PROCEEDINGS OF THE 10TH ANNUAL SPARTAN MEETING
Ankylosing Spondylitis Clinical Registries: Principles, Practices
and Possibilities
Liron Caplan, MD, PhD, Daniel O. Clegg, MD and Robert D. Inman, MD
Abstract: The need for a rigorously developed longitudinal registry of
patients with spondyloarthritis (SpA) is clear and urgent. Like randomized
controlled trials, registries rely on a prospective, systematic protocol-
driven approach to data acquisition to assess outcomes for a prescribed
cohort of patients. Registries seek to capture large numbers of patients
across large geographic zones and can serve as a valuable resource for
patient advocacy, patient education and support, incidence and preva-
lence, and broad demographic profiles. Building on 3 existing registries—
the Prospective Study of Outcomes in Ankylosing Spondylitis, the
Program to Understand the Longterm Outcomes of Spondyloarthritis
(PULSAR) and the University Health Network Spondyloarthritis
Program—these registries and the Spondylitis Association of America
propose to form a combined registry of North American SpA patients.
The combined registry would, ideally, complement ongoing clinical
goals and improve patient care.
Key Indexing Terms: Clinical registry; Spondyloathritis; Ankylosing
spondylitis. [Am J Med Sci 2013;345(6):437–439.]
PRINCIPLES OF REGISTRY DEVELOPMENT
clinical registry has been defined as “an organized system
A that uses observational study methods to collect uniform
data (clinical and other) to evaluate specified outcomes for a pop-
ulation defined by a particular disease, condition or exposure
and that serves a predetermined scientific, clinical or policy pur-
pose(s).”1 Patterned after the design of randomized controlled
trial (RCT), both study types rely on a prospective, systematic
protocol-driven approach to data acquisition to assess outcomes
for a prescribed cohort. Both approaches bear their own
strengths: Whereas RCTs employ randomization in an effort to
distribute characteristics evenly between control and intervention
groups (and thus minimize bias), registries generally assume less
restrictive inclusion criteria that allow investigators to examine
phenomenon in a more “real-world” environment. In essence,
registries attempt to apply the rigorous techniques adopted by
RCTs to observational environments in which therapies are not
randomized or necessarily determined a priori.
As such, clinical registries may fulfill a number of
important functions to (1) identify risk factors for health
outcomes; (2) identify individuals or groups at risk of health
outcomes; (3) compare treatment benefits (ie, comparative
effectiveness research); (4) characterize a disorder/disease
course (incidence/prevalence, survival, temporal changes); (5)
quantify the safety or harm associated with a pharmacological
treatment or device; (6) assess the quality of care delivered by
From the Denver Veterans Affairs Medical Center, University of Colorado
(LC), Denver, Colorado; George E. Whalen Salt Lake Veterans Affairs Medical
Center, University of Utah School of Medicine (DOC), Salt Lake City, Utah; and
University of Toronto (RDI), Toronto, Ontario, Canada.
Presented at the 10th annual SPARTAN-GRAPPA Meeting, Portland,
Oregon, July 27, 2012.
The authors have no financial or other conflicts of interest to disclose.
Correspondence: Daniel O. Clegg, MD, University of Utah Health Care,
50 North Medical Drive, Salt Lake City, UT 84132 (E-mail: daniel.clegg@
hsc.utah.edu).
The American Journal of the Medical Sciences  Volume 345, Number 6, June 2013
a provider or institution and (7) determine the cost-effectiveness
of a treatment or compare treatment costs.1 To realize these
goals, a registry should ideally address a number of important
logistical issues, as outlined below.
In the planning phase, registry developers should devise
a clear mission statement and general purpose of the registry.
This should be done by recruitment of stakeholders, who will
also define the disease to be studied, arrange initial funding
support or craft a funding strategy, construct the formal registry
leadership group and establish a governance structure. The
governing entity or entities will provide regulatory, ethical and
scientific oversight. Importantly, all decisions should be
redacted into a manual of operating procedures (MOP), which
remains current and accessible to all affiliated parties.
