PERSPECTIVES OPINION
Sorting out the risks in progressive multifocal leukoencephalopathy Leonard H. Calabrese, Eamonn Molloy and Joseph Berger Abstract | Progressive multifocal leukoencephalopathy (PML) is a rare, opportunistic infection of the central nervous system, caused by reactivation of the ubiquitous JC virus. PML is a devastating disease that is frequently fatal, and although survival rates have improved, patients who survive PML often experience considerable neurological deficits. PML was associated with a variety of immunosuppressive therapies in the past decade, but attribution of causality is difficult owing to the presence of confounding factors and to an inadequate understanding of the underlying pathogenesis of this disease. This uncertainty has hindered efforts for shared decisionmaking between physicians and their patients and, in some cases, discouraged the use of potentially beneficial therapies. We propose a categorization of immunosuppressive agents according to their risk of PML to support a better-informed decision-making process when evaluating the risks and benefits of these therapies. Calabrese, L. H. et al. Nat. Rev. Rheumatol. advance online publication 14 October 2014; doi:10.1038/nrrheum.2014.167
Introduction
Progressive multifocal leukoencephalo pathy (PML) is a devastating, and often fatal, opportunistic infection of the central nervous system. The diagnosis and clinical features of PML, the discussion of which is beyond the scope of this paper, were con cisely reviewed and formulated in a consen sus statement published in 2013. 1 The epidemiology of PML can be best under stood if divided in three epochs of disease: the first period extends from its initial des cription in the early 1980s, when it was predominantly associated with lympho proliferative diseases; the second coincided Competing interests L.H.C. has served as consultant to Genentech, GlaxoSmithKline and Pfizer in the area of progressive multifocal leukoencephalopathy (PML). E.M. has obtained research support from Roche, has served on advisory boards for BristolMyers Squibb, Pfizer and MSD, and received payment for developing educational materials relating to PML from GlaxoSmithKline. J.B. serves on the PML Adjudication Committees of Amgen, Astra Zeneca, Bristol-Myers Squibb, Eisai, Janssen, Millennium, Parexel, Pfizer, Roche and Takeda; he is a consultant to Genentech, Genzyme, Incyte, Inhibikase Therapeutics, Johnson & Johnson and Novartis, and has received grants from the PML Consortium, Biogen Idec and Novartis.
with the AIDS pandemic, when PML became strikingly prevalent as it affected 4–5% of all patients with AIDS;2 after this period, the epidemiology of PML changed dramati cally with the initial diagnosis of the disease in patients with multiple sclerosis (MS) and Crohn disease who were treated with the biologic agent natalizumab.3–5 No evidence for association of these two autoimmune diseases with PML was observed before the introduction of natalizumab. Subsequently, and over the past decade, the notion of druginduced PML was introduced in the field of rheumatology based on rare reports of the disease associated with the use of immuno suppressive therapies in the context of a variety of rheumatic disorders.6,7 Concerns over PML have been amplified by the knowledge that the condition, while remaining rare, is unpredictable, often fatal and has currently no effective treatment. The disease has a complex biology 8 and, despite advances in clinical and basic research, sig nificant deficits exist in our understanding of its pathogenesis. Because PML is a relatively new disorder for the rheumatology com munity, clinicians might lack confidence in assessing the risks of JC virus opportunis tic infection in their patients. Thus, both clinicians and patients must evaluate the
NATURE REVIEWS | RHEUMATOLOGY
risks of PML by taking into account factors such as the underlying disease, the associ ated immunosuppressive regimen and the level of risk patients are willing to accept. Although PML is associated with a variety of immunosuppressive therapies, several of which carry ‘black box’ warnings with res pect to this potential complication, immuno suppressive agents should not be grouped together, because the risk for PML varies greatly among them. Recognition of this fact is essential for informed, shared, therapeutic decision-making.
