Clinical Trials in Rheumatoid Arthritis: A Statistical Insight

An autoimmune inflammatory disorder, Rheumatoid arthritis (RA), develops when the body’s immune system attacks healthy cells, resulting in inflammation. RA mainly affects the joints, usually more than one joint simultaneously. As a shock absorber for joints, RA damages the cartilage by uncontrolled inflammation. The result is deformed joints over time. Finally, the bone itself starts to erode. In the aftermath of tissue damage, patients may experience long-term pain, instability, and misalignment of their bodies. Hands, wrists, and knees are commonly affected by this disease. The condition can affect other parts of the body, such as the heart, lungs, skin, and eyes.1

As of 2021, it was estimated that 350 million people in the world have arthritis (according to the Global RA network). Since the disease is so endemic, prevention or treatment is incredibly important, and many drugs have been developed to reduce swelling and pain. However, finding a cure for this disorder has been challenging unless the exact cause is determined and unfortunately, clinical trials for drugs designed to address RA are characterized by many challenges.


Rheumatoid Arthritis Drugs: challenges in clinical trials

Several treatment options for RA can lead to different clinical outcomes. Treatment failure may be correlated with established negative predictive or predisposing factors such as genetic factors, comorbidities, extra-articular symptoms (cardiovascular disease), or pregnancy, among others. These factors plague clinical trials and that present many challenges. Every stage of clinical development must have the safety of patients as the primary focus, even if it isn’t the primary end goal. Hence, choosing the correct sample size, dosage, effectiveness, and other aspects are necessary before starting a clinical trial. Statistical challenges, for the most part, emerge at this point. Having reliable comparators, cautiously selecting study participants, and tracking long-term safety are all important aspects of testing new RA drugs effectively 2.

Every situation must be considered during the trial phases of RA. For example, it may coexist with other conditions or lead to several diseases such as cardiovascular disease. When considering the above example, RA also imposes a broader challenge due to the patient group, making sample size a significant concern. Consequently, there may be dropouts or missing information. Thus, biostatisticians face various challenges during the formulation of the model. The dry run phase could be problematic if the correct data is not entered. Because of this, biostatisticians usually devise alternative strategies. We will take a closer look at the major challenges in developing RA drugs and find out how the biostatisticians are tackling them.


Challenge vs. Solution

Challenge 1: Multiplicity Issues during the clinical trial

An overwhelming amount of trial reports usually do not address multiple testing effectively. Multiplicity involves increasing the type I error rate due to various comparisons, such as comparing subgroups, comparing treatment arms, analyzing multiple results, and analyzing the same results for different intervals of time. Clinical trials have multiple testing problems, which can increase the risk of type I and type II errors, resulting in unexpected interpretation of results. Multiplicity issues should be considered for the initial design, analysis, and interpretation.

In RA, multiplicity can occur if the primary endpoint is assessed in more than one way or during interim analyses. It is due to the association of the disorder with several predisposing factors. Consider a group of patients with RA, for example, that have diverse comorbidities; the new drug may have different effects on each patient, resulting in various outcomes 3. A multiplicity problem arises when multiple hypotheses are tested. It is increasingly likely that at least one valid null hypothesis will be rejected with each new test. When multiplicity is not considered, the chances of at least one null hypothesis being wrongly rejected increase. Multiplicity is one of the reasons regulators are interested in seeing a control plan.


Solution: Adjusting Multiplicity

A lot of adjustments must be made by expert biostatisticians to resolve multiplicity issues. Here are some approaches to resolve or adjust this matter of concern.

  • Multiple primary endpoints may be considered co-primary endpoints. Co-primary endpoints can be various monitoring measures targeted at different aspects of a disease, strengthening the evidence for treatment effectiveness. As a result, the trial will only be successful if all primary endpoints are tested successfully (every null hypothesis is rejected).
  • Hochberg adjustments with ‘n’ hypotheses is another approach. It requires that the worst performing hypothesis pass. Obtaining success for one test will achieve success for all the others that haven’t been evaluated yet. A more challenging threshold would be applied to the next badly performing hypothesis when a test fails. It’s important to adjust the threshold because it helps to prevent false rejections due to small p-values (less than 5%) occurring by chance 4.
  • Like the test outlined above, a Holm adjustment approach aims to make passing the most highly performing test challenging. It is followed by a gradual easing of the level of difficulty. One test that fails indicates that all subsequent tests that are not yet assessed have failed as well.
  • Bonferroni correction: This test is specifically designed to evaluate multiple comparators. The main objective is to decrease the error rate in each comparison since it may impact the other results, causing multiple false positives. Bonferroni’s adjustment involves dividing the number of tests by the alpha value (Type-I error) 4.

Among these, we recommend the fixed sequence testing (Using a pre-specified sequence of hypotheses) and the Bonferroni correction model, in which the P value is well-adjusted, as one of the most effective approaches to resolve multiplicity problems.

But now that we have tackled the multiplicity issue, we are faced with a major concern: missing information/data. In trials, a large amount of data is collected (especially considering various factors), and it is possible to be lost or missed. How can this be prevented?


Challenge 2: Missing Information/data

It is common practice to monitor and evaluate a lot of data in RA trials, whether in patient records or published literature. There is always the risk of missing data in clinical research. Nevertheless, they have the potential to diminish the validity of the results. Until recently, statistical methods for dealing with missing data weren’t widely available. When missing data is present, the analysis is usually limited to only complete cases – those with no missing data in any of the variables included in the analysis. These analyses often produce biased results. Moreover, missing data associated with multiple variables often results in a large percentage of the original sample being excluded. It, in turn, reduces the sensitivity and specificity of the analysis.

