Choosing Between Treatment Options When Evidence Is Incomplete

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-11

Key Takeaways

• When evidence is incomplete, prioritize a structured framework integrating evidence quality, biological plausibility, and client factors. Rank available evidence by study design strength (e.g., RCTs > cohort studies > case series) and then by species and population relevance, explicitly acknowledging cross-species extrapolation risks due to physiological differences in drug metabolism or disease pathogenesis.
• Diagnostic confidence significantly modifies treatment decisions; lower confidence in the diagnosis necessitates a higher threshold for initiating risky or irreversible therapies, as demonstrated by the underestimation of dental disease burden without radiography in dogs and cats.
• Require a plausible biological mechanism for treatment efficacy in the specific patient, which aids in predicting non-responders (e.g., clot stiffness affecting lytic susceptibility in stroke therapy) and identifying potential failure modes beyond flawed execution.
• Define objective monitoring parameters and response criteria before initiating treatment, such as clinical sign scores, body weight, or specific laboratory values (e.g., PCV, renal biochemistry), with pre-determined action thresholds to guide continuation, modification, or cessation of therapy.
• Document the reasoning process thoroughly, including the evidence considered, limitations, uncertainties, owner goals, and the monitoring plan, to support continuity of care and professional accountability, mirroring RCVS day one competences for critical evaluation and communication.
• Recognize common failure modes such as premature closure on the first plausible plan, overconfidence in surrogate endpoints (e.g., lab values vs. clinical status), or treatment escalation without reassessment; address these by articulating differential diagnoses, defining response criteria, and revisiting the diagnosis when treatment fails.

Clinical decisions in veterinary medicine rarely rest on a complete evidence base. Published trials may be absent, underpowered, or drawn from different species, and the individual patient often deviates from the population that generated the available data. This article provides a structured framework for selecting among therapeutic options when the literature is incomplete, with emphasis on evidence-based reasoning, explicit uncertainty, and client factors. It is written for veterinary students and early-career clinicians who need a defensible method for decisions that cannot be resolved by consulting a formulary alone.

The framework integrates three domains: the quality and applicability of existing evidence, the biological plausibility of treatment effects in the specific patient, and the practical constraints of the clinical context. Each domain contributes distinct information, and a decision that ignores any one of them is vulnerable to a characteriztic failure mode. The goal is not to manufacture certainty where none exists, but to make the reasoning behind a choice transparent, testable, and revisable as new information emerges.

This article does not provide drug-specific dosing or withdrawal periods. Current formulary and label references must be consulted for those details, and regulatory requirements vary by jurisdiction. The principles here apply across species and production systems, though the weight given to each factor will shift with the clinical setting.

At a Glance

The Structure of Incomplete Evidence

Evidence is incomplete in predictable ways. A therapy may have strong evidence in one species but only anecdotal reports in another. The outcome measured in a trial may be a surrogate marker instead of the clinical endpoint that matters to the owner. The study population may exclude the very patients most likely to receive the treatment, such as those with comorbidities or prior treatment failure. Recognizing the pattern of incompleteness is the first step in deciding how much weight to assign to a given source.

The professional competences expected of veterinary graduates include the ability to evaluate evidence critically and to communicate uncertainty to clients. These are not separate skills from clinical reasoning, they are the same skill exercised at different points in the consultation. A clinician who cannot articulate why a particular study does or does not apply to the patient in front of them cannot make a rational treatment choice when the evidence base is thin.

Categorising Evidence by Design and Applicability

Study Design Strength

Randomised controlled trials sit at the top of the conventional hierarchy, but their absence does not leave a vacuum. Prospective cohort studies, case-control designs, and even well-documented case series each contribute information of a different kind. A case series can establish that a treatment is feasible and can identify adverse effects that warrant monitoring, even though it cannot establish comparative efficacy. The key is to match the claim to the design that can support it.