PML risk of immunosuppression
Prior to 2006, PML had not been diagnosed in patients with a rheumatic disease treated with a biologic agent. However, by that time, 35 patients with PML who had received standard immunosuppressive agents such as azathioprine, chlorambucil, cyclophos phamide, glucocorticoids and methotrexate, used alone or more commonly in com bination, were identified.6 Most reported cases occurred in patients with systemic lupus erythematosus (SLE) who were often exposed to minimal iatrogenic immuno suppression; importantly, evidence suggests that SLE itself predisposes to PML.9,10 The contribution of therapeutic agents to the risk of PML was not widely appreciated until 2005, when three cases of PML were associated with natalizumab, a biologic ther apy under investigation for the treatment of MS3,4 and Crohn disease5 at the time. Soon thereafter, PML was diagnosed in three other patients with psoriasis who were treated with efalizumab.11 The profound importance of these cases was emphasized by the absence of prior reports of PML associated with MS, inflammatory bowel disease (IBD) or psori asis. Ultimately, efalizumab was voluntarily withdrawn from the market. Natalizumab was temporarily withdrawn but later rein troduced for the treatment of both MS and Crohn disease under a rigorous risk mitiga tion plan, the TOUCH (TYSABRI Outreach: Unified Commitment to Health) prescribing programme.12 TOUCH mandates the educa tion of healthcare providers and patients as to the risks of natalizumab therapy, as well as close monitoring of patients that receive natalizumab infusions for warning signs of PML.12 ADVANCE ONLINE PUBLICATION | 1
© 2014 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES Box 1 | Barriers for PML development* ■■ Initial infection (early life) ■■ Establishment of a latent, persistent infection (archetype form) ■■ Genetic rearrangements in the promoter or enhancer regions of the virus, promoting the neurotropic form of the virus ■■ Re-expression of the virus from sites of latency, persistent infection, or both ■■ Virus entry into the brain ■■ Establishment of productive infection of glial tissues ■■ Failed CNS immunosurveillance, in particular by JC virus-specific CD8+ lymphocytes *The ubiquitous nature of JC virus, coupled with the extreme rarity of PML, suggests that multiple barriers block disease development. Abbreviations: CNS, central nervous system; PML, progressive multifocal leukoencephalopathy. Permission obtained from Discovery Medicine © Berger, J. R. & Khalili, K. Discov. Med. 12, 495–503 (2011).
In 2006, two cases of PML were reported in patients treated with rituximab for the (unapproved) indication of SLE.13 Rituximab had been well-described in association with PML in the treatment of lymphoproliferative diseases, but the contribution of this drug to the pathogenesis of PML was unclear due to the potential confounders of com bination chemotherapy and the underlying lymphoma. The link between rituximab and PML in both patients with SLE was hard to ascertain, as the presence of other immunosuppressive therapies (as well the underlying diagnosis) made causality dif ficult to assess.11 Nevertheless, after these observations, the FDA required rituxi mab labels to include a black box warning with respect to this possible complication. A similar warning was subsequently issued for mycophenolate mofetil, mainly because of its association with PML in the transplan tation field—an additional concern for rheu matologists who have increasingly used this therapy to treat SLE and other disorders.14,15 Since then, other cases of PML have been reported in association with rituximab and other immunosuppressive therapies. 16,17 One patient with SLE who received the biologic agent belimumab (a human mono clonal antibody that inhibits B‑cell activat ing factor) also developed PML but, again, previous immunosuppressive therapies and SLE indication prevented strong conclusions regarding causality.18
Mechanism of increased PML risk Pathogenesis of PML To properly assess the risk of PML associ ated with a particular therapeutic agent, one
