The first N-of-1 trial in Ayurveda: An Indian Experience and Lessons Learned

Authors: Swati Sharma, Arun Kaushik, Sankha Shubhra Chakrabarti, and Kishor Patwardhan

A first for Ayurvedic research

We recently published what appears to be the first N-of-1 trial in Ayurveda. The study, which appeared in the Journal of Ethnopharmacology (Sharma et al., 2026), evaluated Kalyanaka Ghrita (KG) in a patient with Mild Cognitive Impairment (MCI). KG has been described in Ayurvedic texts dating back around 2,000 years and is indicated for disorders of cognition. Here, we narrate why the study was designed this way, what we learned, and how it may herald a new direction for Ayurvedic research.

From a clinical question to fruit flies, and an RCT that did not take off

The story began with a simple question: which Ayurvedic formulations are prescribed today for cognitive disorders, and which are worth testing using rigorous methods? In 2019, we approached the Government Ayurveda Research Institute for Mental Health and Hygiene (GARIM) in Kottakkal, Kerala, India, which shared a list of formulations commonly prescribed to its patients. We selected the most frequently used ones and tested them in fruit flies (Drosophila melanogaster). This became the basis of Swati’s postgraduate work (MD in Ayurveda). She studied four formulations, including KG, in models of neurodegeneration (Sharma et al., 2021; Sharma et al., 2023) at Prof. SC Lakhotia’s lab in the Department of Zoology. Those experiments showed early evidence of neuroprotective effects.

As part of her doctoral work, Swati wanted to conduct a conventional randomized controlled trial in patients with dementia. The absence of well-established modern medical interventions for MCI also made patients more willing to volunteer. However, recruiting enough patients who met stringent inclusion criteria and could commit to months of follow-up was difficult.

Around that time, Kishor was asked to review a manuscript proposing N-of-1 trials for evaluating Ayurvedic drugs, which was published more than a year later (Chatterjee et al., 2024). The idea itself was not new. A few authors had already argued that N-of-1 designs suited Ayurveda. What did not exist was a published trial report implementing one. While we were stuck on recruitment, the idea clicked: “Why not us?”

Why an N-of-1 design made sense

Ayurvedic practice holds that prescriptions depend on the individual patient, including their constitution, digestive strength, lifestyle, and other factors. Conventional RCTs, on the other hand, measure an average effect in a group, creating some tension with Ayurveda’s logic.

MCI generally progresses slowly, making a two-month period, the length of each period in our design, reasonably suitable for within-patient comparisons. This made MCI a good fit for N-of-1 trial. The N-of-1 design also allowed us to include patients with multiple comorbidities rather than seek an “ideal” patient meeting stringent inclusion criteria.

We planned four independent N-of-1 trials, based on formulations previously studied by Swati in Drosophila models of Alzheimer’s and Huntington’s disease (Sharma et al., 2021; Sharma et al., 2023).

Addressing the question of reproducibility

We have observed differences of opinion among Ayurvedic physicians about how individualization is carried out in clinical practice. Rather than basing treatment choice on parameters that are difficult to reproduce, we used the classical Ayurvedic indications for each formulation. The patient whose comorbidities matched the largest number of indications was assigned the corresponding formulation. This was a deliberate choice to make the process reproducible.

Designing and running the trial with what we had

We did not have specialized N-of-1 infrastructure or prior experience with this design. What we did have was Sankha Shubhra, a geriatrician and longtime collaborator, and a busy outpatient geriatric clinic. Because of time constraints, we could take only two of the four formulations through to patient recruitment. The paper derived from another trial employing the other formulation is currently under peer review.

We used validated Hindi-language tools, including the Hindi Mental State Examination and Clinical Dementia Rating scale, along with a quality-tested batch of KG with documented HPTLC analysis. Routine blood tests at every visit allowed us to track safety data.

The placebo problem, and why we went open-label

Could this trial have been blinded? In principle, a placebo-controlled design provides stronger evidence, and it has been done for Ayurvedic medicines. A well-known example is Furst et al.’s (2011) trial in rheumatoid arthritis, but its placebo matching worked differently than one might assume. Rather than matching each formulation, the placebos matched the dosage form. Thus, with nearly 150 formulations spanning six dosage forms, the trial needed only six placebos. Its parallel-group design also helped: each patient stayed on the same intervention, real drug or placebo, throughout, without having to compare the taste or smell of one treatment with another.