Species and Population Transfer

Cross-species extrapolation is a routine necessity in veterinary practice, but it carries specific risks. Physiological differences in drug metabolism, receptor distribution, and disease pathogenesis can render a treatment ineffective or harmful even when the underlying condition appears similar. The MSD Veterinary Manual organizes its content by species precisely because such differences are the rule instead of the exception. When extrapolating, state the assumption explicitly and design monitoring that would detect a failure of the assumption.

Diagnostic Confidence as a Modifier

The strength of the evidence for a treatment matters only in the context of the strength of the diagnosis. A well-supported treatment for a misdiagnosed condition is not a rational choice. Full-mouth radiography in dogs, for example, yields clinically important findings in teeth without visible lesions in a substantial proportion of patients, which means that clinical examination alone systematically underestimates disease burden and can lead to undertreatment diagnostic value of full-mouth radiography in dogs. The analogous finding in cats, where radiographs of clinically normal teeth revealed important pathology in a large minority of patients, reinforces the same lesson: diagnostic certainty changes the therapeutic calculus diagnostic value of full-mouth radiography in cats. When the diagnosis is uncertain, the threshold for starting an irreversible or high-cost treatment should rise.

Biological Plausibility and Mechanism

A treatment can fail for two reasons: the mechanism is wrong, or the mechanism is right but the execution is flawed. The first failure is more dangerous because it is harder to detect. Requiring a plausible mechanism does not mean demanding a complete molecular explanation, it means asking whether the proposed treatment could plausibly alter the disease process in this patient at this stage.

Mechanistic reasoning is also the primary tool for identifying patients who are unlikely to respond. In human stroke therapy, for example, the lytic agent recombinant tissue plasminogen activator is effective in only a minority of patients with large vessel occlusions, and the physical properties of the clot itself influence lytic susceptibility clot stiffness and lytic susceptibility in vitro. A clinician who understands the mechanism can predict that a highly retracted, stiff clot is less likely to respond and can plan accordingly, either by choosing a different approach or by setting expectations for a partial response.

Personalised and Adaptive Approaches

The limitations of population-level evidence have driven interest in individualised testing. In human oncology, liquid biopsy assays that genotype circulating tumor DNA are being investigated for their ability to guide treatment selection, monitor response, and detect minimal residual disease liquid biopsy in an advanced clinical trial for lung cancer. The underlying logic is that molecular profiling of the individual tumor should outperform a categorical treatment plan derived from population averages.

A more direct approach is functional testing of the patient's own tissue. Microfluidic platforms that expose live tissue slices to multiple drugs in parallel can generate individualised response data on a timescale rapid enough to inform treatment decisions parallel microfluidic chemosensitivity testing on slice cultures. These technologies are largely experimental in veterinary medicine, but they illustrate the direction of travel: replacing population averages with patient-specific measurements where feasible. When such testing is unavailable, the clinician should ask what patient-specific information could substitute, such as prior response to a similar drug or the results of culture and susceptibility testing.

Defining Control and Response Criteria

A treatment decision is incomplete without a definition of success. In human respiratory medicine, the concept of disease control has been formalised to guide therapeutic decisions, using stability of clinical features and validated questionnaire scores over a defined baseline period validating the concept of COPD control. The same logic applies in veterinary medicine. Before starting a treatment, define the objective parameters that will be measured, the time point at which they will be reassessed, and the threshold that will trigger a change in plan.

The monitoring plan should be specific enough that a different clinician could execute it. Which clinical signs will be tracked? At what frequency? What constitutes worsening that requires immediate re-evaluation versus a minor fluctuation that can be observed? Answering these questions before treatment starts converts a vague hope into a testable hypothesis.

Structured Decision Point Assessment

The transition from evidence appraisal to clinical action requires a formal decision point assessment. This assessment sequences the patient, disease, and contextual variables into a ranked problem list, then maps each problem to candidate interventions. The sequence is deliberate: patient signalment and physiological status first, disease characteriztics second, owner and resource constraints third, and clinician capability fourth.

