How many patients completed the Bunevicius 1999 trial?

Thirty-three patients completed both treatment periods. The paper does not report how many were screened or enrolled before dropouts, as it predates CONSORT reporting standards.

Was there a washout period between treatment arms?

No. Patients transitioned directly from one treatment period to the next without a washout. Given levothyroxine's 6-7 day half-life, residual drug from Period 1 would have persisted into early Period 2.

What was the T4-to-T3 substitution ratio used?

The investigators replaced 50 mcg of levothyroxine with 12.5 mcg of liothyronine, reflecting an approximate 4:1 potency ratio. This kept total thyroid hormone exposure roughly equivalent between arms.

Did the trial correct for multiple comparisons?

No. Seventeen outcome measures were tested without Bonferroni correction or any other multiplicity adjustment. Six reached nominal significance at p < 0.05, but the adjusted threshold would be approximately p < 0.003.

Why haven't later trials replicated the Bunevicius findings?

Possible explanations include type I error in the original, population heterogeneity in replication attempts, varying T3 doses and durations, and the possibility that only a genetic subgroup (DIO2 polymorphism carriers) benefits from combination therapy.

What cognitive tests improved with combination therapy?

The Digit Symbol subtest of the WAIS improved significantly. Memory tests (logical memory, visual reproduction, digit span) and the Trail Making Test did not reach significance.

Is this trial relevant to current clinical guidelines?

Yes. The 2022 ETA guidelines and the 2014 ATA guidelines both reference this trial when discussing combination therapy as an option for selected patients who remain symptomatic on T4 alone, though neither recommends routine use.

What does the crossover design mean for interpreting results?

Each patient received both treatments, eliminating between-person variability. This increases statistical power for a given sample size but introduces carryover risk if effects from Period 1 bleed into Period 2.

Were thyroid hormone levels different between treatment periods?

Yes. Free T3 was higher and free T4 was lower during combination therapy, as expected from the substitution. TSH remained within the normal range in both periods, confirming biochemical equivalence at the pituitary level.

Should patients use this trial to request T3 from their doctor?

This trial provides a rationale for discussion, not a mandate for prescribing. Patients symptomatic despite normal TSH on levothyroxine may reasonably discuss a supervised combination trial with their endocrinologist, but should understand that larger trials have not consistently confirmed the benefit seen here.

Inside the Bunevicius T4+T3 Methodology: What Most Summaries Skip

Q: Why haven't later trials replicated the Bunevicius findings?

Possible explanations include type I error in the original, population heterogeneity in replication attempts, varying T3 doses and durations, and the possibility that only a genetic subgroup (DIO2 polymorphism carriers) benefits from combination therapy.

Q: What cognitive tests improved with combination therapy?

The Digit Symbol subtest of the WAIS improved significantly. Memory tests (logical memory, visual reproduction, digit span) and the Trail Making Test did not reach significance.

Q: Is this trial relevant to current clinical guidelines?

Yes. The 2022 ETA guidelines and the 2014 ATA guidelines both reference this trial when discussing combination therapy as an option for selected patients who remain symptomatic on T4 alone, though neither recommends routine use.

Q: What does the crossover design mean for interpreting results?

Each patient received both treatments, eliminating between-person variability. This increases statistical power for a given sample size but introduces carryover risk if effects from Period 1 bleed into Period 2.

Q: Were thyroid hormone levels different between treatment periods?

Yes. Free T3 was higher and free T4 was lower during combination therapy, as expected from the substitution. TSH remained within the normal range in both periods, confirming biochemical equivalence at the pituitary level.

Q: Should patients use this trial to request T3 from their doctor?

This trial provides a rationale for discussion, not a mandate for prescribing. Patients symptomatic despite normal TSH on levothyroxine may reasonably discuss a supervised combination trial with their endocrinologist, but should understand that larger trials have not consistently confirmed the benefit seen here.

By HealthRX.com Medical Team

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

Clinical image for Inside the Bunevicius T4+T3 Methodology: What Most Summaries Skip Image: HealthRX.com clinical image

At a glance

| Parameter | Detail | |-----------|--------| | N | 33 (completers) | | Intervention | Usual T4 dose minus 50 mcg, plus liothyronine 12.5 mcg | | Comparator | Usual T4 dose (full levothyroxine monotherapy) | | Design | Randomized, double-blind, crossover | | Duration | 5 weeks per period, no washout | | Primary endpoint | Multiple neuropsychological and mood scales (see below) | | Key result | Combination therapy superior on 6 of 17 cognitive/mood measures |

The crossover architecture and what it implies

Bunevicius et al. chose a crossover design rather than a parallel-group RCT. Each patient served as their own control, receiving both treatments in random order. This is a defensible choice for a small sample because it eliminates between-subject variability in baseline thyroid status, dose requirements, and neuropsychological performance. The tradeoff: any carryover effect from Period 1 contaminates Period 2.

