Lumpers, Splitters, and the Move Toward New Taxonomy for Rheumatology

Historians attribute Philip II of Macedon — father of Alexander the Great — as the source of the motto and political principle known as “divide and conquer.” As a ruling technique, the maxim worked, with Phillip ultimately ruling over most of classical Greece; a success that would be repeated with Julius Caesar and Napoleon Bonaparte in the centuries to follow.

Contrary to this historical paradigm, rheumatologists have long been attempting to conquer complex diseases like osteoarthritis, vasculitis and the spondyloarthropathies by lumping them together to be studied and treated as a whole. However, as diagnostic technology has advanced and more targeted therapies have emerged, there has been a shift toward dividing these diseases based on molecular information to clinical phenotype. This ideological fissure is known as the lumper/splitter debate — and there is little doubt on which side Philip of Macedonia might have argued.

While the particulars of the disease states are vastly different, Peter Grayson, MD, MSc, head of the Vasculitis Translational Research Program at the NIH, noted general trends in the discussion. “From a clinical trial standpoint and from a clinical practice standpoint, we lump,” he said. “On the flip side, the biggest point about splitting has to do with the movement toward precision medicine.”

Grayson, who spoke on both sides of the debate in vasculitis, believes that rheumatology should follow the trail blazed by oncology: Using biomarkers, molecular signatures and other personalized information to treat each patient and each tumor individually. “It is going to be a challenge to reach this goal, but a good place to start is more splitting in clinical trials,” he said.

At the crux of the debate is the necessity to categorize rheumatic diseases, with each side assessing the disease based on their own ideologic stance. For example, lumpers in OA tend to view the disease as a mechanical problem; splitters might look at knee/hip OA and hand OA as separate entities, or view post-traumatic OA as something apart from inflammatory OA.

In vasculitis, there is movement away from the longstanding paradigm of grouping vasculitides as large-, medium-, and small-vessel disease; specifically, debate continues about whether giant cell arteritis (GCA) and Takayasu’s arteritis represent a spectrum of the same disease or different diseases. It may sound like a small thing, but the ever-shifting semantics reflect the clinical realities that confound researchers and clinicians.

With the lumper/splitter debate continuing to play out in peer-reviewed journals and on the dais of society meetings around the globe, an examination of lumping and splitting in OA, the vasculitides, and the spondyloarthritides may point the way toward the future of the specialty.

Lumping OA

During a raucous debate at the 2019 Osteoarthritis Research Society International World Congress, Tonia Vincent, MD, PhD, professor of musculoskeletal biology at the Kennedy Institute of Rheumatology at Oxford University, argued in favor of the lumper ideology against Frank Beier, PhD, Canada Research Chair in Musculoskeletal Health at the Schulich School of Medicine & Dentistry at Western University, London, Ontario, Canada.

“In any disease, understanding its pathogenesis is central to target discovery,” Vincent said in a follow-up interview with Healio Rheumatology. “If we can target the central pathway or pathogenic process, then we stand to impact all individuals with the disease rather than a few.”

Vincent takes it as a given that abnormal mechanical load is the common etiological factor in OA. “Mechanically-induced pathways are likely to reveal those important targets,” she said. “Factors such as inflammation and age, among others, impact those mechanical signaling pathways and, in this way, contribute to outcomes. But the role of those factors is likely to be as a modifier rather than a principal target.”

Moreover, most of these factors are continuous variables, so it is not logical to try to select, for example, older patients, or those with inflammation, and treat them differently, according to Vincent.

The OARSI presentation was framed as a breakdown of the generally-accepted endotypes of OA. During her argument, Vincent highlighted why metabolic, post-traumatic, inflammatory and site-specific disease should remain together.

“The epidemiological evidence very clearly points to an increased risk of OA in metabolic syndrome, which includes raised lipids, obesity, diabetes and high blood pressure,” she said. “But once you control for the obesity — a known independent risk factor in OA — this association with metabolic syndrome is lost. Therefore, there does not appear to be any value in selecting these patients for special treatment.”

Vincent stressed that weight loss is the exception, but pointed out that weight loss is already regarded as part of the standard management strategy for OA.

Regarding post-traumatic OA, Vincent said that other than the speed and age of onset, there is no evidence that post-traumatic OA is any different to age-related disease at the level of the macroscopic, microscopic and patient-related symptoms.