Begin with the patient. Signalment, body condition, comorbidity burden, and current medication profile determine which treatment modalities are physiologically tolerable. A geriatric cat with chronic kidney disease and a dental resorption lesion faces different therapeutic options than a young dog with the same radiographic finding. The diagnostic yield of full-mouth radiography in cats and dogs demonstrates that clinically silent lesions are common, with radiographs of teeth without clinical lesions yielding clinically important findings in 41.7% of cats and 27.8% of dogs in the respective referral populations. These findings alter the treatment plan before any modality is selected, because they expand the scope of disease that must be addressed.

Next, characterize the disease by stage, grade, and expected trajectory. Acute, life-threatening conditions compress the decision timeline and favour interventions with rapid effect. Chronic, progressive conditions permit staged decision making and trial of therapy. Neoplastic disease requires explicit discussion of intent: curative, cytoreductive, palliative, or surveillance. The role of molecular diagnostics in oncology illustrates how treatment decisions are shifting from categorical protocols toward individualised plans based on tumor characteriztics, with the goal of avoiding overtreatment and unnecessary toxicity.

Decision Criteria for Modality Selection

Three treatment modalities dominate clinical choice: medical, surgical, and supportive. They are not mutually exclusive, and most plans combine them. The decision framework below assigns priority based on disease biology, patient status, and owner constraints.

The table is a starting point, not a rule. A surgical condition in a poor anesthetic candidate may be managed medically or supportively even when surgery offers the best long-term outcome. Conversely, a medically manageable disease with severe adverse effects in a given patient may justify surgical intervention. The decision tree in the next section formalises this reasoning.

Decision Tree for Modality Selection

The following decision tree applies to a confirmed diagnosis with at least two plausible treatment modalities. It assumes the clinician has completed the diagnostic assessment and has characterized the disease stage.

1. Is the disease immediately life-threatening without intervention?
2. Yes: proceed with the fastest effective intervention that the patient can tolerate. Stabilize first, then revisit modality choice.
3. No: continue to step 2.
4. Is there a curative or disease-modifying option with acceptable morbidity?
5. Yes: compare the curative option against alternatives using the table above. Consider success rate, complication rate, and recovery time.
6. No: continue to step 3.
7. Is there a palliative medical option that meaningfully improves quality of life?
8. Yes: offer this as the primary plan, with supportive care as an adjunct.
9. No: continue to step 4.
10. Is supportive care alone consistent with acceptable welfare?
11. Yes: recommend supportive care with explicit goals and a review date.
12. No: discuss euthanasia or referral to a specialist center for options not available in the current setting.

At each step, the clinician must state the threshold that triggered the decision. For example, "acceptable morbidity" is defined by the expected complication rate, the patient's physiological reserve, and the owner's capacity for aftercare. These thresholds should be documented in the medical record.

Monitoring Parameters and Response Assessment

Once a modality is selected, the plan is incomplete without defined monitoring parameters. Each parameter should detect a specific failure mode or track a specific therapeutic goal. The choice of parameters depends on the drug or procedure used, the disease being treated, and the patient's baseline status.

The concept of disease control, developed in human chronic obstructive pulmonary disease, illustrates the value of defining control criteria before treatment begins. Control is assessed over a defined baseline period and linked to future outcomes such as time to exacerbation. A similar logic applies in veterinary medicine: define what "controlled" means for the specific disease, measure it at baseline, and reassess at fixed intervals. This approach converts vague impressions of response into reproducible clinical judgments.

Documentation and Communication of Uncertainty

The medical record must capture the reasoning behind modality selection, the evidence considered, and the uncertainty attached to that evidence. This documentation serves three purposes: it supports continuity of care, it protects against errors in recall, and it provides a basis for revisiting decisions when new information emerges.

