Research Review By Dr. Demetry Assimakopoulos©

Audio:

Download MP3

Date Posted:

September 2016

Review Title:

Screening Tests to Predict Injuries – Critical Review

Included Papers:

  1. Bahr R. Why Screening tests to Predict Injury do not Work – and Probably Never Will…: A Critical Review. British Journal of Sports Medicine 2016; 50: 776-780.
  2. Wright AA, Stern B, Hegedus EJ et al. Potential Limitations of the Functional Movement Screen (FMS): A Clinical Commentary. British Journal of Sports Medicine 2016; 50(13).

Background Information:

The use of screening tests to predict future injury is becoming increasingly popular in sports medicine. Generally, screening tests are purported to detect disease or dysfunction in individuals not demonstrating overt symptomatology. The intention is to identify pathological conditions early in the disease process, leading to early intervention. Breast cancer screening with mammography and prostate cancer screening with blood testing are likely the best, evidence-based examples in general medicine. In spite of the evidence demonstrating the benefits of early detection, the use of screens in medicine is still a hotly debated topic, because not all screening programs and subsequent interventions have led to favourable clinical outcomes (1, 2).

The sports medicine, rehabilitation and manual therapy literature has discussed a number of physical examination tests and screens that demonstrate a statistically significant association with subsequent injury risk. Unfortunately, while these tests provide clinicians and researchers with a greater understanding of causative factors of injury, they are unlikely to predict which specific athletes will suffer a future injury with sufficient accuracy. The following review discusses the use of screens in the context of sports medicine.

Discussion:

To ensure that screening programs deliver their intended benefits, the World Health Organization published the following criteria:
  1. The condition being screened for is in an important health problem.
  2. The condition has a detectable early stage.
  3. Treatment in the early stage is more beneficial than treatment at a later stage.
  4. A suitable test is available to detect the disease in the early stage.
These criteria work well in the context of general medicine, where screens are used to detect an established disease such as breast cancer, as early as possible. However, in the context of sports medicine, clinicians are using performance tests in an attempt to detect physical impairments which might predispose an athlete to future injury. There are inherent differences between screening for the presence of disease and screening for future injury. In the case of disease, a binary classification is used: either the person is sick, or healthy. In the case of assessing for sports injury risk, the outcome is usually continuous, which makes prediction much more difficult. To detect injury risk accurately and fulfill criteria number 2 above, researchers must take continuous variables and make them binary: is this athlete at risk (yes) or not (no). With the number of variables that contribute to potential injury, converting something complex and continuous into something binary is difficult.

When screening for disease, the objective is to administer early treatment to prevent calamity. However, in sports medicine, injury risk is assessed and an intervention is administered to hopefully minimize injury risk factors.

In sports medicine, injury risk factors are either modifiable or non-modifiable. Modifiable risk factors, such as strength, endurance and motor control can be targeted in many, if not all groups. However, non-modifiable risk factors, such as previous injury or gender, can only be targeted to specific subgroups, and not all individuals. Depending on the athlete, injury prevention might be difficult.

Developing a Screening Program is done in a series of 3 steps:
  1. Prospective cohort studies, demonstrating a statistically significant association between the screening test and future injury.
  2. Repeating the prospective cohort study in hopes of creating a cut-off score, which separates high-risk athletes from low-risk ones. This step also requires researchers to test the tool in different athlete groups. Statistics such as sensitivity, specificity, odds ratios, P-value, etc. are usually done at this stage.
  3. RCTs to determine whether or not an intervention applied according to the screen, changes the screen’s result and mitigates subsequent injury. Ideally, screening programs should also be done in other athletes at low risk of injury.
Screening Test Properties:

Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) are important statistical values often used to describe testing properties. Just for review, here’s what these terms mean:
  • Sensitivity: does the test capture all those with injury?
  • Specificity: does it capture ONLY those with injury?
  • PPV: how many athletes scoring a positive test are injured?
  • NPV: how many athletes scoring a negative test are not injured?
Let’s use a study on the vertical drop screening test for ACL injury (3) as an example. In this study, peak external knee adduction moment during landing had the strongest association with knee injury risk. The researchers concluded that this measurement can predict ACL injury status with 78% sensitivity and 73% specificity. The problem with this conclusion is the substantial overlap in test results between the injured and uninjured groups, meaning the test likely missed people at great risk for injury, or categorized individuals that were not at risk as high risk. From this evidence, the test is imperfect.

Unfortunately, sensitivity and specificity are inversely related. This means that if you’d like your screen to capture ALL players at risk of future injury (100% sensitivity), specificity will likely suffer. This means that many more players who are likely not at great risk will be misclassified as high-risk.