“This accords well with our notion that OA is all mechanically-driven but that the abnormal load may be due to different things,” she said. “I favor the simple view that abnormal mechanical load can either be due to increased load traversing a normal joint — due to such factors as obesity, joint malalignment/abnormal joint shape, occupational risk — or normal load traversing a joint that has lost its mechanoprotective mechanisms, which could be from acute destabilizing injuries or loss of muscle support and gait reflexes with age.”

Inflammation is generally easy to spot, with swollen joints or synovitis, according to Vincent. “Moreover, MRI studies show that pretty much all patients have evidence of low-grade synovitis when you look hard enough for it using contrast imaging — this means that inflammation cannot be regarded as a distinct endotype.”

There also exists some controversy regarding the role of inflammation in disease. “One shouldn’t assume that evidence of synovitis necessarily means OA is driven by inflammation, either at the level of symptoms such as pain and stiffness, or structural damage like changes on an X-ray,” Vincent said. “Even though many groups have used classical inflammatory cytokines such as IL-1 and TNF to demonstrate cartilage breakdown in vitro, their role in disease has been definitively disproved by several negative randomized control trials targeting IL-1 and TNF. Of course, ‘inflammation’ is more nuanced than necessarily only referring to cytokines or activated immune cells.”

Indeed, Vincent and colleagues have shown that simple mechanical injury activates inflammatory signaling pathways in chondrocytes. “This indicates, again, that mechanical injury is sufficient to drive disease pathogenesis in the absence of soluble or cellular mediators of the immune system,” she said.

One of the core arguments surrounding site-specific OA is that the knees and hips bear weight, and the hands do not and are therefore not under the same mechanical stress. “The finger joints are subjected to significant load from associated muscle action, as indeed are weight-bearing joints, so the same mechanosensitive principal applies to hands as it does to other large joints,” Vincent said.

Far from being a lumping hard-liner, Vincent acknowledged there is potential utility of splitting the disease. “It is not that I don’t think there might be distinct molecular endotypes of OA that could have clinical utility,” Vincent noted during her opening remarks at OARSI. “I simply do not believe we have evidence of these endotypes in OA currently.”

Splitting OA

Like Vincent, Beier also aimed to dig into the pathogenesis of OA — he just happens to think that there are multiple pathogeneses for the disease, rather than the single etiological factor of mechanical load. “It has been known for a long time that structural features of OA, especially as defined by X-ray and symptoms such as pain, do not always occur together,” he said. “Some patients show both, but others show structural joint pathology without feeling pain, while a third group might experience severe pain without major X-ray abnormalities. These differences point towards differences in underlying mechanisms.”

A growing number of clinicians and researchers are raising their voices and publishing data in support of diversity in the underlying mechanisms of OA. In their study published in Nature Reviews Disease Primers, Martel-Pelletier and colleagues acknowledged, at the very least, that “the concept of the pathophysiology is still evolving, from being viewed as cartilage-limited to a multifactorial disease that affects the whole joint.”

Some of the myriad factors associated with OA occurrence and progression include injury, aging, obesity, overuse, genetic factors, female sex and malalignment of joints, according to Beier. “While they can all lead to a similar disease phenotype at the end stage, and potentially similar mediators at late stages, including biomechanics, the original trigger is different in each case,” he said. “In order to understand, and potentially treat OA, we need to understand the entire disease progress from the earliest stages and start intervention as early as possible.”

Martel-Pelletier and colleagues noted that it is also critical to parse out the local and systemic factors that may play a role in clinical and structural presentation of OA. Better understanding of these factors may demonstrate where the pathways begin and how and when they converge.

Beier noted that even a clearer perception of disease progression could reveal differences in the underlying mechanism. “Some people show rapid disease progression, such as patients who develop OA after injury, while it takes much longer in others,” he said. “Again, this points towards differences in the underlying mechanisms.”

The crux of Beier’s argument was that if the underlying mechanisms of disease development and progression are different, this could alter the way it is treated. “We might need different treatments for different forms,” he said.

Regarding treatment, Martel-Pelletier and colleagues stressed that pharmacologic therapies, at the moment, only treat symptoms. A clearer picture of the underlying mechanisms could lead to more targeted therapies.

In a recent review paper published in Current Opinions in Rheumatology, Warner and Valdes raised the question of the utility of genome-wide association studies in OA. They suggested that there may be 21 independent susceptibility loci for OA overall, while four distinct loci have been identified in a genome-wide association scan on minimum joint space width.