Document the following elements for each treatment decision:

• The diagnosis and stage, with the diagnostic tests that established them.
• The treatment options considered and the reasons for excluding each one.
• The evidence base for the chosen option, including its source and limitations.
• The monitoring parameters selected and the thresholds that will trigger a change in plan.
• The owner's stated goals, constraints, and understanding of the risks and benefits.
• The date for planned reassessment.

Communication with the owner should mirror this structure. State what is known, what is uncertain, and what will be monitored. The Royal College of Veterinary Surgeons day one competences include the ability to communicate effectively with clients and to recognize the limits of one's own knowledge and skill. These competences apply directly to treatment discussions: the clinician should state when evidence is extrapolated from another species, when it is based on small studies, and when it rests on clinical judgment alone.

Species and production system alter the correct choice in ways that must be stated explicitly. Food animals carry withdrawal period and trade implications that companion animals do not. Herd-level decisions in production medicine may prioritize population outcomes over individual outcomes. The World Organization for Animal Health terrestrial animal health standards address disease control and welfare in ways that can constrain treatment choices in production settings. Wildlife and exotic patients may have limited formulary data and require extrapolation from domestic species with explicit acknowledgement of the uncertainty.

Available equipment changes the correct choice as well. A practice without advanced imaging cannot stage a tumor as accurately as a referral center, and this limitation should influence both the treatment recommendation and the discussion of referral. The MSD Veterinary Manual professional edition and AVMA practice resources provide species-specific guidance and professional standards that support these decisions, but neither replaces the clinician's judgment in applying general principles to a specific patient.

When the evidence base is genuinely limited, say so. Name the gap, state what would be needed to close it, and proceed with the best available option while monitoring closely. This honest framing builds owner trust and creates a rational basis for changing course if the chosen treatment fails.

Recognized Complications and Failure Modes

Therapeutic decisions made under incomplete evidence fail in predictable patterns. The most common failure is premature closure, where the first plausible treatment plan is adopted without systematic consideration of alternatives. This occurs when the clinician anchors on the most familiar diagnosis or the most recent case seen with similar signs. The corrective action is to articulate the differential list in writing before selecting therapy, then test each candidate against the available evidence.

A second failure mode is overconfidence in surrogate endpoints. A laboratory value or imaging finding may improve while the patient's clinical status deteriorates. For example, a radiographic lesion may appear stable while the patient loses condition. The discriminating check is to define response criteria before treatment begins, specifying which clinical parameters will be monitored and what change constitutes meaningful improvement. The concept of control, as validated in human chronic disease management, depends on defining stability across multiple domains instead of a single biomarker.

A third failure mode is treatment escalation without reassessment. When a patient fails to respond, the inexperienced clinician may add another drug instead of questioning whether the diagnosis is correct or whether the original treatment was ever appropriate. Escalation should follow a structured decision point assessment, not occur as an automatic response to lack of improvement.

A fourth failure mode is ignoring the cost of the diagnostic workup itself. Full-mouth radiography in dogs and cats demonstrates this principle: radiographs of teeth without clinical lesions yielded clinically important findings in 27.8% of dogs and 41.7% of cats, but were of no clinical value in 30.5% of dogs and 53.6% of cats. The clinician must weigh the yield of additional testing against its cost and the likelihood that findings will alter management.

Common Errors and Corrective Actions

Students and less experienced clinicians frequently confuse evidence quality with evidence quantity. A large case series may be cited as stronger evidence than a smaller controlled study, when the reverse is often true. The corrective action is to evaluate study design before sample size, asking whether the study included a comparison group and whether outcomes were measured objectively.

A related error is extrapolating across species without justification. Pharmacokinetic differences, metabolic pathways, and receptor distributions vary substantially between species. The MSD Veterinary Manual professional edition provides species-specific guidance that should be consulted before applying a treatment protocol developed in one species to another.

A third error is failure to document the reasoning process. When the evidence base is incomplete, the clinical record must capture the rationale for the chosen approach, the alternatives considered, and the criteria that will trigger reassessment. This documentation serves both continuity of care and professional accountability. The RCVS day one competences include the expectation that graduates can maintain accurate clinical records and justify their decisions.

