There comes a moment, in the journey of a patient treated for cancer, when imaging ceases to be a photograph and becomes a verdict. It is not always the day of diagnosis. It is often later, during treatment, or just after. The physician waits for the images. The patient waits for the physician’s gaze. We no longer simply ask: “What is it?” We ask: “Is it working?”
At that moment, imaging bears a strange responsibility: it must say something alive about a shifting disease, while only capturing indirect signals. It must decide in the uncertain: progression or scar? inflammation or tumor? true response or illusion of response? And, harder still, it must sometimes detect efficacy even before the body has had time to shrink.
In contemporary oncology, this question of therapeutic response is no longer occasional. It repeats. Patients live longer, treatments follow one another, strategies become personalized. Imaging becomes a regular breathing rhythm of the care pathway. And as it repeats, another question rises: do we have to inject every time? Can we monitor treatment response without systematically resorting to iodinated contrast CT or gadolinium MRI?
The topic may seem technical. It is not, not entirely. It touches examination frequency, safety, departmental organization, patient acceptability, and above all the very quality of what we call responding. Because reducing contrast only makes sense if we preserve the essential capacity: detect early, decide right.
Why response assessment was built around injection
For decades, treatment response was told with a very simple metric: size. A tumor shrinks, the treatment works. A tumor grows, it fails. This logic became a common language for world oncology with RECIST, then RECIST 1.1, a framework that standardizes how to measure lesions, how many to follow, and how to define response and progression in trials and, by extension, in practice.
RECIST 1.1, in itself, does not explicitly impose a contrast agent. It imposes a reliable measurement. But in the real world, injection is often what makes measurement possible: it better delineates a hepatic lesion, distinguishes a lymph node from a vessel, reveals metastases that would otherwise blend into the parenchyma. Without contrast, certain targets become less visible, therefore less measurable, therefore less reliably trackable. It is not a matter of comfort; it is a condition of legibility.
And then there are domains where response cannot be reduced to size. In some tumors, the central question is not how big is it but is it still alive. This is precisely the spirit of mRECIST in hepatocellular carcinoma HCC: we do not measure only the mass, we measure the viable portion, identified by arterial enhancement. In other words: contrast is built into the very definition of response.
Same logic in modern response algorithms after locoregional liver treatments: the LI-RADS Treatment Response Assessment TRA relies on imaging signs including the improvement or persistence of hyperenhancement to estimate tumor viability.
In these frameworks, wanting to do without contrast is not a marginal optimization: it is a questioning of the language of response itself.
Why the without contrast question becomes urgent now
The first reason is arithmetic: the number of examinations is exploding. Modern oncology relies on regular, sometimes closely spaced follow-ups because we adjust quickly, because we change therapeutic lines, because we want to avoid losing time. The result: injection, once a punctual act, becomes a repeated exposure.
The second reason is cultural and regulatory: authorities have reminded us that contrast is not innocuous, even when very useful. In MRI, debates on gadolinium retention led to restrictions in Europe for certain linear agents with a precautionary logic: use only when necessary and at the minimum sufficient dose, preferring more stable agents.
The third reason is clinical: treatments have changed. The arrival of immunotherapies revealed a disconcerting phenomenon: lesions sometimes grow before responding, or new lesions appear transiently. This is not always true progression, it is sometimes inflammation linked to the immune response. To frame these atypical trajectories, iRECIST was proposed as a guide, standardizing how to confirm progression under immunotherapy and avoid premature treatment discontinuation.
In this context, imaging must do more than measure: it must interpret a biological narrative. And the more complex the reading, the more fragile the idea of reflexively repeating injections becomes.
Finally, there is a very simple reason, rarely stated: injection is tiring. Tiring for patients, tiring for teams. It adds flow constraints, surveillance, consent, risk management. If one can answer the clinical question without injecting, then injection becomes a complexity cost that must be justified.
Clarifying what without contrast means
In this article, without contrast refers mainly to the absence of iodinated products CT and gadolinium MRI. There exist other forms of injection in imaging notably tracers in nuclear medicine such as FDG PET which do not share the same risk profile or the same logic. They do not answer the same question: where iodinated or gadolinium contrast highlights vascularization or permeability, PET measures metabolism, a biological activity.
It is therefore possible, in certain pathways, to reduce contrast by relying more on other injections. This is not necessarily progress or regression: it is a change in strategy. But it must be named, in order not to sell a parsimony that does not exist.
The central problem: the late response of size
Size-based response monitoring RECIST has an enormous advantage: it is simple, reproducible, shared. But it has a structural flaw: size often changes late.