Still, in spite of excellent sensitivity and specificity, individual risk likelihood is difficult to ascertain. In this case, likelihood ratios (also represented by receiver operating characteristic curve analyses, or area under the curve) are helpful. (WRITER’S NOTE: likelihood ratios are often used for assessing the value of performing a diagnostic test. These values are built upon sensitivity and specificity of the test to determine whether a test result usefully changes the probability that any condition exists). For this statistic, a value of 1.0 is a perfect prediction, while 0.50 indicates that the test is no better than flipping a coin. While vertical drop testing demonstrates acceptable sensitivity and specificity for predicting ACL injury, the likelihood ratio is .56, indicating that the test is not as good at predicting future injury as previously thought.

Clinical Application & Conclusions:

The authors of this paper (#1 above) described the essential steps researchers and clinicians must use to create an injury screening test. Step 1 consists of identifying a strong relationship between a screening test and injury risk in a prospective study. In step 2, researchers must test the properties of the screen and validate it in relevant populations using appropriate statistical tools. Step 3 must consist of an RCT to test the effect of an intervention program to prevent injury in athletes with a remarkable screen. To date, there is no good example of a screening test which predicts future sports injury adequately.

The Functional Movement Screen (FMS) is a tool that has become hugely popular in the sports science world. To date, a lot of work has been done both clinically and empirically in an attempt to validate the use of the tool to predict future injury in a variety of professional, amateur and occupational athletes. A composite score of < 14 on the FMS is commonly considered the threshold for a potential injury risk. Unfortunately, this number was conceived out of studies with small, homogenous populations, which limits generalizability. (WRITER’S NOTE: generally, we like to see homogeneous groups. This is because it limits the variability of the population sample, and decreases the chances of individual differences accounting for the results. I surmise that the authors of Study #2 are commenting on the homogeneous samples because one unfortunate side-effects of using only homogeneous groups is limited generalizability to other athletic populations.) Also, the score of < 14 has not been adequately replicated in subsequent studies. Additionally, it is likely that this number might change, depending on the athletic event.

Generally, a useful screening test has high sensitivity. A meta-analysis of 6 studies (4) evaluating the FMS for injury prediction reported sensitivity of only 0.24 (low!) and 0.85 specificity. A sensitivity of 24% suggests that the FMS will be positive in approximately 24% of athletes who eventually suffer an injury. This means that a cut-off of < 14 will overlook up to 76% of those who will eventually become injured. The meta-analysis also concluded the FMS has a negative likelihood ratio of 0.87. These poor sensitivity and LR scores suggest that the FMS cannot be used confidently as a screen for future injury, and clinicians should exercise caution when using this in clinical practice as an injury prediction tool.

What about working with athletes in real life? Truthfully, we have to ask ourselves whether or not using a specific methodology or algorithm to help you progress an athlete through a training program will hurt them in the end. If you see dynamic knee adduction and pain, and improving the athlete’s biomechanics decreases their pain, then the athlete has benefitted no matter what the likelihood ratio is. An approach such as this will not be effective in 100% of athletes, and realistically, it is unreasonable to expect it to be. No test will be able to account for every variable that can possibly result in injury, so hanging your hat on one test is just silly.

Additionally, we as clinicians do not make decisions in a vacuum. One test, or series of tests in the case of the FMS, doesn’t tell you everything you need to know about an athlete’s potential for injury. However, it might guide you towards something you or the athlete hadn’t thought of or noticed, which might benefit them in the end. Assessing movement quality and quantity is only one angle of athletic assessment. Great clinicians have a variety of tools, spanning from orthopedics, ‘functional’ sports medicine, palpation, psychology and intuition that they use to provide a comprehensive assessment and treatment approach. Objectively, one test alone will not be a magic bullet. Statistics should not be required to prove that.

Study Methods:

No statistical methods were done. These papers were a critical review and a commentary.

Study Strengths / Weaknesses:

Strengths:
  1. This commentary provided an accurate definition of what a screen is, and what it is not.
  2. It described that screens might not be as helpful as once thought in medicine generally, and not just in sports injury management.
Weaknesses:
  1. It would have been nice if the authors included a chart describing 10-15 common “functional” tests, their accompanying statistics and a take-home message about its actual clinical utility.

Additional References:

  1. Ilic D, Neuberger MM & Djulbegovic M. Screening for prostate cancer. Cochrane Database Syst Review 2013.
  2. Wilson JMG & Junger G. Principles and practice of screening for disease. Public Health Papers No 34. Geneva: WHO, 1968.
  3. Hewett TE, Meyer GD, Ford KR, et al. Biomechanical measures of neuromuscular control and valgus loading of the knee predict anterior cruciate ligament injury risk in female athletes: a prospective study. Am J Sports Med 2012; 40:521-526.
  4. Dorrell BS, Long T, Shaffer SW, et al. Evaluation of the Functional Movement Screen as an injury prediction tool among active adult populations: a systematic review and meta-analysis. Sports Health 2015; 7:532-537.