantation.

Oncogenes and Tumor Suppressor Genes

Q64.5 Two categories of genes with a significant role in carcinogenesis are oncogenes and tumor suppressor genes. An example of an oncogene that is of importance to carcinogenesis in dermatology, is the ras oncogene. The p53 tumor suppressor gene is perhaps the best-studied example concerning mutations responsible, at least in part, for various cutaneous malignancies. It is well beyond the scope of this chapter to expand on this very important area of study in carcinogenesis.

However, one ‘model’ concerning the role of oncogenes and tumor suppression genes in carcinogenesis that has assisted this author’s understanding can be cited. In this ‘automotive’ model of carcinogenesis, the oncogenes are the ‘gas pedal’ and tumor suppressor genes are the ‘brake pedal’. If the gas pedal is stuck down to the floor (= oncogene mutation), but the brakes are functioning adequately, the ‘car’ does not proceed out of control (no malignancy). If the brake pedal is not functioning (= tumor suppressor gene mutation), but the gas pedal is not ‘stuck’ down to the floor, the car again does not proceed out of control (no malignancy). Only with the gas pedal stuck and the brake pedal malfunctioning, does the car proceed ‘out of control.’ In this model, the car proceeding ‘out of control’ is analogous to malignant cells dividing in an uncontrolled fashion (because of mutations of both an oncogene and tumor suppressor gene).

The reader is invited to delve further into the intriguing topic of carcinogenesis, particularly as the topic is applied to cutaneous malignancies. Some additional reviews of potential interest are cited here.

Malignancies Having Increased Risk with Organ Transplantation

A tremendous amount of data has been published over the past 5 decades concerning malignancies that occur at an increased incidence, in patients undergoing solid organ transplantation, compared with the population at large. Substantial long-term data, concerning azathioprine and cyclosporine derived from these transplantation databases, provide some initial insights into the immunosuppression-related malignancies. A major database initiated in the 1970s by Penn and Starzl evaluated predominantly renal transplantation patients. The primary conclusions from this database include the following:

• 1.

  Both azathioprine and cyclosporine have a similar risk of malignancy induction in this population.

• 2.

  Cyclosporine-treated patients tend to develop selected malignancies earlier than patients treated with azathioprine.

• 3.

  Malignancies most consistently increased, versus the control population, include NHL (generally B-cell phenotype), SCC, and Kaposi sarcoma.

• 4.

  Smaller increases in basal cell carcinomas (BCC) have been demonstrated.

• 5.

  The most common malignancies in the general population, such as breast, colorectal, and lung carcinomas, do not appear to have a substantially increased incidence compared with control populations.

Viral Cofactors

Q64.6 One interesting feature of the three malignancies, with the highest relative risk in the transplantation databases, is that each tends to have a viral cofactor in a high percentage of cases ( Table 64.3 ). EBV is responsible for a significant percentage of non-Hodgkin B-cell lymphomas. HPV subtypes are commonly associated with SCC of the skin, oral mucosa, and female genitourinary tract. Malignancies at all three SCC sites are increased in transplantation databases. Finally, human herpesvirus-8 (HHV-8) has been shown to have a causal role in organ transplantation patients developing Kaposi sarcoma. Q64.4 More recently, viral causation and an increased incidence of solid organ transplantation have been established for Merkel cell carcinoma.

For the purposes of the remainder of this chapter, only EBV-induced lymphomas and HPV-induced SCC of various sites will be discussed, given their relative frequency and related clinical importance to the practicing dermatologist.

A reasonable conclusion is that drug-induced immunosuppression preferentially impairs the human body’s immune surveillance for viral-induced malignancies. It is not clear how immunosuppression preferentially supports the development of virally induced malignancies. One can speculate that immunologic effector mechanisms, such as cytotoxic T cells (CD 8), natural killer cells (NK cells), and tumor-infiltrating lymphocytes might be relatively impaired compared with patients who have never had a solid organ transplantation. Further discussion and speculation, regarding this specific aspect of immune surveillance of viral oncogenesis, is clearly beyond the scope of this chapter.

Unique Aspects of Transplantation Risk for Malignancy Induction

It is important to question whether the clinical circumstances in solid organ transplantation are an adequate model for assessing the possibility of drug-induced malignancy with autoimmune diseases, such as psoriasis, RA, and inflammatory bowel disease (IBD). The two most notable differences in comparing solid organ transplantation versus autoimmune disorders are:

• 1.

  In organ transplantation patients, there is an indefinite presence of a substantial source of ‘foreign’ antigens, whether a kidney, liver, or heart is transplanted; and

• 2.

  More aggressive, multidrug immunosuppressive regimens are generally required in the organ transplantation setting compared with the drug regimens used for autoimmune disorders (including psoriasis), in general.

Regarding the first issue, the antigen-processing cells, T cells, and B cells, albeit subdued somewhat by the immunosuppressive regimen, will never completely ‘view’ the foreign antigens as ‘self.’ The logical result is an indefinite low-grade immunologic stimulation of various lymphocyte subsets. It is not clear why B-cell lymphomas predominate, although EBV in infectious mononucleosis appears to preferentially populate this lymphocyte population. Regarding the second issue, it is intuitive that the greater the degree of drug-induced immunosuppression, the greater the impairment of immune surveillance for malignant cells, whether a virally-induced malignancy or not.

Currently, the term ‘posttransplantation lymphoproliferative disorders’ is used to include the spectrum of B- and T-cell lymphomas, Hodgkin disease, and lymphoma precursors occurring in this population. It is important to note that early on, these lymphomas may be reversible, although once well established, the lymphomas can be very aggressive.