Listen

Module 1. What HTA is — and where it sits in Saudi Arabia. Definition, domains of value, and who does what between approval and access.

HTA informs a decision; it is not the decision. The 2020 HTAi/INAHTA definition describes HTA as a multidisciplinary process that uses explicit methods to determine the value of a health technology at different points in its lifecycle. Its purpose is to inform decision-making that promotes an equitable, efficient and high-quality health system. The output is advice: a structured judgement on clinical benefit, cost, and wider consequences compared with what is done today. The decision itself belongs to the payer, ministry or hospital committee that owns the budget.

The domains of value. A full assessment looks at more than a trial result. The usual domains are clinical effectiveness and safety, economic evaluation, budget impact, ethical and social issues, organisational impact on services, and the patient and public perspective. Different decisions weight the domains differently. A drug decision leans on clinical and economic evidence. A decision on a new test leans on how results change management, and on workflow.

Who does what between approval and access. In simplified terms: the regulator (SFDA) decides whether a product may be marketed. Insurers and public payers decide what is covered; the Council of Cooperative Health Insurance sets the benefit rules for private cover. A national procurement body runs tenders. Health clusters and hospital drugs-and-therapeutics committees decide what is on the local formulary. Professional societies write guidelines. Independent HTA bodies and societies build capacity and can offer advice. Mandates change, so confirm current roles before relying on this map.

A worked example. A regulator asks: is this advanced therapy for ulcerative colitis safe, effective and well made? A guideline panel asks: where does it sit in the treatment sequence? An HTA asks: compared with what we already fund, is the extra benefit worth the extra cost, for which patients, and what else would that money have done?

Module 2. Framing the decision problem. PICO, comparators, perspective, time horizon and patient-relevant outcomes.

From a vague wish to a decision problem. Every assessment starts with a decision problem: a precise statement of the population, intervention, comparator, outcomes, time frame and setting (PICOTS). It is written before anyone reads the results, and it is the contract against which the evidence is later judged. A vague question such as “is this biologic good?” cannot be answered; “does drug X improve sustained remission versus current standard therapy in biologic-naive adults with moderate-to-severe UC over one year?” can.

The comparator is the most argued-over choice. The comparator should be the best alternative that is actually available to the same patients: usually current standard of care, not placebo. Where practice varies, include the commonly used options. A trial against placebo cannot show whether a new drug beats what patients would otherwise receive. When no head-to-head trial exists, the gap is filled by indirect comparison (Module 3).

Perspective, time horizon and outcomes that matter. The perspective states whose costs and benefits count: usually the health-system payer, sometimes society. The time horizon must be long enough to capture all important differences. For a chronic disease such as IBD that means years or a lifetime. Outcomes should matter to patients: remission, avoiding surgery, hospitalisation, quality of life. A surrogate such as endoscopic healing is useful only to the extent it predicts those outcomes.

A worked example. P: adults with moderate-to-severe UC, biologic-naive. I: new advanced therapy. C: infliximab biosimilar (what clusters actually use first). O: clinical remission at week 52, steroid-free remission, colectomy, serious infection, quality of life. T: 52 weeks, with a lifetime model for cost. S: hospital gastroenterology services.

Module 3. Clinical evidence: synthesis, indirect comparison and certainty. Systematic review, network meta-analysis, GRADE and the limits of surrogate endpoints.

A review that others could repeat. An HTA starts with a systematic review: pre-specified question, documented searches of several databases, duplicate screening, risk-of-bias assessment and a transparent synthesis. Industry dossiers should be checked against an independent search, because selective inclusion is easy to miss. Meta-analysis pools comparable trials; heterogeneity (I², differing populations or doses) limits how far they can be pooled.

When there is no head-to-head trial. Most IBD biologics were tested against placebo, not each other. An adjusted indirect comparison (Bucher method) uses a common comparator: the relative effect of A versus B is the ratio of A versus placebo to B versus placebo, and the uncertainty adds on the log scale. Network meta-analysis generalises this across many treatments. Both rest on an assumption that the trials are similar enough in patients and design. Head-to-head trials, such as the VARSITY trial of vedolizumab versus adalimumab in UC, are rare and valuable.