Beier thinks this is a low-ball estimate. “Genome-wide association studies have identified approximately 100 genetic loci that are linked to the risk of OA,” Beier said. “These loci are linked to genes with very different functions and in very different pathways.”

Research begets research. In this case, the genome-wide studies have begun to steer researchers in the direction of the next generation of studies, according to Beier. Some of these data sets have shown various growth factor pathways that may be involved in the pathogenesis of OA, along with extracellular matrix genes and transcription factors. “This means that deregulation of many different pathways can predispose to OA, which provides clear evidence for high molecular variability within the spectrum of OA,” he said.

Despite these advances, studies have yet to clearly link specific molecular pathways to specific OA phenotypes or subtypes, according to Beier. However, he outlined steps toward reaching this goal.

“Better characterization of patient phenotypes and variability — and, in parallel, better description of molecular changes in various forms of OA — plus better linkage of these two types of information, will be required to establish the existence of OA phenotypes,” Beier said. “This will also tell us whether different OA phenotypes will benefit from different therapeutic strategies. I am not sure if Dr. Vincent agrees on that point.”

For the record, she does.

Lumping in Vasculitis

Although Eric J. Gapud, MD, PhD, physician scientist and director of Research for the Vasculitis Center in the division of rheumatology at Johns Hopkins University, is a splitter, he also acknowledges the utility of classifying by size as part of the clinical workup.

“With vasculitis, at least initially, you need some big categories to start with in order to triage your workup to figure out where you are going in terms of large, medium or small vessel disease,” he told Healio Rheumatology. “But those of us who spend the majority of time in the laboratory know that on the cellular and molecular level, there is no doubt that there are different disease processes.”

In general, the advantage to lumping when dealing with a family of rare diseases like vasculitis is that it enables more efficient recruitment into clinical trials, noted Grayson. “It allows us to overcome obstacles to studying an extremely rare condition,” he said. “Most of the landmark clinical trials in vasculitis were lumper trials. This was done to enable sufficient numbers to conduct these studies so they were powered for meaningful results.”

In something of a microcosm of the lumper/splitter debate in vasculitis, Grayson dug deep into both the similarities and differences between GCA and Takayasu’s arteritis in a 2015 paper published in the Journal of Rheumatology. He noted common genetic variants in the FCGR2A/FCGR3A pathway, along with similarities in the distribution of arterial lesions. “Lesions tended to be contiguous in the aorta and symmetric in branch vessels in both diseases,” he wrote. “The distribution of affected vessels was similar with few exceptions.”

Looking at diseases together in this way has allowed researchers to extrapolate what is known about more common forms of vasculitis and put it into clinical practice for rarer forms of the disease, according to Grayson. “From these lumper trials, we have seen the overlapping or related areas that allowed us to repurpose drugs,” he said.

However, the lumper trials have only brought the vasculitis community so far. “When you look at the subanalyses, those trials have not been sufficiently powered to detect subtleties in different forms of vasculitis,” Grayson said.

Another consideration is the patient experience, he noted. “Patients tend to view these diseases a bit differently than physicians,” Grayson said. “The things they value in terms of how they feel are often different from what we value in terms of response to drugs.”

A key example of this is fatigue, which unifies patients across vasculitis types. “Fatigue is a very important symptom to patients, and it is present in small-, medium- and large-vessel vasculitis,” Grayson said. “As physicians, we often do not consider fatigue as part of active disease, and therefore we do not do a good job of dealing with it.”

Splitting in Vasculitis

Shifting to the splitter side, Grayson said that updating classification criteria as new data emerge is an important step in understanding the differences between diseases. His 2015 paper demonstrated that age is used to classify both GCA and Takayasu’s arteritis, but he suggested that its use is arbitrary. “Involvement of the aorta and primary branches with disease onset at < 40 years is classified as [Takayasu’s arteritis], and large vessel involvement with disease onset at > 50 years is often classified as GCA,” he wrote.

Grayson highlighted other potential ways of differentiating this disease grouping. “There was notably increased axillary artery involvement in GCA, and a high prevalence of left subclavian artery involvement was seen in [Takayasu’s arteritis],” he wrote. “We suggested that [Takayasu’s arteritis] and GCA might exist on the same spectrum of disease with subtle skewing of disease phenotypes influenced by immunosenescence.”

