Improving Response Evaluation Criteria in Solid Tumors (RECIST)

September 13, 2013

What is RECIST?

RECIST (Response Evaluation Criteria in Solid Tumors) is a set of guidelines for clinicians to follow to determine cancer treatment effectiveness. Cancer treatment is complicated, multi-pronged, lasts for years, and significantly impacts the lives of the cancer patient and their family and friends. Determining whether a patient is “cancer free” is challenging. Some cancers can be detected with a blood test but most cannot. These must be monitored using medical imaging modalities such as CT, MRI, and PET imaging.

Standard cancer treatments primarily focus on surgery, radiation, and chemotherapy. Typically, a treatment follows these steps:

  1. Detect cancer.
  2. Decide on treatment (surgery, radiation, or chemotherapy).
  3. Image patient prior to treatment.
  4. Perform treatment.
  5. Image patient every three to six months during treatment.
  6. Measure tumor response to treatment and decide whether or not to continue treatment.
  7. Repeat.

Each cycle through this process constitutes a “treatment plan” and can last several months to years. Furthermore, a cancer patient will likely need to repeat this treatment cycle multiple times trying different treatments each cycle. This can mean decades of painful treatments for patients.

The goal of RECIST is to quantitatively assess tumor response to treatment in an effort to know when to stop ineffective treatments and when to continue effective ones.

How is RECIST implemented?

RECIST involves measuring tumors prior to and during a treatment. This is accomplished by capturing several medical image scans (CT, MRI, etc.) throughout this time. Usually just one modality is used for RECIST assessment to maintain consistency over time. The goal of these measurements is to provide a quantitative assessment of whether the tumor is “changing” size. Size, in this case, is used as a surrogate for determining whether tumor necrosis (death) is occurring. Increasing size is generally considered “growth” and decreasing size is correlated with tumor “death.”

Once the patient is imaged (by CT, MRI, etc.), the reading radiologist typically selects and examines tumor size throughout treatment and provides reports to the medical oncologist. The RECIST guidelines provide selection criteria and measurement methods for this purpose.

In practice, up to about five or so large well-formed solid tumors are usually selected for tracking during treatment. Each tumor is measured by the radiologist who usually selects a single image slice through each tumor that best represents the maximal impression of the tumor through one of the image slices. Then, a measuring tool is dragged across the tumor image along the longest axis of the tumor yielding a length which is used as the guiding metric for size quantitation.

What are RECIST limitations?

RECIST has provided a significant improvement in treatment response assessment since its introduction, however, there are now many notable shortcomings that introduce error into RECIST assessment:

  1. Longest axis does not accurately represent size. A tumor is typically non-spherical in shape and naturally has complex 3D surfaces that are ameba-like in stature. A tumor can change size drastically but still yield the same longest-axis measurement. For example, a tumor that changes from an ellipsoid to a sphere in shape will yield the same longest axis (or vice-versa). This means that a tumor may be incorrectly assessed to be “stable” when in fact it is growing (or shrinking) drastically. Sometimes, two axes are used to address this problem (the longest axis and then a second axis roughly perpendicular to the first).
  2. Malformed tumors make selecting a longest axis challenging. This is especially the case when different radiologists perform the analysis. The radiologist must select a 2D image slice that maximally intersects the tumor and must also manually place longest axis endpoints. Each of these steps have some inter- and intra-operator error associated with them.
  3. A tumor may stay the same size, but in fact have necrosis in the tumor interior. Tumor necrosis, detected by intensity changes in imaging, are not currently considered in RECIST.
  4. Tumor orientation relative to the imaging device can introduce error. Changes in patient positioning, organ shifting, surgery, and/or breathing can affect the 2D image slice axis alignment by imaging equipment (CT, MRI, PET, etc.). Ellipsoid tumors can easily be oriented such that a 2D image slice never accurately portrays the full ellipsoid length. Thus, the 2D length measurement tools confined to image slices cannot capture the true length of the tumor.
  5. Operator error can be introduced by computer screen size. Each pixel on the computer screen can represent upwards of one to two millimeters of tissue. If a radiologist misplaces a longest axis endpoint by a single pixel, the total measurement is susceptible to a few millimeters of error. This is problematic in tumors around one centimeter in diameter or smaller.
  6. A subset of tumors may not accurately represent a patient having dozens of tumors. Time constraints prevent radiologists from considering all tumors a patient may have. Thus, the radiologist must select a subset of tumors for longitudinal analysis. It is possible that this subset does not accurately represent the total tumor burden of the patient.

How to improve RECIST

RECIST can be improved by the following:

  1. 3D Volume assessment. Considering the entire 3D tumor volume can significantly improve quantitative assessment accuracy. Volume represents tumor size better than the longest axis and is a better indicator of tumor growth (or shrinkage).
  2. 3D shape assessment. The 3D tumor shape can be considered when volume does not change from scan to scan. For example, morphometric analysis can be applied to determine shape change between an ellipsoid and sphere shaped tumor sharing the same volume.
  3. Intensity profile assessment. Image intensity changes are known to indicate tumor necrosis. This manifests as intensity profile changes from scan to scan (i.e., consider a histogram of intensity values throughout a tumor). Statistical changes in intensity values across the entire tumor can indicate tumor activity (i.e., growing, necrosis, stability, etc.)
  4. Density and mass assessment. Tumor density and mass may also indicate tumor activity.
  5. Include all tumors. A patient having dozens of lung nodules may better be assessed by considering the total tumor burden if all tumors are assessed en masse.

What is required to implement RECIST improvements clinically?

The RECIST improvements described above are impeded by the following barriers:

  1. 3D image segmentation. Delineation of tumor boundaries in 3D is required in order to assess the tumor volume and shape. 3D tumor segmentation could be performed manually by a wide variety of software tools but introduce operator error and time constraints that impede clinical utility. Furthermore, these tools are often difficult to use.
  2. Automatic 3D image segmentation. To eliminate operator error and relieve time constraints, automatic tumor segmentation is required in a clinical environment. However, automatic image segmentation algorithms are very difficult to design and build and are generally unreliable in practice.
  3. Practical assessment software. Even if tumors could be segmented automatically, practical assessment software that is easy to use needs to exist. This software would enable the radiologist to select all tumors and view their changes longitudinally across a wide variety of characteristics.
  4. Correlating research to clinical outcome. Clinical research needs to be performed to correlate volume, intensity, shape, density, and mass changes to clinical outcome.

David F. Wiley, PhD

President & CTO

Leave a Reply

Your email address will not be published

Log in or register to post comments