A tumor can be biologically dying without having shrunk yet. It can necrose, fibrose, hollow out, and yet retain comparable volume. Conversely, some lesions may diminish through inflammatory or vascular effects without the disease being truly controlled. This is one of the reasons why adapted criteria have emerged by organ and treatment including mRECIST in HCC, based on viability rather than mass.
To follow a response without contrast, one must therefore find markers that tell biology earlier than size.
This is where functional imaging enters the scene.
Diffusion: measuring cell death without injecting
The most emblematic without contrast follow-up technique in oncology is MRI diffusion DWI. Its principle can be told in one sentence: it measures the freedom of water movement in tissues.
In a highly cellular tumor, water circulates less: diffusion is restricted. When cells die, membranes disorganize, extracellular space increases, water circulates more: diffusion increases. This increase is often expressed as a rise in ADC apparent diffusion coefficient, a number derived from DWI.
This marker is fascinating because it speaks the language of the living. It does not say only it is big or it is small. It says something like: the microstructure has changed.
This potential is not merely theoretical. In breast cancer, for example, a major multicenter trial ACRIN 6698 showed that diffusion parameters could predict pathological response in patients receiving neoadjuvant chemotherapy, in link with ADC variations during treatment.
In cervical cancer, a meta-analysis found a significant correlation between ADC values measured early during treatment and ADC variation and subsequent response, suggesting an interest of diffusion for predicting or monitoring response to radiochemotherapy.
And in other cancers, the same pattern returns: diffusion is studied as an early marker, sometimes faster than morphological changes. A meta-analysis in nasopharyngeal carcinoma evaluated ADC’s ability to predict treatment response, illustrating this dynamic beyond the classical MRI cancers.
The interest, for our subject, is obvious: all of this can be done without gadolinium.
But diffusion has a problem, and it is a crucial one: the measurement is not identical everywhere.
When a number becomes fragile: ADC variability
ADC has an aura: it resembles an objective biomarker, almost like a blood test. Yet it depends heavily on technique: machine, parameters, b values, reconstruction, chosen ROI, motion, artifacts. Two hospitals can produce ADCs that are not directly comparable if their practices are not harmonized.
This is precisely why standardization organizations have taken up the topic. The QIBA profile for DWI insists on the need to establish acceptable variance levels so that ADC becomes a truly usable biomarker, and details requirements and precautions for robust measurements.
Put differently: diffusion can be a powerful tool if it is treated as a measurement tool, not as a visual add on.
And this draws a key point of the without contrast response: it is not simply less injection. It often demands more technical discipline.
The breast: a laboratory of gadolinium free monitoring
Breast cancer is interesting because it concentrates several challenges: neoadjuvant treatments where one wants to know quickly if it works, repeated examinations, and an imaging history very tied to contrast dynamic MRI.
Yet in recent years, diffusion has taken a growing place in follow up to the point of being evaluated against injected tools. Analyses have compared DWI and DCE MRI for predicting pathological complete response, discussing their respective performances.
The idea is not that diffusion always replaces injected MRI. The idea is more pragmatic: in certain pathways, notably follow up, one can reduce gadolinium dependence by relying more on non injected markers provided thresholds, timings, and acquisition conditions are validated.
This approach tells something about the general evolution of oncology: we are learning to see the response otherwise than by watching contrast enhancement.
The liver: where contrast absence becomes a real dilemma
Here the subject grows complex, because the liver is precisely the territory where the notion of viability rests most heavily on enhancement.
In HCC, as seen, mRECIST is based on the viable portion identified by contrast. In modern assessments after locoregional treatments, frameworks like LI RADS TRA also use criteria based on the improvement or persistence of enhancement, among other signs.
So how can one imagine response monitoring without contrast in a field where response is defined by contrast?
The honest answer is: it can be done in certain contexts, but one cannot pretend it is equivalent.
Diffusion is studied as an alternative or complement including for response assessment of certain systemic or targeted treatments, or for hepatic metastases. Recent reviews discuss DWI use as an early response marker in different cancers, highlighting its non injected nature and the derived parameters it allows.
But the crucial point is the following: without contrast, one often loses the most direct ability to distinguish viable tumor tissue from necrosis especially after embolizations, ablations, or treatments that alter vascularization. One can partially compensate with other markers, but compensation is not always perfect.
This observation imposes a realistic strategy: the without contrast response is not a uniform revolution. It is an adaptation by indication.
The brain: when response depends as much on contrast as on what does not enhance
In neuro oncology, the with or without contrast debate has a particular tone. Because contrast enhancement is a major marker for certain tumors but also because contrast can mislead, notably after radiotherapy, creating pseudo progressions or inflammatory changes resembling recurrence.