The International Conference on Harmonization (ICH) E9 guidelines state that missing data should be prevented at any cost. It is acknowledged that there is no one solution for rectifying missing data because every design and measurement is different. A sensitivity analysis is suggested, and missing data handling will be defined in the protocol, as well as reasons for revocation. Data can be missing randomly (completely or not) or not random at all. As far as new approaches are concerned, several can be implemented statistically.


Solution: Handling Missing Data

In order to handle missing data, several approaches are available. Here are some examples 5:

  • Last observation carried forward (LOCF): After those who have opted out of follow-up investigations, LOCF carries forward their last observation to their last time point. If this approach is used, the carried-forward values can be treated as the observed values for the rest of the observation period.
  • Baseline observation carried forward: It is generally used in trials where the endpoint will return to its baseline value after withdrawal, such as chronic pain in RA trials.
  • Mixed models repeated measures (MMRM): MMRM is preferred for longitudinal continuous outcomes in individually randomized trials. In addition to avoiding model ambiguity, this model is unbiased for missing data absent at random or completely absent at random 6.
  • Multiple Imputation (MI): Managing missing data using MI is a well-used approach. MI entails calculating multiple probable values and filling in the missing data for each subject with missing data related to a given variable. In turn, multiple completed sets of data are created. It enables the user to properly account for the ambiguity about the true value of imputed variables. Binary variables are imputed using logistic regression models, and continuous variables are imputed using linear regression models 7.

When missing data is present, these are typically the approaches used. As a best practice, we recommend multiple imputations in such cases of missing data. Depending on the kind of data that is missing and the level of complexity of the problem, the model can be tailored to meet the needs of the project.


Challenge 3: High Failure Rate/Failing to meet expectations

Innovation in RA pharmaceuticals is becoming increasingly challenging and expensive. The recruitment target and timeline of clinical trials often do not adhere to expectations. Time, workforce, funding, and patient pool resources are vital in randomized controlled trials. Despite the availability of these resources, it is important to remember that they are not always plentiful.

The price of successfully bringing a breakthrough drug to market is high due to late-stage drug failures and the rising costs of confirmatory or follow-up trials. In traditional clinical trials, predefined elements are used to design the study. Several shortcomings of randomized clinical trials have been noted in addressing factors and variables with rigid models. These trials’ success depends on the quality of their clinical assessments. Unfortunately, they are often inaccurate due to numerous anomalies, multiple endpoints, and missing data. As a result of excessive patient dropouts during the clinical stage, profound changes in design are necessary and mostly unavoidable 8.


Solution: Adaptive Design

It is a method of modifying an ongoing trial’s design or statistical procedures during its testing phase in response to variables generated by the trial in question, thus improving the trial’s flexibility. The goal of adaptive clinical trial designs is to make trials more efficient by shortening durations, reducing participant numbers, adjustment/modification of statistical hypotheses, and increasing the likelihood that the drug benefits the trial participants. Also, they effectively address issues related to trial design, for instance, patient variability and relevant treatment effects.

Predetermined changes that adaptive designs may permit statistically include:

  • Optimizing the sample size
  • Abstinence from various treatments or doses
  • Allocating patients to trial arms in a different sequence
  • Selecting patients with the highest likelihood of benefiting from recruitment programs
  • Terminating a trial early because of failure or lack of effectiveness
  • Early conclusions are possible to facilitate the development of new, effective drugs

To some extent, the FDA and EMEA have demonstrated leadership in implementing adaptive trial design 9. It’s imperative to follow the guidelines of the FDA/EMEA while implementing an adaptive design



Rheumatoid arthritis (RA) affects millions of people every year. RA can be effectively treated and managed by following a comprehensive dose regimen and self-management practices. The disease’s complexity presents several challenges in planning a clinical trial. A successful trial requires understanding the perspectives of all stakeholders. Consequently, overcoming these issues is imperative during a trial’s dry run. Hence, the best simulation models are used to achieve optimal results.

Experts develop strategies or design principles that combine experience and modern solutions to bring novel RA drugs to market in compliance with upcoming guidelines. With the right models, development timelines can be shortened, and transitions between phases can be seamless. Many prerequisites need to be in place to increase the likelihood of successful trials. It includes:

  • Expert biostatisticians with broad experience in statistical analysis, especially adaptive design
  • Staff with clinical research experience to efficiently integrate cutting-edge strategies/solutions into the research and trial timeline
  • A centralized infrastructure for performing a deep literature analysis and leveraging resources to the fullest
  • Regulatory analyst interpreting and implementing newly introduced or updated guidelines for advanced modeling


About the Authors


Deepak Manohar; Associate Director, Biostatistics

As an experienced biostatistician in the clinical development industry, Deepak has been working with clinical trials, biologics, biomarkers, and regulatory studies for several years. He has also formulated comprehensive statistical analysis plans, worked with large databases, and supervised teams in a collaborative environment. He specializes in SPSS, SAS, and programming languages to perform Statistical Analysis.


Shrutimita Pokhariyal; Principal Biostatistician

Shrutimita has worked in the pharmaceutical industry as a biostatistician expert in statistics and statistical programming for many years. She specializes in Electronic Data Capture (EDC), Clinical Data Management, Good Clinical Practices (GCP), CDISC Standards, and Data Analysis. At the moment, she is involved in numerous leading studies.


About Us:

ClinChoice is a leading global Contract Research Organization (CRO) with over 3400 clinical research professionals across North America, Asia, and Europe. For more than 27 years, ClinChoice has provided high-quality contract research services to pharmaceutical, biotechnology, medical device, and consumer products clients, encompassing a broad range of services and therapeutic areas.