The MOP typically includes a description of the principal
research foci of the registry, with sufficient detail to dissuade an
overly ambitious approach to data collection. The MOP should
specify the study design (eg, whether there is nested data
collection), inclusion/exclusion criteria and the rules documenting
how exposures and outcomes are quantified. The data collection
methods and sources of data should be explicitly mentioned, with
a precise and reproducible protocol describing the procedures for
assembly of the data and/or specimens. Preferably, these methods
should anticipate areas of likely bias and inconsistency and
develop schemes for ameliorating such problems. Some MOP
will establish a target sample size or the timetable necessary to
answer the major scientific questions posed by the registry,
whereas others adopt a more open-ended approach.
Importantly, the registry design should reflect the
priorities for the registry, with the amount of attention in the
MOP reflecting the relative significance of individual registry
components in furthering these priorities. The reliability,
burden (in terms of both cost and time) and importance of
individual data elements will drive the crucial decision whether
to designate data elements as critical or complementary. Newly
initiated registries are often undermined by a research agenda
that aspires toward unrealistic goals. Limiting the initial number
of fields guards against this tendency.
A number of principles help to optimize a registry’s utility.
For example, employing only validated measures, when they are
available, increases the likelihood of conducting good science.
Given the broader cohort diversity observed in clinical registries
(compared with the restrictive populations of RCTs), inclusion of
comorbidity measurement also often represents a wise use of
resources. These comorbidity indices may allow registry inves-
tigators to adjust for important differences between subjects.
Although RCTs mitigate such differences through randomization,
observational studies derived from registries usually rely on sta-
tistical adjustment to control for identifiable sources of bias.
Similar to RCTs, registries should consider pilot testing
of data collection, a strategy for resolving missing data from the
outset, and methods to maximize retention (of both subjects and
investigators). However, because registries usually store data
and specimens for longer periods of time compared with RCTs,
designers should also develop a plan for ensuring the integrity
437
Caplan et al
of data/specimens over time. In addition, the longer duration of
observation for registries means that training of registry staff
should consist of an iterative process, emphasizing ongoing
training, rather than concentrating on the training at the
initiation of the study.
Finally, the approach to data analysis and interpretation
frequently varies between RCTs and registries. Whereas RCTs
typically engage a particular subject matter only once, registry
investigators can expect to reuse analytic approaches in
numerous related studies, often led by a number of different
investigators. Thus, registry analysts and data managers should
employ a transparent method of coding data and routines that
can be easily recycled among projects and adapted for future
use. Instead of an ad hoc analytic approach, registry statisticians
should construct independent modular programs that may be
combined and recombined for subsequent studies. Lastly, since
data acquisition occurs longitudinally, the integration of newer
and older data should be as seamless as possible.
In general, clinical databases differ from registries by
being a more in-depth profile of a patient cohort over time,
commonly in a prospective longitudinal cohort study. The
database captures great breadth and depth of information,
includes laboratory and radiographic outcomes and usually
involves biobanking of patient specimens for translational
research studies. Despite these differences, there is commonly
overlap between registries and databases, depending largely on
the utilization of the data and the purpose of the study at hand.
By virtue of their greater depth of clinical information, cohorts
documented in clinical databases can be usefully extracted for
registries, but not the converse.
CURRENT ACTIVE ANKYLOSING SPONDYLITIS
REGISTRIES IN NORTH AMERICA
The Spondylitis Association of America (SAA) contin-
ues to explore the feasibility of combining elements from 3
active North American registries. It is valuable to examine the
current status and briefly describe some of the individual
characteristics of these registries.