needs to understand the mechanisms that underlie that risk. Unfortunately, the patho genesis of PML is incompletely understood. PML is the consequence of reactivation of a ubiquitous polyomavirus referred to as JC virus, a name derived from the initials of the first person from whom it was isolated (John Cunningham).19 Despite the fact that at least half the adult population worldwide is infected with the causative virus, the rarity of patients with PML clearly points to the exist ence of considerable barriers to the develop ment of the disease (Box 1). JC virus exists almost exclusively in a genetic configuration referred to as the ‘archetype JC virus’, which is incapable of replicating effectively in glial tissues. Therefore, JC virus cannot readily cause PML—only upon genetic rearrange ments, such as deletions and duplications in the noncoding control region, can the virus become capable of growing in glial tissues (prototype form) and induce PML.20 In the central nervous system, the prototype form can complete its full replication sequence to produce new viral particles in oligo dendrocytes, whereas replication in astro cytes is incomplete. The productive infection of oligodendrocytes leads to their death by necrosis, resulting in islands of demyelina tion, spread of the virus to adjacent regions and expansion of the demyelinated region. The means by which JC virus is transmit ted remain unknown. Serological studies indicate that infection occurs before adult hood, although the prevalence of JC virus seropositive individuals increases with every decade thereafter.21,22 The primary phase of infection has no recognizable pheno type, which consequently hinders the study of both the timing and the precise nature of primary JC virus infection. The detection of JC virus in ~40% of harvested tonsils23 suggests that the initial infection might be oropharyngeal. One hypothesis suggests that the virus first infects B cells in tonsillar and parapharyngeal lymph nodes, from where it disseminates to other sites, especially the renal uroepithelium (Figure 1). Whether other sites of viral latency exist (such as the brain) remains controversial. B cells are believed to harbour the virus,24,25 and the unique genetic recombination machinery of these cells is thought to enable conversion from the archetype to the prototype form. Notably, humoral immunity to JC virus does not protect against the development of PML, as high anti-JC virus antibody titres are detected in advanced stages of the disease. However, mechanisms of cellular immunity such as IFN‑γ-producing CD4+ T cells and
2 | ADVANCE ONLINE PUBLICATION
JC virus-specific CD8+ T cells are critical for controlling the disease.26,27
PML triggering Natalizumab and efalizumab Of the drugs that trigger PML, the best studied is the monoclonal antibody natali zumab, used in the treatment of MS and Crohn disease. The survival of patients with natalizumab-associated PML is ~76%.28 Several mechanisms are believed to be involved in natalizumab-induced PML. First, natalizumab administration was dem onstrated to result in the release of prema ture B cells from bone marrow stores. These B cells might be sites of viral latency, and transcription factors associated with their maturation might result in an increase of JC virus replication; importantly, B cells have the appropriate genetic machinery to facilitate generation of mutations in the noncoding control region of the virus, and perhaps elsewhere, potentially resulting in the conversion to the pathogenic prototype strain of JC virus. Secondly, natalizumab, an α4β1 integrin and α4β7 integrin inhibitor, prevents lymphocytes and other inflamma tory cells from binding to vascular cell adhe sion protein 1 (VCAM‑1) and crossing the blood brain barrier, as this type of immune cell migration is dependent on α4β1 integrin. Elimination of the monoclonal antibody by plasmapheresis in natalizumab-associated PML permits entry of JC virus-specific CD8+ T cells into the brain, which almost invariably results in PML immune reconsti tution inflammatory syndrome (PML-IRIS). Efalizumab is a lymphocyte functionassociated antigen 1 (LFA‑1) antagonist used in the treatment of psoriasis, and is believed to cause PML through pathogenic mecha nisms similar to those associated with natali zumab. These two drugs are unique in that latency from the time of their introduction to the time of PML onset is typically meas ured in years: 84% of natalizumab-associated PML occurs after 2 years of treatment, with the average length of treatment being 41 months.28 Only one patient was reported to have developed PML with 1/100 for patients treated with natali zumab with certain quantifiable cofactors, to 3 million patients
Abatacept
RA
No confirmed cases; two PML cases with belatacept (a closely related compound) in patients with solid organ transplantation
Tocilizumab
RA, JIA
No confirmed cases
Anakinra
RA, autoinflammatory disorders
No confirmed cases
Ustekinumab
PsA, psoriasis
No confirmed cases
Tofacitinib
RA
No confirmed cases
Class 1 (high risk)
Class 2 (low risk)
Class 3 (very low risk)