RANO 2.0 insists precisely on the complexity of follow up: one must consider enhancing and non enhancing components, therapeutic contexts, and sometimes confirmatory imaging to avoid concluding too quickly on progression.
Why does this matter here? Because it recalls an evidence: even when contrast is central, reading the response can no longer be contrast equals tumor. Non injected sequences notably FLAIR and diffusion play a role in overall interpretation.
In other words, even in a very contrast dependent field, response is already no longer solely injected.
PET: changing the question rather than removing injection
When one seeks to monitor response without iodinated or gadolinium contrast, one quickly encounters a major alternative: PET notably FDG PET which measures the metabolic activity of lesions.
It is a conceptually different approach. Where RECIST measures size, PERCIST proposes response criteria based on uptake intensity and evolution suggesting thresholds to define a metabolic response.
Here again, it is not without injection. But it is often without iodine or gadolinium and sometimes earlier: metabolism can decline before the mass shrinks.
The downside is that PET has its own constraints: availability, cost, irradiation, inflammatory false positives particularly under immunotherapy, interpretation sometimes delicate depending on tumor type.
This point matters because it reveals a general principle: monitoring response is not merely reducing a gesture. It is choosing a signal. Size, vascularization, diffusion, metabolism each signal tells a different facet of treatment. And sometimes, the best strategy is not to remove contrast, but to stop depending on a single signal.
Immunotherapy: when response becomes a narrative, not a number
The without contrast question takes on a particular color with immunotherapy, because it made visible a deeper problem: response is not always monotonic. A lesion may grow before regressing. Inflammation may mimic progression. A new lesion may appear then disappear.
iRECIST was designed precisely to provide a rule of conduct: do not conclude too quickly, confirm certain progressions, standardize data collection to make trials comparable.
This framework says something essential: in contemporary oncology, the difficulty is not only seeing, but not over interpreting. Reducing injections is not the core of the problem; the core is constructing a robust reading of the trajectory.
And in this context, non injected tools diffusion, fine morphological sequences may have a complementary role: not to replace contrast, but to help interpret what contrast shows.
AI and radiomics: the ambition to read response in the invisible
Much is said about artificial intelligence in oncology, but one of its most interesting promises when one is concerned with contrast reduction is not to manufacture artificial images. It is to extract more information from what we already have, including non injected sequences.
Radiomics approaches consist in extracting hundreds of features from an image texture, shape, heterogeneity to find patterns associated with a response or a prognosis. When properly conducted, they can reveal tendencies the human eye does not see. But they are also vulnerable to overfitting, to acquisition variability, to lack of external validation.
In our subject, this path has a clear interest: to reduce contrast, one must increase the value of non injected sequences. AI can be part of this augmentation, provided it is validated as a clinical tool and not as a demonstration.
What is gained, and what is risked, in without contrast follow up
One gains first an obvious parsimony: fewer gestures, fewer constraints, fewer repeated exposures. In long term follow ups, this matters.
One gains sometimes an earlier response. This is the strongest argument for diffusion: detecting a microstructural modification before size changes. Data in breast and cervix show that early treatment ADC variation may be associated with subsequent response.
One also potentially gains better organizational tolerance: faster, smoother examinations, sometimes easier to multiply when imaging access is strained.
But two major risks are taken.
The first is losing the notion of viability in cancers where it is central. HCC and locoregional treatments are the typical example: the absence of contrast can make certain response criteria simply inapplicable.
The second is relying on non injected markers without sufficient standardization. Diffusion is powerful, but ADC is sensitive to variability. Without control over variance, an ADC change may reflect protocol variation rather than tumor evolution.
The conclusion is therefore neither blind enthusiasm nor skepticism. It is pragmatism: without contrast follow up is possible, useful, sometimes superior for certain signals but requires methodological rigor and careful indication selection.
Toward a multi-signal response: the end of all contrast, and the end of all size
Looking at the overall evolution, a transition emerges: response is no longer a single measurement, it is a combination.
Size remains a common language, structured by RECIST 1.1. Enhanced viability remains indispensable in certain cancers and treatments, as mRECIST and hepatic response frameworks remind us. Diffusion provides a non injected marker of microstructural change, potentially early. PET provides a metabolic reading, structured by criteria like PERCIST. And in domains such as neuro oncology, modern criteria explicitly acknowledge the need to read the enhancing and non enhancing together, cautiously, in the face of false signals.
In this landscape, monitoring response without contrast is not a revolution replacing everything. It is a shift: reducing dependence on a single gesture, a single signal, a single routine.
The real issue is not whether to inject or not. It is whether you know exactly what signal you are relying on when you decide that a treatment works or fails.