The trial allocated no washout between periods. Because levothyroxine has a half-life of roughly 6 to 7 days, residual T4 from the first period would persist into week one of the second period. Liothyronine, with its much shorter half-life (approximately 1 day), clears faster. This asymmetry means that patients crossing from combination therapy back to T4-only might carry a brief residual T3 effect, while those crossing in the opposite direction would not. The authors did not report a formal test for period-by-treatment interaction, which is the standard analytic guard against carryover in crossover trials.

Randomization and blinding details

The original paper states that randomization was performed by the hospital pharmacy. Patients received identically encapsulated tablets for both periods. Blinding was maintained for participants, investigators, and the psychologist administering tests. The pharmacy held the allocation code until data collection was complete.

What the paper does not report: the method of sequence generation (block randomization, simple randomization, or coin flip), allocation concealment beyond pharmacy custody, or whether any unblinding events occurred. For 33 subjects, even one unblinded case represents roughly 3% of the dataset. The absence of a CONSORT-style flow diagram (published before CONSORT became standard) means we cannot confirm how many patients were screened, how many declined, or whether dropouts occurred before completion.

Inclusion and exclusion: a narrow slice of hypothyroidism

Eligible patients were adults with overt hypothyroidism already on stable levothyroxine replacement for at least the preceding 3 months, with TSH in the normal range. This design selects for adherent, biochemically euthyroid patients on established doses. It explicitly excludes the more heterogeneous population of newly diagnosed or poorly controlled patients.

Key exclusions: psychiatric illness requiring medication, pregnancy, and use of medications known to affect thyroid hormone metabolism. By excluding active psychiatric disease, the trial tests whether combination therapy improves cognition and mood in patients who are "well" at baseline. This ceiling-effect problem becomes relevant when interpreting the results: the room for improvement is narrow in already-euthyroid, psychiatrically healthy adults.

The endpoint problem: 17 instruments, no declared primary

This is the single most important methodological detail that most summaries omit. The trial administered 17 separate neuropsychological and mood measures, including the Profile of Mood States (POMS), Visual Analogue Scales (VAS) for mood, the Wechsler Memory Scale, the Trail Making Test, digit span, and others. The publication does not designate one of these as the primary endpoint.

When 17 outcomes are tested at alpha = 0.05, the probability of at least one false positive (assuming all null hypotheses are true) is approximately 58%. Finding 6 out of 17 tests significant is more compelling than finding 1, but the lack of multiplicity correction (no Bonferroni, no Holm, no false discovery rate adjustment) means the nominal p-values are liberal estimates of the true error rate.

The authors reported improvements in:

POMS total mood disturbance
Anger-hostility subscale
Depression subscale
Confusion-bewilderment subscale
Composite score of 11 VAS items
Digit Symbol subtest of the WAIS

No correction was applied across these comparisons. A Bonferroni-adjusted threshold for 17 tests would be p < 0.003 rather than p < 0.05. The paper does not provide exact p-values for all comparisons, making post hoc recalculation difficult.

Statistical approach and the estimand question

The analysis used paired t-tests comparing end-of-period scores between conditions. This is appropriate for a crossover design under the assumption of no carryover and no period effect. The estimand is implicitly: "the average treatment effect of combination therapy versus monotherapy on each scale score, within the same patient, after 5 weeks of exposure."

What this estimand does not capture:

Duration dependence. Five weeks may be insufficient for T3-mediated neuroplastic or receptor-occupancy changes to reach steady state. The ATA 2014 guidelines later noted that longer treatment durations might be necessary to detect meaningful cognitive differences.
Responder heterogeneity. Paired means can mask a subgroup of strong responders diluted by non-responders. The paper did not report individual-level response data or test for effect modification by baseline mood scores.
Clinical significance. Statistical significance on a subscale of the POMS does not automatically equate to a clinically meaningful change. The paper does not report minimal clinically important differences (MCIDs) for any instrument used.

The comparator choice and dose substitution logic

Rather than adding T3 on top of existing T4 (which would raise total thyroid hormone exposure), the investigators subtracted 50 mcg of levothyroxine and replaced it with 12.5 mcg of liothyronine. The substitution ratio (50 mcg T4 : 12.5 mcg T3) approximates a 4:1 potency assumption, roughly consistent with the FDA-approved prescribing information for liothyronine, which suggests T3 is 3 to 4 times as potent as T4 microgram-for-microgram.

This approach kept serum TSH within the normal range in both arms. Mean TSH values did not differ significantly between periods. However, free T3 levels were (as expected) higher during combination therapy, while free T4 was lower. The trial confirmed that the substitution was biochemically "safe" in the short term but did not address whether long-term T3 exposure at this ratio would maintain euthyroidism without TSH drift.