A fourth error is neglecting to revisit the diagnosis when treatment fails. The clinician should ask whether the original diagnostic tests were adequate, whether the results were interpreted correctly, and whether new information has emerged that changes the differential ranking.

Limitations of Current Evidence and Areas of Expert Disagreement

The evidence base for many veterinary therapeutic decisions remains limited. Personalised approaches that tailor chemotherapy to individual patients are under development, but existing models of drug activity based on tumor cells in culture or animal models cannot accurately predict how drugs act in patients in time to inform treatment. Similarly, liquid biopsy techniques for plasma genotyping of solid tumors are entering clinical research, but their role in routine therapeutic decision making is not yet established.

Expert opinion differs on several questions. The threshold for adding a second therapeutic agent when the first is partially effective is debated. Some clinicians favour early combination therapy to maximize response, while others prefer sequential monotherapy to identify which drug is responsible for benefit or toxicity. The optimal frequency of reassessment is also contested, with some advocating fixed intervals and others recommending response-triggered evaluation.

Where guidance differs between species, production systems, or regions, the clinician must recognize that local factors may override general recommendations. The WOAH terrestrial animal health standards provide international frameworks for disease control and trade-related decisions, but these must be interpreted in the context of local disease prevalence and regulatory requirements. The AVMA practice resources offer additional professional guidance that may reflect regional practice norms.

Referral, Consultation, and Regulatory Reporting

Referral is warranted when the therapeutic decision exceeds the clinician's expertise, when the required diagnostic capability is unavailable, or when the patient's condition deteriorates despite appropriate treatment. Specialist consultation should occur before irreversible interventions are undertaken on the basis of incomplete evidence.

Laboratory involvement is indicated when monitoring requires specialised assays, when therapeutic drug monitoring is needed to guide dosing, or when unexpected adverse effects suggest a pharmacokinetic abnormality. The decision to involve a laboratory should be made early instead of after treatment failure.

Regulatory reporting obligations vary by jurisdiction and by the nature of the adverse event. Suspected adverse drug reactions, unexpected treatment failures, and incidents involving controlled substances may all trigger reporting requirements. The clinician should know the reporting pathways in their jurisdiction and should document the circumstances of any reportable event. Where the evidence base is incomplete, transparent reporting of outcomes contributes to the collective knowledge that will improve future therapeutic decisions.

Frequently Asked Questions

How Do I Choose a Treatment When Cost Limits the Options?

Start by separating the clinical ideal from the affordable alternative. Present both honestly to the client, then build a decision framework around the constraint. Identify which component of the ideal protocol carries the most prognostic weight, and protect that element first. For example, if advanced imaging is unaffordable but changes surgical planning, a full-mouth radiographic series may still be feasible where computed tomography is not, and the diagnostic yield can be substantial. In cats, radiography of teeth without clinical lesions yields incidental or clinically important findings in a meaningful proportion of patients, so the cheaper test is not necessarily the worthless test. Document the financial discussion and the client's informed choice in the record.

What Should I Do When the Recommended Diagnostic Test Is Unavailable?

Use the nearest available surrogate and state its limitations explicitly. A test that answers a related question is often better than no test, provided you adjust your confidence thresholds accordingly. For example, where molecular profiling is unavailable, histopathology and clinical staging still guide most solid tumor decisions, though they cannot identify targeted therapy candidates the way plasma genotyping can. When you substitute a less informative test, widen your monitoring intervals and lower your threshold for re-evaluation. Record the reason the ideal test was unavailable and the rationale for the substitute. If the surrogate result could change the treatment choice, refer to a facility with the required capability instead of proceeding on incomplete information.

How Does the Decision Framework Change for Production Animals Versus Companion Animals?

