🧪 This platform is in early beta. Features may change and you might encounter bugs. We appreciate your patience!
length_time_bias
Length-time bias occurs when screening programs preferentially detect slow-growing, less aggressive disease variants because they have a longer pre-symptomatic window during which screening can detect them. Fast-progressing cases cause symptoms and are detected clinically before screening, while slow cases are overrepresented in screened populations. This makes screened patients appear to have better outcomes, not because screening helps, but because their disease was less severe from the start.
A cancer screening program shows that screened patients survive an average of 7 years post-diagnosis while unscreened patients survive only 3 years. However, this difference may simply reflect that screened patients had slow-growing tumors that would have been less dangerous regardless.
A new screening program for thyroid cancer detects a large number of small, slow-growing tumors. Patients found through screening appear to have much better five-year survival rates than those diagnosed after symptoms appear. Critics note this likely reflects the fact that screening preferentially catches indolent tumors that would have remained harmless, not that the screening genuinely saves lives.
A workplace screening initiative for Type 2 diabetes shows that employees identified through routine testing live years longer post-diagnosis than those identified after symptoms emerge. However, the screened group predominantly has slow-progressing metabolic disease that would not have caused serious harm for decades, inflating the apparent benefit of the program.
Binary (yes/no) questions an LLM must answer to identify this aspect:
Does the study involve screening or early detection of a condition?
Type: binaryAre screened cases compared to clinically detected cases without controlling for disease aggressiveness?
Type: binaryIs a better prognosis in screened patients attributed to the screening program itself?
Type: binaryIs there evidence that the screened group disproportionately has slow-progressing disease variants?
Type: binaryLength-time bias occurs when screening programs preferentially detect slow-growing, less aggressive disease variants because they have a longer pre-symptomatic window during which screening can detect them. Fast-progressing cases cause symptoms and are detected clinically before screening, while slow cases are overrepresented in screened populations. This makes screened patients appear to have better outcomes, not because screening helps, but because their disease was less severe from the start.
The biological aggressiveness of a disease directly affects both its detectability by screening and its natural prognosis, creating a strong confound that mimics a treatment effect.
Look for randomized controlled trials of screening programs that compare mortality rates in screened vs. unscreened populations. Survival time from diagnosis is an unreliable endpoint for screening efficacy.
Length-time bias has complicated evaluation of prostate cancer (PSA) screening, where screened men appeared to do better largely due to detection of indolent tumors.
Use these tools to detect, analyze, or train this aspect.