What the results actually showed

| Measure | T4 alone (mean ± SD) | T4+T3 (mean ± SD) | p-value | |---------|----------------------|---------------------|---------| | POMS Total Mood Disturbance | 28.6 ± 35.7 | 17.3 ± 27.2 | 0.03 | | POMS Depression | 7.6 ± 8.3 | 4.6 ± 5.9 | 0.03 | | POMS Anger | 7.4 ± 8.4 | 5.0 ± 5.3 | 0.04 | | Digit Symbol (WAIS) | 55.2 ± 12.6 | 59.4 ± 11.7 | 0.009 | | VAS Composite | 53.0 ± 23.5 | 43.8 ± 22.3 | 0.04 |

Non-significant results included the Trail Making Test (Part A and B), Digit Span, logical memory, visual reproduction, and several individual VAS items. The selective pattern of significance (mood and processing speed improved, but memory and executive function did not) suggests a pharmacological signal rather than a global placebo-driven improvement, but this interpretation is post hoc.

Acknowledged limitations and the replication problem

The authors themselves noted the small sample size and suggested larger confirmatory trials. Over the next decade, at least 11 randomized trials attempted to replicate these findings. The majority failed to demonstrate a consistent advantage of combination therapy over T4 monotherapy for cognitive or mood outcomes. A 2006 meta-analysis by Grozinsky-Glasberg et al. concluded that available evidence did not support routine use of T4+T3.

Why the discrepancy? Several possibilities:

The original finding was a type I error amplified by multiple testing without correction.
Population differences. Later trials enrolled broader populations, potentially diluting a signal present only in a cognitive-complaint subgroup.
Dose and ratio differences. Later trials used varying T3 doses and substitution ratios, making direct comparison difficult.
Duration. Some replication attempts used shorter periods (4 weeks); others used longer ones (up to 12 weeks in the Saravanan 2005 trial).
Deiodinase polymorphisms. Post hoc subgroup analyses from the WATTS trial suggest that patients with the DIO2 Thr92Ala polymorphism may preferentially benefit from combination therapy, a hypothesis that did not exist in 1999.

How design choices shaped interpretation

The Bunevicius trial occupies an unusual position in endocrinology. It is simultaneously the most-cited evidence for combination therapy and the least methodologically rigorous among the RCTs in this space. Its crossover design was appropriate for the sample size but introduced carryover risk. Its lack of a primary endpoint inflated the chance of finding something significant. Its 5-week treatment periods may have been too short. Its exclusion of psychiatric comorbidity limited generalizability.

The trial's lasting contribution is not the definitive proof of combination therapy's superiority. It is the hypothesis that intracellular T3 concentrations in the brain may not be fully normalized by T4 monotherapy alone, and that a subset of hypothyroid patients may benefit from exogenous T3. That hypothesis remains open, with the 2022 ETA guidelines acknowledging combination therapy as a reasonable trial in selected patients who remain symptomatic despite normal TSH.

The bottom line for clinicians reading this trial

Read Bunevicius 1999 as a hypothesis-generating study, not a confirmatory one. The signal is real enough to have sustained two decades of research. But the methodology, specifically the absence of multiplicity correction across 17 endpoints and the unaddressed carryover risk, means the effect size estimates carry wide uncertainty. Patients asking about T4+T3 combination therapy deserve to know that this single trial of 33 patients, while well-designed for its era, has not been consistently replicated in larger studies.

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
Jonklaas J, Bianco AC, Bauer AJ, et al. Guidelines for the treatment of hypothyroidism: prepared by the American Thyroid Association task force. Thyroid. 2014;24(12):1670-1751. PubMed
Grozinsky-Glasberg S, Fraser A, Nahshoni E, Weizman A, Leibovici L. Thyroxine-triiodothyronine combination therapy versus thyroxine monotherapy for clinical hypothyroidism: meta-analysis of randomized controlled trials. J Clin Endocrinol Metab. 2006;91(7):2592-2599. PubMed
Saravanan P, Simmons DJ, Visser TJ, Dayan CM. Randomized controlled trial examining the effects of thyroxine plus triiodothyronine combination therapy in hypothyroidism. J Clin Endocrinol Metab. 2005;90(2):805-812. PubMed
Wouters HJCM, van Loon HCM, van der Klauw MM, et al. No effect of the Thr92Ala polymorphism of deiodinase-2 on thyroid hormone parameters, health-related quality of life, and cognitive functioning in a large population-based cohort study. Thyroid. 2017;27(2):147-155. PubMed
Wiersinga WM, Duntas L, Fadeyev V, Nygaard B, Vanderpump MPJ. 2012 ETA guidelines: the use of L-T4 + L-T3 in the treatment of hypothyroidism. Eur Thyroid J. 2012;1(2):55-71. PubMed