How Should Diagnostic Imaging Be Chosen? An Evidence-Based Reading of LDCT, X-ray, PET/CT, MRI, and Gadolinium-Based Contrast Agents

Benefits, Limitations, and Risks of Diagnostic Imaging

Bottom line

Diagnostic imaging is not a binary choice between “radiation is dangerous” and “earlier detection must be beneficial.” The examination most likely to change treatment, with overall benefits exceeding harms, should be selected according to the individual's risk, the clinical objective, the imaging protocol, and the actions that will follow. PMID: 33687470

LDCT: Effective screening for high-risk groups does not mean population-wide screening is beneficial

Professor Ching-Shun Lin's original article opposed marketing low-dose computed tomography to all women over age forty. Its point was not to reject LDCT, but to reject the omission of risk stratification. Randomized trials show that LDCT can reduce lung-cancer mortality among high-risk smokers. The same trial also found that a large proportion of positive results were false positives, exposing examinees to follow-up imaging, invasive procedures, and anxiety. Benefit and harm therefore coexist; campaigns cannot promote only “early detection.” PMID: 21714641

An ecological cohort study of lung cancer among Taiwanese women found that, during the expansion of LDCT, the incidence of early-stage lung cancer increased more than sixfold, while the incidence of late-stage disease and mortality did not decline in parallel. This supports Professor Lin's warning about substantial overdiagnosis: some cancers detected might never cause symptoms or premature death during a person's lifetime. Such research, however, cannot determine how an individual patient's tumor will progress and cannot be used to tell a patient to forgo necessary treatment on their own. PMID: 35040922

Overdiagnosis is not an abstract concept but a real harm that must be explained before informed decisions about lung-cancer screening are made. Current guidelines restrict LDCT eligibility to high-risk people who meet age and smoking-exposure criteria, further demonstrating that benefits established in trials of high-risk groups cannot be directly extrapolated to adults in general. PMID: 24322569 PMID: 33687470

X-ray and CT: Optimize dose without calling low risk zero risk

CT uses ionizing radiation. Repeated exposure, or exposure without clinical necessity, must therefore be considered in relation to the risk of radiation-induced cancer. A reasonable approach is first to determine whether the examination will change diagnosis or management, and then use a protocol sufficient to answer the clinical question—not to reject all necessary imaging out of fear. PMID: 42438930

Professor Lin's original article cited the understanding at the time that epidemiologic evidence remained inconclusive for diagnostic exposures below one hundred millisieverts. With the literature now updated, a more precise formulation is that cumulative low-dose exposure may contribute to long-term risk, but the risk to any individual remains difficult to quantify accurately because of dose-estimation uncertainty and confounding. “Inconclusive” should therefore not be read as “entirely risk-free,” nor does it mean that the benefit of appropriate medical imaging is overturned by a very small and uncertain risk. PMID: 41977229 PMID: 39021204

For mammography, current research has not found an increase in peripheral-blood circulating tumor cells after compression. Another study likewise found no clinically meaningful overall increase in circulating tumor DNA or tumor cells. These findings support the procedure's safety at the biomarker level, but the size and endpoints of the studies are insufficient, on their own, to prove that there is absolutely no risk of dissemination under any circumstances. PMID: 23990353 PMID: 31236809

Thyroid exposure from mammography is very low, and routine thyroid shielding is unnecessary; a shield may instead interfere with imaging and increase the need for repeat examinations. For dental X-rays, observational studies have not established causality, although associations with repeated exposure have been reported and are limited by recall bias and insufficient dose data. Radiation-protection judgments in dentistry cannot be directly transferred from conclusions about mammography. PMID: 22358013 PMID: 31502516 PMID: 32557182

PET/CT and MRI: Function, structure, and safety cannot be reduced to a slogan

PET can provide functional and metabolic information, but a blanket claim that CT or MRI can show only structure no longer adequately describes the capabilities of multimodal imaging and advanced MRI. Imaging choices should return to the clinical question the physician genuinely needs to answer, rather than rank modalities by a single technological label. PMID: 42254367