How certain are we? GRADE. GRADE rates certainty for each outcome as high, moderate, low or very low. Randomised evidence starts high and is downgraded for risk of bias, inconsistency, indirectness, imprecision and publication bias. Observational evidence starts low and can be upgraded for large effects. Certainty is what lets decision-makers know how much to trust the number they are about to put in a model.

A worked example. For advanced therapies in UC, an independent search for randomised trials of adults with moderate-to-severe disease, reported clinical remission at induction and maintenance, finds trials the sponsor omitted because they used different dosing. Including them changes the pooled estimate and the confidence interval.

Module 4. Economic evaluation: costs, QALYs and the ICER. Cost-effectiveness analysis, thresholds, Markov models and uncertainty.

ICER, QALY and the threshold. Cost-effectiveness analysis compares incremental cost with incremental health gain. With health measured in quality-adjusted life years (QALYs), the incremental cost-effectiveness ratio is ICER = (Cost_new − Cost_comparator) / (QALY_new − QALY_comparator). A technology that costs less and gives more QALYs dominates; one that costs more and gives fewer is dominated. In between, the ICER is compared with a willingness-to-pay threshold, which expresses what the system is prepared to pay for one QALY. Countries set thresholds differently and there is no universal value; older “1 to 3 times GDP per capita” rules of thumb are widely criticised.

Modelling a chronic disease: Markov models. Trials are short. A Markov model extends them by moving a cohort between health states in fixed cycles, with each state carrying a cost and a utility. A UC model typically has remission, mild-to-moderate activity, severe activity, post-colectomy and death. Transition probabilities come from the evidence synthesis (Module 3). Future costs and QALYs are discounted to present value, and a half-cycle correction avoids over-counting.

Uncertainty is part of the answer. One-way sensitivity analysis varies parameters one at a time to find which matter. Probabilistic sensitivity analysis (PSA) draws all parameters from their distributions many times and plots the results on the cost-effectiveness plane. A cost-effectiveness acceptability curve (CEAC) shows the probability that the technology is cost-effective at each threshold. Where uncertainty is large and the decision is costly to reverse, value-of-information analysis asks whether more research is worth doing before deciding.

A worked example. Illustrative: the new therapy costs SAR 90,000 more per patient over the horizon and adds 0.45 QALYs. ICER = 90,000 / 0.45 = SAR 200,000 per QALY. Against an illustrative threshold of SAR 150,000, it is above the line. Net monetary benefit = 0.45 × 150,000 − 90,000 = −SAR 22,500, which says the same thing without dividing.

Module 5. Budget impact and affordability. Why a good-value technology can still be unaffordable, and how biosimilars change the picture.

Value is not the same as affordability. Cost-effectiveness asks whether a technology is good value per patient. Budget impact analysis (BIA) asks what it does to total spending for a particular payer over the next few years. A cheap-per-QALY drug for a common condition can still overwhelm a budget, while an expensive-per-QALY drug for a rare condition may barely register. Both questions are needed, and they are answered by different models with different perspectives, horizons and no discounting.

Building a budget impact estimate. Each year: eligible patients × uptake of the new technology × net incremental cost per patient. Net cost is the new technology’s cost minus the cost of what it displaces, minus any downstream savings that fall within the same budget period. Uptake usually ramps up; ignoring the ramp overstates early cost. Time horizon is typically 3 to 5 years. Sensitivity analysis should vary eligible population and uptake first, as these are usually the most uncertain.

Biosimilars and what savings can fund. When originator biologics lose exclusivity, biosimilars of the same molecule enter at lower prices. Switching studies, such as the NOR-SWITCH trial of infliximab, support the safety and efficacy of switching from originator to biosimilar in stable patients. The HTA question is then how much is saved, how fast, and what those savings could fund. Savings are real only if the switch actually happens and prices fall, which depends on tenders, clinician confidence and patient communication.

A worked example. Direct-acting antivirals for hepatitis C were cost-effective in many settings even at high launch prices, but the number of eligible patients made the first-year bill a major concern. Elimination programmes needed negotiated prices and phased treatment, not a different ICER.

Module 6. Beyond the numbers: values, equity and deliberation. Multi-criteria thinking, patient voice, transparency and conflicts of interest.