the risks of PML. First, all patients with sys temic autoimmune diseases are at some risk of PML; in particular, this risk appears to be highest for patients with SLE.7,32 Second, when it comes to selection of an immuno suppressive regimen utilizing drugs from Class 2 or Class 3, risks should be graded as very low regardless of the therapy. The Council for International Organizations of Medical Sciences35 defines an event as very rare when
Sorting out the risks in progressive multifocal leukoencephalopathy Leonard H. Calabrese, Eamonn Molloy and Joseph Berger Abstract | Progressive multifocal leukoencephalopathy (PML) is a rare, opportunistic infection of the central nervous system, caused by reactivation of the ubiquitous JC virus. PML is a devastating disease that is frequently fatal, and although survival rates have improved, patients who survive PML often experience considerable neurological deficits. PML was associated with a variety of immunosuppressive therapies in the past decade, but attribution of causality is difficult owing to the presence of confounding factors and to an inadequate understanding of the underlying pathogenesis of this disease. This uncertainty has hindered efforts for shared decisionmaking between physicians and their patients and, in some cases, discouraged the use of potentially beneficial therapies. We propose a categorization of immunosuppressive agents according to their risk of PML to support a better-informed decision-making process when evaluating the risks and benefits of these therapies. Calabrese, L. H. et al. Nat. Rev. Rheumatol. advance online publication 14 October 2014; doi:10.1038/nrrheum.2014.167
Introduction
Progressive multifocal leukoencephalo pathy (PML) is a devastating, and often fatal, opportunistic infection of the central nervous system. The diagnosis and clinical features of PML, the discussion of which is beyond the scope of this paper, were con cisely reviewed and formulated in a consen sus statement published in 2013. 1 The epidemiology of PML can be best under stood if divided in three epochs of disease: the first period extends from its initial des cription in the early 1980s, when it was predominantly associated with lympho proliferative diseases; the second coincided Competing interests L.H.C. has served as consultant to Genentech, GlaxoSmithKline and Pfizer in the area of progressive multifocal leukoencephalopathy (PML). E.M. has obtained research support from Roche, has served on advisory boards for BristolMyers Squibb, Pfizer and MSD, and received payment for developing educational materials relating to PML from GlaxoSmithKline. J.B. serves on the PML Adjudication Committees of Amgen, Astra Zeneca, Bristol-Myers Squibb, Eisai, Janssen, Millennium, Parexel, Pfizer, Roche and Takeda; he is a consultant to Genentech, Genzyme, Incyte, Inhibikase Therapeutics, Johnson & Johnson and Novartis, and has received grants from the PML Consortium, Biogen Idec and Novartis.
with the AIDS pandemic, when PML became strikingly prevalent as it affected 4–5% of all patients with AIDS;2 after this period, the epidemiology of PML changed dramati cally with the initial diagnosis of the disease in patients with multiple sclerosis (MS) and Crohn disease who were treated with the biologic agent natalizumab.3–5 No evidence for association of these two autoimmune diseases with PML was observed before the introduction of natalizumab. Subsequently, and over the past decade, the notion of druginduced PML was introduced in the field of rheumatology based on rare reports of the disease associated with the use of immuno suppressive therapies in the context of a variety of rheumatic disorders.6,7 Concerns over PML have been amplified by the knowledge that the condition, while remaining rare, is unpredictable, often fatal and has currently no effective treatment. The disease has a complex biology 8 and, despite advances in clinical and basic research, sig nificant deficits exist in our understanding of its pathogenesis. Because PML is a relatively new disorder for the rheumatology com munity, clinicians might lack confidence in assessing the risks of JC virus opportunis tic infection in their patients. Thus, both clinicians and patients must evaluate the
NATURE REVIEWS | RHEUMATOLOGY
risks of PML by taking into account factors such as the underlying disease, the associ ated immunosuppressive regimen and the level of risk patients are willing to accept. Although PML is associated with a variety of immunosuppressive therapies, several of which carry ‘black box’ warnings with res pect to this potential complication, immuno suppressive agents should not be grouped together, because the risk for PML varies greatly among them. Recognition of this fact is essential for informed, shared, therapeutic decision-making.