The framework stays the same, but the weight assigned to each criterion shifts. In production medicine, population-level outcomes, withdrawal periods, and trade implications often outweigh individual response. The World Organization for Animal Health terrestrial standards provide a structured basis for decisions that affect herd health and international movement. In companion animals, the individual's quality of life and the client's emotional and financial capacity carry more weight. The evidence base also differs. A study showing diagnostic value in dogs does not automatically transfer to cats, and neither transfers to cattle. Re-evaluate the species and production system transfer of every source before applying it, and consult species-specific references for conditions where cross-species extrapolation is known to be unreliable.

What Must I Document When Choosing a Treatment on Incomplete Evidence?

Record the clinical question, the options considered, the evidence for each, and the reasoning that selected one option. Note the specific uncertainties, including gaps in species-specific data and unknown long-term outcomes. Document the client's informed consent, including what was disclosed about the limits of the evidence. The Royal College of Veterinary Surgeons day one competences expect graduates to recognize the limits of their knowledge and to communicate those limits professionally. If you consulted a colleague or specialist, record that too. Finally, set a clear review date and specify which monitoring parameters will trigger a change in plan. This record protects the patient, supports continuity if another clinician takes over, and provides a basis for learning when the outcome is known.

How Do I Explain Uncertainty to a Client Without Undermining Confidence?

Use a structured explanation that separates what is known from what is not. State the diagnosis and the recommended treatment first, then explain the evidence gap in plain terms. For example, you might say that the treatment is supported by studies in similar patients but that individual responses vary. Offer a concrete monitoring plan so the client knows what will be watched and when. This approach mirrors the concept of disease control used in human respiratory medicine, where control status is defined by stability and symptom burden over a defined period, and it gives the client a tangible way to judge whether the treatment is working. Avoid vague statements about uncertainty. Name the specific unknowns and the specific signs that would prompt a change.

When Should I Stop Treating and Refer or Reassess?

Stop and reassess when the patient fails to meet predefined response criteria within the agreed timeframe, when new information changes the diagnosis, or when the client can no longer sustain the treatment plan. Set these criteria before treatment starts, not after. If the patient deteriorates despite treatment, re-examine the diagnosis instead of simply escalating therapy. Refer when the required expertise, equipment, or monitoring capability is outside your practice, or when the client requests a second opinion. In research settings, adaptive platforms can test multiple compounds in parallel and identify the most promising therapy on a timescale useful for clinical decisions, but such tools are rarely available in practice. In their absence, a timely referral is the adaptive strategy.

Related Clinical & Scientific Guides

• Veterinary Case Presentation: Structure and Delivery
• Veterinary Communication in the Workplace: Team Dynamics
• Monitoring Plans for Hospitalized Veterinary Patients

References and Further Reading

• Liquid biopsy and its role in an advanced clinical trial for lung cancer.. 2018.
• Parallel microfluidic chemosensitivity testing on individual slice cultures.. 2014.
• Effect of Clot Stiffness on Recombinant Tissue Plasminogen Activator Lytic Susceptibility in Vitro.. 2018.
• Diagnostic value of full-mouth radiography in cats.. 1998.
• Validating the Concept of COPD Control: A Real-world Cohort Study from the United Kingdom.. 2017.
• Diagnostic value of full-mouth radiography in dogs.. 1998.
• RCVS Day One Competences. RCVS.
• MSD Veterinary Manual, Professional Edition. MSD Veterinary Manual.
• American Veterinary Medical Association Practice Resources. American Veterinary Medical Association.

Related Articles

• Ethical Reasoning in Veterinary Practice: Frameworks for Decision-Making
• Veterinary Treatment Plan Development: A Structured Approach
• Bandaging Materials in Veterinary Practice: A Guide to Selection and Use
• Building a Differential Diagnosis List: A Step-by-Step Framework
• Canine Physical Examination: A Template for Thorough Assessment

This article is educational professional reference material for veterinary audiences. It is not a substitute for veterinary diagnosis, individual clinical judgment, current product labeling, or applicable regulatory requirements.