The radiation dose from PET/CT is not a fixed value. Professor Lin's original article discussed a video's claim of twenty-five millisieverts. More recent real-world data indicate that the average for standard protocols may be lower, while more complex whole-body imaging workflows may approach or exceed that value. Twenty-five millisieverts should therefore not be treated as the fixed, typical dose for every PET/CT examination. PMID: 41961220 PMID: 41996412

MRI does not use ionizing radiation and can be an alternative in suitable clinical circumstances. Yet “no ionizing radiation” does not mean that every patient, every implant, and every examination purpose is free of other safety constraints or limitations of applicability. PMID: 41973695

Gadolinium-based contrast agents: Confirmed retention is not the same as proven harm

Professor Lin's original article distinguished two matters that are often conflated. Gadolinium-based contrast agents can be retained in the brain or other tissues, and this may be observed even when kidney function is normal; however, resulting clinical neurologic harm has not been established. Reviews continue to support this distinction. News that “residue remains” should not be taken directly to mean that it “must cause disease.” PMID: 28361260 PMID: 37665796

Nephrogenic systemic fibrosis is associated primarily with severe kidney dysfunction, but risk is not identical across gadolinium-based agents, and modern data concerning lower-risk agents also differ. In practice, the healthcare team should know the patient's kidney function, the necessity of the examination, and the type of agent being considered, rather than treating all gadolinium-based contrast agents as carrying the same risk. PMID: 32976265 PMID: 34997298

Scientific progress has also made safety assessment more nuanced. The earlier reassurance that “no problem other than nephrogenic systemic fibrosis has been established” should now be qualified: the absence of proven clinical harm does not eliminate the need to consider agent stability, tissue retention, and the minimum necessary use. This is the language of risk management after the literature has been updated, not a rewriting of the position Professor Lin expressed at the time. PMID: 41069292 PMID: 40574249

Shielding and special populations: One rule cannot cover every examination

In modern projection radiography, routine contact shielding of patients usually offers little benefit. If positioned inaccurately, a shield may obscure anatomical information, cause repeat imaging, or interfere with automatic exposure control and increase the dose to the examined area. Nevertheless, high-quality comparative evidence remains limited for different examination types and for circumstances involving pregnant women, children, and other groups. “Routine shielding is no longer recommended in most cases” cannot be turned into “individual assessment is never needed under any circumstances.” PMID: 40965682 PMID: 22695996

For gonadal or fetal exposure, doses from modern diagnostic imaging are usually very low, and existing data have not conclusively demonstrated a risk of fetal cancer. Human data at low doses are limited, however, so assessment should still account for the stage of pregnancy and type of examination, avoiding absolute claims that harm is “completely impossible.” PMID: 28418814 PMID: 15191441

Reports must also be read in context: LSV in preterm infants

Lenticulostriate vasculopathy on cranial ultrasonography in preterm infants is associated with multiple infectious and noninfectious conditions. Its cause remains uncertain, and congenital cytomegalovirus is an important association. Current evidence does not support testing broadly for every congenital infection solely because LSV is detected, but it does support screening for congenital cytomegalovirus. “Avoid comprehensive testing” therefore cannot be simplified into “observation alone is enough for everyone.” PMID: 25960415

For isolated, low-grade LSV, evidence remains insufficient regarding long-term neurodevelopmental effects. This is not equivalent to proof that there is no effect at all. Severity, coexisting cytomegalovirus infection, and other abnormalities may all change interpretation and follow-up. PMID: 28192815 PMID: 35112561

Questions to ask before undergoing imaging

Before any diagnostic imaging examination, ask what question the test is intended to answer, whether the result will change treatment, whether a suitable alternative without ionizing radiation exists, and what follow-up will be required after a positive result. For screening in particular, “a mortality benefit has been shown in a high-risk group” is not the same as “more lesions can be found in the general population.” Only when benefit, false positives, overdiagnosis, radiation, and contrast-agent risks appear on the same decision map does imaging truly serve health rather than fear. PMID: 41220098