Moving back to the clinical realm, Grayson proposed that the current medical landscape of precision medicine has significantly shifted expectations for both doctors and patients. “In many ways, vasculitis is viewed as a chronic condition now,” he said. “We as physicians expect to put patients into remission with currently available drugs; the patients expect this kind of individualized treatment as well.”

A more thorough understanding of genetic information may help reach this goal. In their study published in Journal of the American College of Cardiology, Zhang and colleagues investigated the role of CD28 signaling in both medium and large vessel vasculitis. They suggested that in GCA, CD4 T-cells and macrophages infiltrate vessel walls and form tissue-destructive granulomatous infiltrates. The artery mounts a “maladaptive response” that can lead to a number of outcomes, including intramural neoangiogenesis, intimal hyperplasia and luminal occlusion. Blocking CD28 may lead to targeting of glycolytic activity in T-cell populations that reside in vessel walls, according to the researchers.

“As we become more advanced with therapeutics, we are gravitating toward splitting in our study designs,” Gapud said. He acknowledged, however, that there may come a point when clinical trials are over-splitting. “We are not there yet, but we are at a point where the ability to split reflects the progress we have made in understanding our patient populations.”

For example, biomarkers could prove a fruitful place to further split patients in clinical trials to learn whether more subgroups exist. “In one-third of Takayasu cases, we do not see ESR or CRP elevation,” Gapud said. “However, in GCA, we do not accept the diagnosis without ESR/CRP elevation.”

As the field transitions from using these blood tests to more sophisticated tools like molecular signatures, thinking beyond individual disease states may also have utility, Grayson added. “There may be patients with vasculitis who have certain molecular signatures that look like a subset of patients with lupus,” he said. “We have to allow the science and the biologic pathways to let us look beyond traditional diagnosis patterns.”

Whether any of these approaches are ready for prime time remains to be seen, according to Grayson. “We have to ask ourselves whether much of this is technically feasible yet,” he said. “There are many mountains to climb with regard to the computational ability and the cost of doing this kind of work.”

In the meantime, there are still simpler ways of understanding the differences between diseases. “The prevalence of large vessel disease in patients with GCA can vary widely depending upon the imaging modality used to screen for disease (angiography 20% to 30%, positron emission tomography 30% to 80%, ultrasound 30%),” Grayson wrote.

Clinicians can also consider real-world results when designing the next generation of clinical trials, according to Gapud. “Patients with GCA and large-vessel disease often do well on tocilizumab (Actemra, Genentech),” Gapud said. “But we look to other treatment regimens for Takayasu because the diseases have different pathogenic mechanisms.”

For Grayson, ultimately, the paradigm shift will come when patients are no longer lumped together based on clinical symptoms or diagnosis, but on more sophisticated information. “As we gain further understanding of, say, specific interferon or cytokine signatures in the blood, we will redefine how we lump or split patients,” he said.

Lumping the Spondyloarthropathies

Philip J. Mease, MD, from the Swedish Medical Center and the University of Washington in Seattle, at various points in his career, has argued both sides of the debates in spondyloarthropathy.

“When it comes to clinicians making a diagnosis, there is value in lumping,” Mease said. “There are certain key patterns or features that unite many of the spondyloarthritis conditions.”

This is particularly true for a very practical reason: The ever-increasing number of nonrheumatologists who must try to diagnose rheumatic diseases due to the rheumatology workforce shortage.

“A family practitioner probably has a hard time understanding the differences between rheumatoid arthritis and spondyloarthritis, and they may be unlikely to be able to differentiate between PsA and axial spondyloarthritis,” he said. “What we need to do as a rheumatology community is explain that there are key symptom areas that shunt me — or the diagnosing clinician — in the direction of rheumatoid arthritis, lupus or the spondyloarthritides. It is helpful to have those big buckets.”

As a diagnosing clinician who specializes in the spondyloarthritis area, Mease looks for inflammatory or immunologic back pain to distinguish how much of the symptomatology may be due to a spondyloarthritis condition, either wholly or in combination with degenerative arthritis and/or fibromyalgia. “If it is inflammatory, I will start thinking along the spondyloarthritides line,” he said. “Then I will start asking questions about whether there is evidence of lower-extremity asymmetric arthritis, enthesitis, whether the Achilles tendon or plantar fascia is involved, whether there is uveitis or inflammatory bowel disease. I want to know about skin involvement and psoriasis. All of these symptoms and conditions have association with the spondyloarthritides.”

