How many patients were in the Bunevicius T4+T3 trial?

Thirty-three patients completed the crossover study. All were women with hypothyroidism on stable levothyroxine replacement. The small sample size is one of the most frequently cited limitations, as it limits statistical power for detecting differences in subjective psychometric endpoints.

Did the Bunevicius trial use a washout period between treatment arms?

No. Patients switched directly from one arm to the other with no washout period. Because liothyronine's downstream effects on mood and cognition may outlast its serum half-life, residual carryover could have influenced results in the second treatment period.

Why didn't larger trials replicate the Bunevicius results?

At least four larger randomized trials (Clyde 2003, Sawka 2003, Walsh 2003, Saravanan 2005) and a 2006 meta-analysis failed to confirm that T4+T3 combination therapy produced consistent neuropsychological benefits over T4 alone. The original positive finding may have reflected the small sample, multiple uncorrected endpoints, or population-specific factors.

Was the Bunevicius trial funded by a pharmaceutical company?

Yes. The study received partial funding from SmithKline Beecham, which manufactured brand-name liothyronine (Cytomel). While industry funding does not automatically invalidate results, it is a relevant disclosure that later reviewers flagged when evaluating the trial's outsized influence on prescribing patterns.

What does the ATA say about T4+T3 combination therapy?

The 2012 ATA guidelines concluded that available evidence did not support routine use of combination T4+T3 therapy. The guidelines acknowledged persistent patient interest and called for better-designed trials but stopped short of recommending the combination as standard practice.

How long was each treatment period in the Bunevicius trial?

Each arm lasted five weeks, for a total study duration of 10 weeks. Several guideline bodies have since recommended that trials of thyroid combination therapy use treatment periods of at least three to four months to capture stable psychometric outcomes.

Did the trial correct for multiple statistical comparisons?

No. The study tested multiple neuropsychological and mood instruments without applying a correction for multiplicity (e.g., Bonferroni). Running six or more hypothesis tests without adjustment inflates the family-wise error rate, meaning some statistically significant results may have occurred by chance.

Can the Bunevicius results be applied to men?

Not directly. All 33 completers were women. While hypothyroidism is more common in women, the absence of any male participants means the trial provides no evidence about whether men respond differently to T4+T3 combination therapy.

What T4-to-T3 substitution ratio did the trial use?

The investigators replaced 50 mcg of levothyroxine with 12.5 mcg of liothyronine, implying a 4:1 potency ratio. The actual bioequivalence ratio remains debated (estimates range from 3:1 to 5:1), and slight over-replacement could have produced mild tissue-level hyperthyroidism in the combination arm.

Is the Bunevicius trial still relevant today?

It remains historically significant as the catalyst for the T4+T3 research agenda. Its clinical relevance, however, has been largely superseded by subsequent larger trials and meta-analyses that did not confirm its primary findings. It is best understood as a hypothesis-generating pilot study, not as practice-changing evidence.

Honest Criticisms and Limitations of the Bunevicius T4+T3 Trial

By HealthRX.com Medical Team

Published May 25, 2026Updated May 25, 2026Last reviewed May 25, 2026

Clinical image for Honest Criticisms and Limitations of the Bunevicius T4+T3 Trial Image: HealthRX.com clinical image

At a glance

| Detail | Value | |---|---| | Sample Size | 33 (completed) | | Population | Adults with hypothyroidism on stable levothyroxine | | Intervention | 50 mcg of the daily T4 dose replaced with 12.5 mcg liothyronine | | Comparator | Usual levothyroxine dose (T4 monotherapy) | | Design | Double-blind, randomized, crossover | | Duration | 5 weeks per arm, 10 weeks total | | Primary Endpoint | Composite of cognitive, mood, and quality-of-life psychometric tests | | Key Result | T4+T3 arm scored better on multiple neuropsychological measures and was preferred by 19 of 33 patients |

Why This Trial Still Draws Fire

The Bunevicius et al. 1999 NEJM paper is probably the most cited single study in the T4-versus-T3 debate. It was the first well-publicized randomized trial to suggest that adding liothyronine to levothyroxine could improve neuropsychological outcomes. But a closer reading reveals methodological choices that make the results harder to interpret than the abstract suggests.

What follows is not a dismissal of the trial. It is a structured accounting of the criticisms that peer reviewers, subsequent trialists, and endocrine society guideline committees have raised over the past 25 years.

Sample Size and Statistical Power

Thirty-three patients completed both arms of the crossover. By modern standards, this is an underpowered sample for detecting differences in subjective psychometric scores, where effect sizes tend to be small to moderate. The trial did not publish a formal power calculation in the original NEJM report. Without one, it is impossible to know whether the study was designed to detect a clinically meaningful difference or simply any statistically significant shift.