Our situation was different. The same patient had to receive both the drug and placebo in alternating periods. KG has a distinct smell, taste, consistency, and texture, making it difficult to reproduce in a placebo. Ghee itself is not an inert carrier either, making a truly neutral substitute difficult to identify. Manufacturing a placebo that matched the exact smell, taste, and texture without containing ghee was almost impossible in our setting.

Why a quasi-randomized approach?

We attempted randomization using a computer-generated sequence. It returned a strictly alternating A:B:A:B:A:B pattern, resulting in a quasi-randomized design. Since the trial was open-label, both the patient and assessor knew the treatment administered during each period, so we proceeded with the existing sequence. The treatment order was therefore predictable.

A tryst with Bayesian statistics

Our initial plan was to use only descriptive statistics and visualizations. As the visualizations suggested potentially interesting patterns, we explored whether statistical methods could add value. That was when we turned to Arun, an expert in Bayesian statistics. We used Bayesian linear regression to estimate posterior probabilities and 95% highest-posterior-density intervals for each outcome. With so few observations at each follow-up and only one patient, conventional frequentist methods have limited usefulness. Bayesian methods, in contrast, allowed us to express the posterior probability of a given effect, along with a credible range for its magnitude.

Fourteen months, one patient

The trial was meant to run for twelve months. It took fourteen. Family responsibilities, two hospitalizations for an unrelated illness, caregiver illness, festival holidays, and seasonal disruptions all pushed the timeline out.

What we would do differently, and what’s next?

Looking back, we would want biomarker or neuroimaging data to complement the clinical scales. Since our patient also had depression, and their depression scores improved, a future trial could test whether this improvement was driving the cognitive improvement.

Pooling results from a series of N-of-1 trials could strengthen the evidence but would bring its own challenges. Classical ayurvedic formulations are marketed by multiple pharmacies. Batch quality, ingredient sourcing, processing methods, and textual sources can also vary. Such differences would need to be measured and reported consistently for meaningful pooled analysis. Otherwise, we would not know how much the “same” formulation differed from one trial to another.

Could this model scale?

N-of-1 trials do not need large recruitment pipelines or complex blinding. They need one willing patient, validated outcome measures, and statistics suited to small samples. For a system of medicine already centered on the individual, this may be a natural fit. Whether this becomes a broader tool for building evidence for Ayurveda will depend on more groups trying it and pooling their results to identify emerging patterns.

References

Chatterjee, A., Subramaniam, S., & Sankaranarayanan, A. (2024). Scientific evaluation of Ayurvedic drugs: The use of N-of-1 clinical trials. Indian Journal of Medical Ethics, 9(1), 41–44. https://doi.org/10.20529/IJME.2023.067

Furst, D. E., Venkatraman, M. M., McGann, M., Manohar, P. R., Booth-LaForce, C., Sarin, R., Sekar, P. G., Raveendran, K. G., Mahapatra, A., Gopinath, J., & Kumar, P. R. K. (2011). Double-blind, randomized, controlled, pilot study comparing classic Ayurvedic medicine, methotrexate, and their combination in rheumatoid arthritis. JCR: Journal of Clinical Rheumatology, 17(4), 185–192. https://doi.org/10.1097/RHU.0b013e31821c0310

Sharma, S., Singh, G., Patwardhan, K., & Lakhotia, S. C. (2021). Beneficial effects of two Ayurvedic formulations, Saraswata Ghrita and Kalyanaka Ghrita, on survival and on toxic aggregates in Drosophila models of Huntington’s and Alzheimer’s disease [Preprint]. Europe PMC. https://europepmc.org/article/PPR/PPR405455

Sharma, S., Singh, G., & Patwardhan, K. (2023). Effects of two Ayurvedic formulations, Dhanwantaram Kashaya and Saraswatarishta, on life history parameters and toxic aggregates in Drosophila models of Huntington’s and Alzheimer’s disease. Journal of Ethnopharmacology, 305, Article 116080. https://doi.org/10.1016/j.jep.2022.116080

Sharma, S., Kaushik, A., Chakrabarti, S. S., & Patwardhan, K. (2026). Role of an Ayurvedic formulation ‘Kalyanaka Ghrita’ in mild cognitive impairment: A single-patient, quasi-randomized, controlled, open-label N-of-1 trial. Journal of Ethnopharmacology, 371, Article 122089. https://doi.org/10.1016/j.jep.2026.122089