Mease noted that the diseases under this umbrella are generally equal in terms of gender, whereas RA and lupus tend to occur more frequently in women.

Once he has moved along this diagnostic pathway, Mease acknowledged that the treatment paradigms are similar for many of the spondyloarthropathies.

Splitting the Spondyloarthropathies

Reclassifying disease states is also happening in the spondyloarthropathies. An important step was taken in 2011, when Philipose and Deodhar described the updated Assessment of SpondyloArthritis International Society (ASAS) classification criteria for axial SpA and peripheral SpA.

The previous criteria did not specifically differentiate between them. Key research showed that the HLA-B27 gene is critical in differentiating between axial and peripheral SpA. In addition, the updated criteria allow for the use of MRI, as opposed to conventional X-ray, in detecting sacroiliitis with or without structural damage, enabling clinicians to make an earlier diagnosis.

However, among rheumatology specialists, capturing granular data to detail the features and phenotypes of disease is essential, according to Mease.

One way to accomplish this is to ensure that measures of disease are as specific as possible. “Looking at PsA, we have to have composite criteria that capture peripheral joint and skin involvement, as opposed to using ACR or DAS28 response, which are purely focused on arthritis,” he said. “Enthesitis is a more prominent issue in PsA than axial spondyloarthritis. We need to be looking at all of these domains, and using PsA criteria to determine PsA and axial spondyloarthritis criteria to determine axial spondyloarthritis.”

Understanding and differentiating disease states can then lead to more targeted therapies, Mease added. “IL-6 inhibitors work well in RA, but have not performed well in the spondyloarthritis conditions,” he said. “IL-6 completely failed in axial spondyloarthritis and was ho-hum in PsA.”

Conversely, drugs in the IL-17 and IL-23 pathways have been “fantastic” in psoriasis, according to Mease. “They have also been strong in axial spondyloarthritis,” he said. “We are getting more precise. There are a lot of good people looking at how these treatments work in each disease.”

Perhaps the final question to consider is how “big pharma” reacts to the lumper/splitter debate in the context of conducting clinical trials, according to Mease, who began talking about splitting patients as far back as a meeting in Berlin in 2000.

Since then, the gradual shift toward splitting in the conduct of clinical trials has occurred. That said, Mease is certain that there is much more work to be done.

“When I fast-forward 20 years, I think we will be looking back on this time as one of very crude differentiation,” he said. “We will have specific SNP biomarkers and epigenetic information that we can combine with pain scores and presence or absence of factors like IBD or uveitis that will tell us exactly what this patient needs in terms of therapy.” – by Rob Volansky

• References:
• Braun J, Sieper J. Z Rheumatol. 2010;doi:10.1007/s00393-009-0591-7.
• Grayson PG. J Rheumatol. 2015;doi:10.3899/jrheum.141376.
• Martel-Pelletier J, et al. Nat Rev Dis Primers. 2016;doi:10.1038/nrdp.2016.72.
• Philipose J, Deodhar A. J Musculoskel Med. 2011;28:454-457.
• Warner SC, Valdes AM. Curr Opin Rheumatol. 2017;doi:10.1097/BOR.0000000000000352.
• Zhang H, et al. J Am Coll Cardiol. 2019;doi: 10.1016/j.jacc.2019.01.049.

• For more information:
• Frank Beier, PhD, can be reached at Medical Sciences Building, Room 216, London, Ontario, Canada, N6A 5C1; email: fbeier@uwo.ca.
• Eric Jonas Gapud, MD, PhD, can be reached at 5200 Eastern Ave., Baltimore, MD 21224; email: egapud1@jhmi.edu.
• Peter Grayson, MD, MSc, can be reached at 9000 Rockville Pike, Bethesda, MD 20892; email: peter.grayson@nih.gov.
• Philip J. Mease, MD, can be reached at 601 Broadway, Seattle, WA 98122; email: pmease@philipmease.com.
• Tonia Vincent, MD, PhD, can be reached at Roosevelt Drive, Headington, Oxford, OX3 7FY, UK; email: tonia.vincent@kennedy.ox.ac.uk.

Disclosure: Beier, Gapud and Grayson report no relevant financial disclosures. Mease reports grants, consulting fees and/or speaking fees from AbbVie, Amgen, Boehringer Ingelheim, Bristol Myers Squibb, Celgene, Eli Lilly, Galapagos, Genentech, Gilead, Janssen, Novartis, Pfizer, SUN and UCB.