Small samples inflate the risk of both Type I errors (false positives from chance clustering) and Type II errors (missing real effects in underpowered subgroups). When a small trial reports positive results across multiple endpoints, the probability that at least some of those results are spurious rises with each additional comparison.

The Multiple-Comparisons Problem

The trial assessed a large battery of neuropsychological and mood instruments. The published results reported improvements on several individual scales without a pre-specified correction for multiple comparisons (such as Bonferroni or Holm-Sidak adjustment).

HealthRX.com Multiple-Endpoint Audit of the Bunevicius Trial:

| Outcome Domain | Instruments Used | Reported p <0.05? | Adjusted for Multiplicity? | |---|---|---|---| | Mood | Profile of Mood States (POMS), Visual Analogue Scales | Yes (several subscales) | No | | Cognition | Digit Symbol, Digit Span, others | Yes (Digit Symbol) | No | | Physical symptoms | Billewicz index | Yes | No | | Global preference | Patient preference question | Yes (19/33 preferred T4+T3) | N/A (descriptive) | | Thyroid function | TSH, free T4, total T3 | Mixed | N/A (biochemical) |

When six or more hypothesis tests are run without multiplicity correction, the family-wise error rate can exceed 25%. This does not prove the findings were false. It means the statistical confidence is weaker than any single reported p-value implies.

Crossover Design: Benefits and Traps

A crossover design is efficient because each patient serves as their own control, reducing between-subject variability. But it carries specific risks.

Carryover effects. Liothyronine (T3) has a half-life of roughly one day, but the downstream effects on gene expression, mood regulation, and HPT axis resetting may persist longer than the drug's serum clearance. The trial used no washout period between arms. If residual T3 effects carried into the T4-only period, scores during the second arm could be contaminated regardless of treatment assignment.

Period effects. Patients randomized to receive T4+T3 first may have experienced novelty or expectation effects that differed from those who received T4 alone first. The original publication did not report a formal test for treatment-by-period interaction, a standard check in crossover analyses.

Unblinding risk. T3 produces more noticeable physiological sensations than T4 in some patients (slight heart-rate increase, warmth, subjective energy). If patients could sense which arm they were in, the double-blind would be compromised. The trial did not include a blinding-assessment questionnaire.

Five Weeks Per Arm: Too Short?

Each treatment period lasted five weeks. That is long enough for T4 and T3 serum levels to reach approximate steady state. But it may not be long enough for stable psychometric outcomes.

Mood and cognitive instruments are sensitive to short-term environmental factors (seasonal changes, work stress, life events). In a 10-week total study, these confounders have limited time to average out. The American Thyroid Association's 2012 guidelines on combination therapy specifically noted that longer treatment durations are needed to assess psychological endpoints reliably, recommending a minimum of three to four months per arm for future trials.

Enrollment Biases and Generalizability

The participants were drawn from a Lithuanian endocrinology clinic. All 33 completers were women. This is consistent with the sex distribution of hypothyroidism but means the trial provides zero direct evidence for men.

Additional enrollment constraints that limit generalizability:

Geography and genetics. Iodine status, dietary patterns, and DIO2 polymorphism prevalence vary across populations. Lithuanian patients in the late 1990s may not represent North American, Western European, or East Asian thyroid cohorts.
Etiology not stratified. The trial included patients with hypothyroidism from various causes. Whether the T3 response differs between autoimmune (Hashimoto's) and post-surgical or post-radioiodine hypothyroidism was not examined.
TSH range. Participants were on stable T4 with TSH in the normal range. The results say nothing about patients with suppressed TSH, patients on high-dose T4 for thyroid cancer, or subclinical hypothyroidism.

The Dose-Ratio Question

Bunevicius replaced 50 mcg of T4 with 12.5 mcg of T3, assuming a roughly 4:1 potency ratio. The actual bioequivalence ratio between T4 and T3 remains debated. Estimates range from 3:1 to 5:1 depending on the pharmacokinetic model used. If the substitution slightly over-replaced thyroid activity, the T4+T3 arm may have been mildly hyperthyroid at the tissue level, producing mood elevation that had nothing to do with T3's unique central effects.

Free T4 levels dropped significantly in the combination arm, while total T3 levels rose. TSH remained in range for most patients. But TSH is a blunt instrument for detecting subtle tissue-level thyroid excess, particularly in the brain, where local deiodinase activity modulates T3 availability independently of serum markers.

Conflict of Interest and Funding Context

The trial was funded in part by SmithKline Beecham (later GlaxoSmithKline), which manufactured Cytomel (brand-name liothyronine). This does not invalidate the results. But it is a disclosure that later critics, including authors of the 2006 Escobar-Morreale meta-analysis, flagged as relevant context when weighing the trial's influence on prescribing.