The Toronto Western Hospital ankylosing spondylitis
(AS) clinic is a longitudinal observation cohort, which is currently
following more than 900 spondyloarthritis (SpA) patients, most
of whom meet the classification criteria for AS. Other important
subsets include reactive arthritis, non-radiographic axial spondy-
loarthritis and undifferentiated SpA. Twenty percent of the cohort
has onset of disease at younger than 16 years, thus defining
juvenile onset SpA. Patients have complete protocol visits
annually and clinic checkup visits every 6 months. In comparison,
the Prospective Study of Outcomes in Ankylosing Spondylitis
(PSOAS) registry includes patients from several sites: Cedars-
Sinai Medical Center, National Institutes of Health/National
Institute of Arthritis and Musculoskeletal and Skin Diseases,
University of California at San Francisco and University of Texas
Health Science Center at Houston. The PSOAS is a comprehen-
sive prospective NIH-funded AS registry that includes approxi-
mately 1212 patients with AS. The registry was initiated in 2003,
with probands continuously recruited and longitudinally followed
at regular intervals since that time. The PSOAS data set captures
standard clinical phenotypes and outcomes for patients with AS;
special attention has been devoted toward acquiring rich
psychological impact measures for affected patients. The PSOAS
also contains genome-wide association study data. Thus, the
clinical variables captured in these 2 registries are largely
congruent. Table 1 provides some comparative data between
these 2 registry cohorts.
438
TABLE 1. Select Clinical Characteristics of 2 Existing
Ankylosing Spondylitis Registries
PSOAS Toronto
No. of patients 1212 577
Male, % 70.1 76.4
White, % 80.9 77.5
Black, % 4.1 2.5
Hispanic, % 6.2 1.4
Asian, % 5.7 10.8
HLA-B27 positive, % 84.3 80
Uveitis, % 32.2 25.6
IBD, % 5.6 13
Psoriasis, % 11 6.9
Age, yr 45 6 14 41 6 13
TNFi usage, % 45 48
IBD, inflammatory bowel disease; PSOAS, Prospective Study of
Outcomes in Ankylosing Spondylitis.
With respect to the Toronto cohort, the clinic continues
to experience significant growth, with the addition of more than
100 new patients with AS per year. Referrals to the clinic derive
primarily from general practitioners and community-based
rheumatologists, but increasingly, the clinic has drawn referrals
from gastroenterologists, ophthalmologists and dermatologists
for the investigation of back symptoms in patients with
inflammatory bowel disease, uveitis or psoriasis, respectively.
There is still a prolonged period of back symptoms before
diagnosis in most patients; so, one of the challenges is to
develop a more effective early recognition/early referral pattern
among allied health professionals caring for chronic back pain,
such as physiotherapists and chiropractors. Active outreach and
education programs are being piloted for both groups at present.
Patient follow-up visits are scheduled for a 6-month
cycle. One of the challenges is to address barriers to compre-
hensive and regular follow-up assessments. These include
geographic distance, scheduling challenges (particularly for
students studying far afield) and concurrent follow-up by local
community internists or rheumatologists. Many patients who
have achieved excellent symptomatic control of AS with
biological therapies lack incentive to return for clinic visits
when they are doing very well.
The primary challenge to sustainability is support for
infrastructure. At present, the AS program incorporates 1
physiotherapist, 1 clinical secretary, 1 clinic research assistant,
1 database manager, 1 program administrator, 1 laboratory
technician and 2 pharmacovigilance nurses. Each member of
the team plays an important role in the clinical, research and
education aspects of the AS program. The clinic continues to
attract outstanding AS fellows, both domestic and foreign. One
of the successes has been the development of satellite AS
clinics in Dublin, Dubai, Singapore, Seoul, Vancouver, Winni-
peg and St. John’s, with both international and Canadian train-
ees. Strong collaborative links with the AS fellows has allowed
development of an international network of AS clinics.
Established in 2007, Program to Understand the Longterm
Outcomes of Spondyloarthritis (PULSAR) is a prospective
registry of patients with SpA within the U.S. Department of
Veterans Affairs health-care system. Patients are extensively
phenotyped clinically on entry to the database, and biological
specimens (sera, plasma, DNA and RNA) are collected for future
analyses. VA medical centers currently recruiting patients include
those in Albany (NY), Dallas (TX), Denver (CO), Jackson (MS),
Volume 345, Number 6, June 2013

Phoenix (AZ), Salt Lake City (UT) and Washington, DC. As of
2013, more than 550 patients have been enrolled. The PULSAR
data set has especially comprehensive pharmacy dispensing data
and linked diagnostic coding, which allows for the collection of
detailed comorbidity data. Because the vast majority of veterans
receive virtually all of their health care through the VA system,
data capture is relatively comprehensive. As is the case with the
aforementioned Toronto and PSOAS registries, standardized
SpA-specific metrology data and patient reported outcome
measures are available for subjects. A functioning independent
Scientific and Ethics Advisory Committee oversees all research
protocols, and the PULSAR registry conforms to the require-
ments of the VA Office of Research and Development and the
Department of Health & Human Services policy on the protec-
tion of human subjects.