What the model cannot see. An ICER treats a QALY as equal whoever gains it. That is a value judgement. Many systems also care about severity of illness, unmet need, age, rarity, access in underserved regions, and effects on carers. These are not errors in the model; they are reasons a committee may reach a different conclusion than a threshold alone would. The task is to make those reasons explicit, not to hide them in the discussion.

Multi-criteria decision analysis (MCDA). MCDA makes the weighing explicit. The committee agrees criteria (for example: clinical benefit, safety, unmet need, cost-effectiveness, budget impact, feasibility), sets weights before looking at the products, scores each product on each criterion, and computes a weighted total. MCDA does not replace deliberation: the scores are a structured starting point for discussion, and large disagreements between members are where the useful conversation happens.

Transparent, accountable deliberation. Trust in HTA depends on process. Members declare financial and non-financial interests each year and for each topic; those with significant conflicts step out. The committee records abstentions, writes reasons, and publishes the decision with the evidence used. Patients and the public contribute lived experience of the condition and treatment burden. Each recommendation carries a review date and the events that would trigger earlier review.

A worked example. Two candidate technologies: an advanced therapy for paediatric-onset Crohn’s disease and a repeat screening test in low-risk adults. If both have the same ICER, a committee may reasonably favour the one that addresses a severe condition in children with few alternatives. It should say so and give the reason.

Module 7. Devices, diagnostics and AI. Why drug-style evidence rules do not fit, with colonoscopy AI, FIT and calprotectin as examples.

Devices are not small drugs. Devices change through repeated design iterations, so the version tested may not be the version sold. Trials are often smaller and cannot always be blinded. Real-world performance depends heavily on the operator and a learning curve. A fair assessment states which version the evidence describes, does not penalise unavoidable design limits as it would for a drug, and separates the effect of the device from the skill of the person using it.

Tests are judged in two steps. A diagnostic test is first judged on analytic validity (does it measure what it claims?) and clinical validity (does the result correspond to the condition, as sensitivity and specificity?). The second step is clinical utility: does using the test change management in a way that improves outcomes or saves resources? Accuracy alone does not establish value. A test can be accurate yet change nothing.

AI that keeps learning. Computer-aided detection (CADe) in colonoscopy is an example: randomised trials show higher adenoma detection, though the benefit for interval cancer or mortality is not yet established, and some effects shrink with experienced endoscopists. Algorithms that continue to learn after deployment raise a new problem: the version assessed is not the version in use. An assessment should ask whether the algorithm is frozen, require a predetermined change-control plan with agreed limits and triggers for reassessment, and demand performance by subgroup, not only on average.

A worked example. Endoscopic bariatric procedures: outcomes depend on operator experience. Trial results from expert centres may not hold in new services, so the assessment asks about training requirements, minimum case volumes and outcomes by operator experience.

Module 8. Managed entry, real-world evidence and reassessment. Sharing risk, learning after launch, and revisiting decisions.

Managed entry agreements. When benefit is promising but uncertain, or the price is high, a payer and manufacturer can agree terms that share risk. Financial agreements (price–volume, discounts, caps, rebates) manage budget. Outcomes-based agreements link payment to results in practice, for example refunds for patients who do not respond. They only work if outcomes are measurable, data collection is funded and the administrative cost is lower than the benefit. Agreements should be transparent enough to be evaluated, even if exact prices are confidential.

Real-world evidence and registries. Registries and routine data can answer questions trials cannot: how a therapy performs in older patients, with comorbidity, in routine centres, over long periods. To be useful for decisions the data need a defined population, validated outcomes, complete follow-up and a protocol agreed before collection starts. Without these, “real-world evidence” is just uncontrolled observation.

Reassessment, disinvestment and horizon scanning. A recommendation is not permanent. Triggers for reassessment include new trial results, safety signals, a large price change, biosimilar or generic entry, a new competitor, or real-world data that contradict the model. Disinvestment means withdrawing or restricting a technology that no longer offers value, which needs the same transparency as adoption. Horizon scanning identifies what is coming (for example new MASH or IBD therapies) so that budgets, diagnostics and training are ready.

A worked example. List price SAR 70,000 per patient-year; 45% respond at one year; non-responders’ cost refunded at 50%. Expected net price = 70,000 × (1 − 0.55 × 0.50) = SAR 50,750. Cost per responder = 50,750 / 0.45 ≈ SAR 112,800, against SAR 155,600 without the rebate.