The case for bringing traditional medicine into primary care was made formally in 1978. The Declaration of Alma-Ata listed traditional practitioners, suitably trained, among the workers on whom primary health care could rely (WHO, 1978). Since then the World Health Organization has issued successive strategies urging member states to regulate traditional and complementary medicine and to integrate it into national health systems (WHO, 2013), and its global survey records a steady rise in the number of countries with national policies, laws and offices devoted to it (WHO, 2019). India has gone further than most, with a dedicated ministry and a national health policy that commits to mainstreaming the AYUSH systems and to a pluralistic model of care (Government of India, 2017).

These are real achievements. They have given traditional systems legal standing, training institutions and a place in public facilities. The argument of this article is that they are not sufficient, and that treating policy recognition as the goal has allowed a harder task to be postponed. Integration is worth having only if it improves care, and whether it does depends on evidence that policy cannot supply by declaration.

Integration in practice has mostly meant co-location. A Unani or Ayurvedic physician is given a room in a primary health centre, a supply of medicines from an approved list and a place in the staffing table. Patients choose which door to enter. This arrangement increases access and respects preference, which were the public health arguments for it (Bodeker and Kronenberg, 2002). It does not, by itself, create any working relationship between the two kinds of practitioner. There are seldom shared records, agreed referral criteria or common protocols for conditions that both treat.

The reason is not hostility alone. A conventional clinician asked to refer a patient with osteoarthritis, irritable bowel or early hypertension to a traditional colleague will reasonably ask what the patient can expect. A traditional practitioner asked when to refer in the other direction needs to know at what point the limits of the system have been reached. Neither question has an agreed answer for most conditions, because the studies that would answer them have not been done or have been done badly. Policy has built the house and left the rooms without connecting doors.

Evidence-based medicine was defined by its founders as the conscientious, explicit and judicious use of current best evidence in decisions about the care of individual patients, combined with clinical expertise and the patient's values (Sackett et al., 1996). That definition is not hostile to traditional practice. It asks only that claims of benefit be open to test. For a traditional intervention to be used with confidence in primary care, several kinds of evidence are needed, and they build on one another, as Figure 1 shows.

The base is description. Much of what traditional practitioners do in clinics has never been recorded systematically: which patients come, with what complaints, what is prescribed and what happens next. Above that comes the quality of the product. A trial of a herbal formulation means little if the next batch differs, and the CONSORT extension for herbal interventions asks authors to report the plant species, the part used, the extraction and the chemical characterisation for that reason (Gagnier et al., 2006). Safety follows. Analyses of traditional preparations sold to the public have found lead, mercury and arsenic in a proportion of them (Saper et al., 2004), and clinically significant interactions between herbal medicines and prescribed drugs are well documented (Izzo and Ernst, 2009). In a primary care population, where many patients take several medicines, that is a daily concern.

Only on that footing do questions of efficacy and of value in ordinary practice become answerable. Table 1 summarises the five kinds of evidence, the question each answers for a primary care service and the general state of each for traditional systems. The pattern is uneven. Pharmacopoeial standards and pharmacovigilance schemes exist on paper in several countries, but the upper layers are thin for all but a handful of interventions.

The last row matters because primary care runs on guidelines. The GRADE approach, now used by most guideline bodies, rates the certainty of evidence and separates it from the strength of a recommendation (Guyatt et al., 2008). A traditional intervention will enter a mainstream guideline when its evidence can be graded, and it will stay out, whatever the policy climate, when it cannot.

The position taken here invites three objections, each with some force.

If the argument is accepted, the content of policy support should change. Funding that now goes to expanding posts and dispensaries would do more good if a share of it built research capacity within traditional medicine: training in clinical research methods in Unani and naturopathy colleges, clinical trial units attached to teaching hospitals, and biostatistical support. Regulators could require adverse event reporting from integrated clinics and make the results public. Health services could treat co-located clinics as places of study as well as service, with simple shared records that capture diagnoses, treatments and outcomes for patients on both sides.

Priorities are needed, since not everything can be studied at once. The sensible candidates are conditions that are common in primary care, for which conventional treatment is unsatisfactory or burdensome, and for which traditional systems make specific and testable claims: chronic musculoskeletal pain, functional bowel disorders, and the early stages of metabolic disease are examples. For each, the work should run in order, from a clear description of what is done to a defined and quality-controlled intervention, and then to pragmatic comparison with usual care. Practitioners have their own part, which is to record outcomes honestly, to report harms, and to accept that some cherished treatments will not survive testing.

Policy has opened the door of primary care to traditional systems of medicine. Whether they are used well once inside depends on knowledge that is still mostly missing: what works, for whom, how safely and in combination with what. Supplying that knowledge is slower and less visible than issuing a strategy, and it is the form of support that would serve patients and the traditional systems best.

To be declared by the author(s) before publication.

To be declared by the author(s) before publication.