A PROPOSED COMBINED REGISTRY
The need for a rigorously developed longitudinal registry
of patients with AS and other SpA is clear and urgent. The
challenge at hand is to develop a consortium of longitudinally
followed, ethnically representative North American patients
with AS and axial SpA, collecting prespecified subjective and
objective clinical data. The SAA has provided seed funding to
facilitate the preliminary work necessary to explore whether
spondyloarthritis patients from existing registries in North
America could be successfully and effectively combined.
Existing registries would furnish the data underlying such
a proposed registry. Phase I of the project, supported through
the generous funding of the SAA, will explore the common data
elements among the 3 contributing registries and document the
level of harmony that exists between these data elements.
Building on the outcome of the phase I effort, phase II
will affirm the feasibility and sustainability of the registry,
carefully define the specific purposes of the registry and finalize
a design for the proposed registry. Our goal will be to har-
monize and leverage the unique strengths of the 3 ongoing
registries; the resulting registry would become a vibrant pro-
spective initiative that could integrate additional patients from
other sources, including the rheumatology practices of clini-
cians in the SPondyloArthritis Research and Treatment Network
(SPARTAN) and the Spondyloarthritis Research Consortium of
Canada (SPARCC) throughout North America. Phase II would
require the convening of a planning committee to define the
potential synergies among SAA, SPARTAN and SPARCC and
to construct the combined registry’s governance structure. The
planning committee’s initial charge would also include articulat-
ing the purpose of the registry going forward, identifying and
Ó 2013 Lippincott Williams & Wilkins
Ankylosing Spondylitis Registries
engaging potential stakeholders, defining the patient population
(eg, AS, axial spondyloarthritis or more broadly defined SpA)
and ensuring the feasibility and sustainability of the registry. It is
anticipated that ongoing funding opportunities will be of interest
to SAA, SPARTAN, SPARCC, local and federal government
agencies and pharmaceutical companies in North America and
throughout the world.
The planning committee could identify and assemble
a collaborative committee of investigators experienced in registry
development to ensure that necessary data elements of the registry
are identified and captured from the outset. Key questions to be
answered from this effort will be to prospectively define the target
population and specify inclusion and exclusion criteria.
As outlined above, following a thoughtful and deliberate
planning process—if a registry is deemed to be feasible and
sustainable—a combined registry committee may be constituted
to provide long-term governance and assist in the with day-to-
day administration of the registry. Additionally, an independent
board could be appointed to provide oversight of registry oper-
ations. This advisory board could ensure that there is strict
adherence to human subject protection and review, prioritize
and approve all requests for data analyses and biological spec-
imen utilization. Advisory board members will be selected so
that they represent areas of necessary technical expertise, as
well as balance the value of the registry information with the
competing needs of stakeholders, including patients.
In summary, current longitudinal North American spon-
dyloarthritis registries are individually and collectively labors of
love. Individually, they have provided and will continue to
provide important information on the natural history and
evolution of disease in the cohorts they were designed to
identify and follow. Collectively, we believe that they could be
used to much better advantage to improve disease definition,
understanding and, ultimately, patient care. Presuming that the
founding registries are able to be harmonized, our hope would
be that a combined registry would be feasible, sustainable and
attractive. Ideally, the registry elements could be developed and
captured in a manner that would lend itself to improving the
care of our patients and systematically capturing and document-
ing crucial clinical and biological information that would
complement ongoing clinical care goals, as well.
REFERENCE
1. Gliklich RE, Dreyer NA, editors. Registries for evaluating patient out-
comes: a user’s guide. Rockville (MD): Agency for Healthcare Research
and Quality; 2007.
439