PML risk of immunosuppression
Prior to 2006, PML had not been diagnosed in patients with a rheumatic disease treated with a biologic agent. However, by that time, 35 patients with PML who had received standard immunosuppressive agents such as azathioprine, chlorambucil, cyclophos phamide, glucocorticoids and methotrexate, used alone or more commonly in com bination, were identified.6 Most reported cases occurred in patients with systemic lupus erythematosus (SLE) who were often exposed to minimal iatrogenic immuno suppression; importantly, evidence suggests that SLE itself predisposes to PML.9,10 The contribution of therapeutic agents to the risk of PML was not widely appreciated until 2005, when three cases of PML were associated with natalizumab, a biologic ther apy under investigation for the treatment of MS3,4 and Crohn disease5 at the time. Soon thereafter, PML was diagnosed in three other patients with psoriasis who were treated with efalizumab.11 The profound importance of these cases was emphasized by the absence of prior reports of PML associated with MS, inflammatory bowel disease (IBD) or psori asis. Ultimately, efalizumab was voluntarily withdrawn from the market. Natalizumab was temporarily withdrawn but later rein troduced for the treatment of both MS and Crohn disease under a rigorous risk mitiga tion plan, the TOUCH (TYSABRI Outreach: Unified Commitment to Health) prescribing programme.12 TOUCH mandates the educa tion of healthcare providers and patients as to the risks of natalizumab therapy, as well as close monitoring of patients that receive natalizumab infusions for warning signs of PML.12 ADVANCE ONLINE PUBLICATION | 1
© 2014 Macmillan Publishers Limited. All rights reserved
PERSPECTIVES Box 1 | Barriers for PML development* ■■ Initial infection (early life) ■■ Establishment of a latent, persistent infection (archetype form) ■■ Genetic rearrangements in the promoter or enhancer regions of the virus, promoting the neurotropic form of the virus ■■ Re-expression of the virus from sites of latency, persistent infection, or both ■■ Virus entry into the brain ■■ Establishment of productive infection of glial tissues ■■ Failed CNS immunosurveillance, in particular by JC virus-specific CD8+ lymphocytes *The ubiquitous nature of JC virus, coupled with the extreme rarity of PML, suggests that multiple barriers block disease development. Abbreviations: CNS, central nervous system; PML, progressive multifocal leukoencephalopathy. Permission obtained from Discovery Medicine © Berger, J. R. & Khalili, K. Discov. Med. 12, 495–503 (2011).
In 2006, two cases of PML were reported in patients treated with rituximab for the (unapproved) indication of SLE.13 Rituximab had been well-described in association with PML in the treatment of lymphoproliferative diseases, but the contribution of this drug to the pathogenesis of PML was unclear due to the potential confounders of com bination chemotherapy and the underlying lymphoma. The link between rituximab and PML in both patients with SLE was hard to ascertain, as the presence of other immunosuppressive therapies (as well the underlying diagnosis) made causality dif ficult to assess.11 Nevertheless, after these observations, the FDA required rituxi mab labels to include a black box warning with respect to this possible complication. A similar warning was subsequently issued for mycophenolate mofetil, mainly because of its association with PML in the transplan tation field—an additional concern for rheu matologists who have increasingly used this therapy to treat SLE and other disorders.14,15 Since then, other cases of PML have been reported in association with rituximab and other immunosuppressive therapies. 16,17 One patient with SLE who received the biologic agent belimumab (a human mono clonal antibody that inhibits B‑cell activat ing factor) also developed PML but, again, previous immunosuppressive therapies and SLE indication prevented strong conclusions regarding causality.18
Mechanism of increased PML risk Pathogenesis of PML To properly assess the risk of PML associ ated with a particular therapeutic agent, one