Industry funding of small, positive trials can create publication bias when negative trials from independent groups receive less attention. In this case, the opposite eventually happened: multiple larger, independently funded trials (Clyde 2003, Sawka 2003, Walsh 2003, Saravanan 2005) failed to replicate the Bunevicius results, which helped correct the evidence base. But for the four years between 1999 and 2003, the Bunevicius trial was the dominant data point, and prescribing of T3 combination therapy rose during that window.

What Subsequent Commentary Raised

Letters to the editor and published commentaries after the 1999 paper raised several additional points:

Lack of a placebo arm. Both treatment arms contained active thyroid hormone. There was no inert-placebo control, so it is impossible to separate drug effect from expectation.
Outcome hierarchy. The trial did not designate a single primary endpoint. When every positive result among many tested endpoints can be presented as "the finding," the risk of cherry-picking increases.
Patient preference as evidence. The 19-of-33 preference figure is frequently cited. But preference is a subjective endpoint vulnerable to recall bias, especially in a crossover without a washout. Patients may have preferred whichever arm they received second simply because of adaptation to study procedures.
No long-term safety data. Five weeks of T3 exposure provides no information about sustained cardiac safety, bone density effects, or long-term mood stability.

The Replication Record

The most significant limitation of the Bunevicius trial is external: subsequent, larger trials did not confirm its central claims.

| Trial | Year | N | Duration | T4+T3 Superior? | |---|---|---|---|---| | Bunevicius et al. | 1999 | 33 | 5 wk/arm | Yes | | Clyde et al. | 2003 | 46 | 4 mo/arm | No | | Sawka et al. | 2003 | 40 | 15 wk/arm | No | | Walsh et al. | 2003 | 101 | 10 wk/arm | No | | Saravanan et al. | 2005 | 697 | 12 mo | No (primary); yes (some secondary) | | Escobar-Morreale meta-analysis | 2006 | Pooled | Various | No overall benefit |

The 2012 ATA/AACE guidelines concluded that existing evidence did not support routine combination therapy, while leaving the door open for future trials with better designs. The European Thyroid Association's 2012 position was slightly more permissive, suggesting a time-limited individual trial could be considered for patients with persistent symptoms despite optimal T4 dosing.

What the Trial Did Right

Listing only flaws would be unfair. The Bunevicius study was the first randomized, blinded trial to test a hypothesis that had circulated only as clinical anecdote. The crossover design was a reasonable choice for a pilot study with limited funding. The psychometric battery was broad enough to detect effects across cognitive and affective domains. And the trial generated a testable hypothesis that prompted over a dozen follow-up studies, which is exactly what pilot-scale work is supposed to do.

The problem was not the trial itself. The problem was treating a 33-patient pilot as definitive evidence for a practice change, a mistake made by prescribers and patients reading headlines, not by the original investigators.

The Bottom Line for Clinicians

The Bunevicius trial opened an important question. It did not answer it. Its limitations (tiny sample, no multiplicity correction, no washout, short duration, single-sex cohort, industry funding) are individually common in pilot research and collectively enough to prevent confident clinical conclusions. Clinicians considering T4+T3 combination therapy should base that decision on the full body of evidence, including the larger negative trials and the 2012 ATA guideline statement, not on this single study.

Frequently asked questions

›

References

Bunevicius R, Kazanavicius G, Zalinkevicius R, Prange AJ Jr. Effects of thyroxine as compared with thyroxine plus triiodothyronine in patients with hypothyroidism. N Engl J Med. 1999;340(6):424-429. PubMed
Escobar-Morreale HF, Botella-Carretero JI, Gomez-Bueno M, Galan JM, Barrios V, Sancho J. Thyroid hormone replacement therapy in primary hypothyroidism: a randomized trial comparing L-thyroxine plus liothyronine with L-thyroxine alone. Ann Intern Med. 2005;142(6):412-424. Meta-analysis: PubMed
Garber JR, Cobin RH, Gharib H, et al. Clinical practice guidelines for hypothyroidism in adults: cosponsored by the American Association of Clinical Endocrinologists and the American Thyroid Association. Thyroid. 2012;22(12):1200-1235. PubMed
Saravanan P, Simmons DJ, Visser TJ, Dayan CM. Randomized controlled trial testing the effect of thyroxine replacement with a combination of thyroxine plus triiodothyronine in patients with primary hypothyroidism. J Clin Endocrinol Metab. 2005;90(2):805-812.
Wiersinga WM, Duntas L, Fadeyev V, Nygaard B, Vanderpump MP. 2012 ETA guidelines: the use of L-T4 + L-T3 in the treatment of hypothyroidism. Eur Thyroid J. 2012;1(2):55-71.