Skip to main content
Home / Resources / Blog / A Clearer Crystal Ball: A Meta-regression Model Predicts TB Relapse

A Clearer Crystal Ball: A Meta-regression Model Predicts TB Relapse

What would you guess is the world’s most neglected disease? I bet that you wouldn’t pick tuberculosis (TB)— a disease that causes an estimated 9 million new cases and 1.3 million deaths annually. This infectious disease is caused by the bacterium Mycobacterium tuberculosis. TB usually attacks the lungs, but can attack any part of the body.

Patients infected with TB are typically treated with a standard six-month course of multiple antimicrobial drugs. It is quite difficult to get patients to adhere to this long course of treatment. Often, patients will fail to complete the entire drug course. This increases the likelihood of relapse and antibiotic resistance developing. Thus, there is an urgent need for shorter treatment regimens that minimize the risk of relapse. In this blog post, I’ll discuss how meta-regression modeling of relapse can inform TB clinical trial design.

Predicting TB relapse

The gold standard for diagnosing TB is the sputum culture. This test involves collecting sputum from a patient, putting it in a lab container, and observing it for growth of TB bacteria. If bacteria are observed, the culture is deemed “positive” and is evidence that the patient is infected with TB. “2 month culture status” refers to whether a patient’s sputum is culture positive after two months of treatment. Several studies have shown that 2 month culture status (a common Phase 2 endpoint) can predict long-term relapse (a common Phase 3 endpoint). This means that regimens with higher rates of positive cultures after two months of treatment also have higher rates of long-term relapse. Unfortunately, the findings from these studies were not able to be extended to informing the likelihood of success of new shorter regimens in Phase 3 trials.

Informing the design of studies of new TB regimens

In 2013, a study was published that used meta-regression analysis to identify 2 month culture status and treatment duration as independent predictors of TB relapse. Unsurprisingly, all things being equal, longer tuberculosis treatment regimens have lower relapse rates than shorter treatment regimens. The study used data from more than 7000 patients treated with more than 50 drug regimens of various durations. These data were published from the early 1970s to the late 90s. Since this meta-regression analysis used data from trials conducted decades ago, it was unclear whether this model would be able to predict the relapse rates of modern clinical trials enabling the translation from the Phase 2 endpoint to the Phase 3 endpoint.

To address this concern, we refined the 2013 model by incorporating data from three recent Phase 3 trials that evaluated treatment regimens with an abbreviated 4 month duration. We then tested the model’s ability to predict the proportion of patients that relapsed in recently completed trials as well as a 2009 treatment shortening study. The test of the model showed that the predicted relapse rates were consistent with the observed relapse rates from historic data.

The use of pharmacometrics approaches—including techniques such as meta-regression modeling — during drug development has expanded greatly in recent years. I hope that studies like ours will help inform trial design and decision making to increase the likelihood of success for future TB drug trials. By leveraging the power of model-based drug development, we are taking an important step towards eradicating the scourge of TB.

For more information, please read our recently published PLOS ONE article, “Month 2 Culture Status and Treatment Duration as Predictors of Recurrence in Pulmonary Tuberculosis: Model Validation and Update.” I’d also recommend that you check out this case study on how we helped a client obtain regulatory approval for a new treatment for complicated infections.

About the author

By: David Hermann

Powered by GlobalLink OneLink Software