needs to understand the mechanisms that underlie that risk. Unfortunately, the patho genesis of PML is incompletely understood. PML is the consequence of reactivation of a ubiquitous polyomavirus referred to as JC virus, a name derived from the initials of the first person from whom it was isolated (John Cunningham).19 Despite the fact that at least half the adult population worldwide is infected with the causative virus, the rarity of patients with PML clearly points to the exist ence of considerable barriers to the develop ment of the disease (Box 1). JC virus exists almost exclusively in a genetic configuration referred to as the ‘archetype JC virus’, which is incapable of replicating effectively in glial tissues. Therefore, JC virus cannot readily cause PML—only upon genetic rearrange ments, such as deletions and duplications in the noncoding control region, can the virus become capable of growing in glial tissues (prototype form) and induce PML.20 In the central nervous system, the prototype form can complete its full replication sequence to produce new viral particles in oligo dendrocytes, whereas replication in astro cytes is incomplete. The productive infection of oligodendrocytes leads to their death by necrosis, resulting in islands of demyelina tion, spread of the virus to adjacent regions and expansion of the demyelinated region. The means by which JC virus is transmit ted remain unknown. Serological studies indicate that infection occurs before adult hood, although the prevalence of JC virus seropositive individuals increases with every decade thereafter.21,22 The primary phase of infection has no recognizable pheno type, which consequently hinders the study of both the timing and the precise nature of primary JC virus infection. The detection of JC virus in ~40% of harvested tonsils23 suggests that the initial infection might be oropharyngeal. One hypothesis suggests that the virus first infects B cells in tonsillar and parapharyngeal lymph nodes, from where it disseminates to other sites, especially the renal uroepithelium (Figure 1). Whether other sites of viral latency exist (such as the brain) remains controversial. B cells are believed to harbour the virus,24,25 and the unique genetic recombination machinery of these cells is thought to enable conversion from the archetype to the prototype form. Notably, humoral immunity to JC virus does not protect against the development of PML, as high anti-JC virus antibody titres are detected in advanced stages of the disease. However, mechanisms of cellular immunity such as IFN‑γ-producing CD4+ T cells and
2 | ADVANCE ONLINE PUBLICATION
JC virus-specific CD8+ T cells are critical for controlling the disease.26,27
PML triggering Natalizumab and efalizumab Of the drugs that trigger PML, the best studied is the monoclonal antibody natali zumab, used in the treatment of MS and Crohn disease. The survival of patients with natalizumab-associated PML is ~76%.28 Several mechanisms are believed to be involved in natalizumab-induced PML. First, natalizumab administration was dem onstrated to result in the release of prema ture B cells from bone marrow stores. These B cells might be sites of viral latency, and transcription factors associated with their maturation might result in an increase of JC virus replication; importantly, B cells have the appropriate genetic machinery to facilitate generation of mutations in the noncoding control region of the virus, and perhaps elsewhere, potentially resulting in the conversion to the pathogenic prototype strain of JC virus. Secondly, natalizumab, an α4β1 integrin and α4β7 integrin inhibitor, prevents lymphocytes and other inflamma tory cells from binding to vascular cell adhe sion protein 1 (VCAM‑1) and crossing the blood brain barrier, as this type of immune cell migration is dependent on α4β1 integrin. Elimination of the monoclonal antibody by plasmapheresis in natalizumab-associated PML permits entry of JC virus-specific CD8+ T cells into the brain, which almost invariably results in PML immune reconsti tution inflammatory syndrome (PML-IRIS). Efalizumab is a lymphocyte functionassociated antigen 1 (LFA‑1) antagonist used in the treatment of psoriasis, and is believed to cause PML through pathogenic mecha nisms similar to those associated with natali zumab. These two drugs are unique in that latency from the time of their introduction to the time of PML onset is typically meas ured in years: 84% of natalizumab-associated PML occurs after 2 years of treatment, with the average length of treatment being 41 months.28 Only one patient was reported to have developed PML with 1/100 for patients treated with natali zumab with certain quantifiable cofactors, to 3 million patients
Abatacept
RA
No confirmed cases; two PML cases with belatacept (a closely related compound) in patients with solid organ transplantation
Tocilizumab
RA, JIA
No confirmed cases
Anakinra
RA, autoinflammatory disorders
No confirmed cases
Ustekinumab
PsA, psoriasis
No confirmed cases
Tofacitinib
RA
No confirmed cases
Class 1 (high risk)
Class 2 (low risk)
Class 3 (very low risk)
the risks of PML. First, all patients with sys temic autoimmune diseases are at some risk of PML; in particular, this risk appears to be highest for patients with SLE.7,32 Second, when it comes to selection of an immuno suppressive regimen utilizing drugs from Class 2 or Class 3, risks should be graded as very low regardless of the therapy. The Council for International Organizations of Medical Sciences35 defines an event as very rare when
