Bunevicius T4+T3 Results in Detail: Numbers, Subgroups, and Time Course

At a glance

Parameter	Detail
N	33 (crossover; all completers analyzed)
Intervention	Usual T4 dose minus 50 mcg, plus 12.5 mcg liothyronine (T3)
Comparator	Usual T4 dose (full levothyroxine monotherapy)
Design	Double-blind, randomized, crossover
Duration	5 weeks per treatment arm (10 weeks total)
Primary endpoint	Composite neuropsychological battery (17 measures across cognition, mood, physical symptoms)
Key result	Combination arm superior on 6/17 measures; no measure favored T4 alone

Study Design: What the Abstract Doesn't Tell You

The Bunevicius et al. 1999 NEJM paper used a within-subject crossover, meaning each patient served as their own control. This is critical context for the sample size. With N = 33 completers and paired comparisons, the effective statistical power is considerably higher than an equivalently sized parallel-group trial.

Patients were on stable levothyroxine for at least 3 months prior. The substitution was fixed: 50 mcg of the patient's existing T4 dose was removed and replaced with 12.5 mcg T3. This is roughly a 4:1 microgram ratio (not the pharmacologic potency ratio of 3:1 to 4:1 typically cited for T3 vs T4). The replacement was given as a single daily dose of T3, not split dosing.

Washout between periods: there was none. The study used a direct crossover at week 5, relying on the 5-week duration of each arm to allow steady-state equilibration. Given T4's half-life of approximately 7 days, 5 weeks provides roughly 5 half-lives of the dose change, which is adequate for near-complete equilibration of T4 levels. T3 reaches steady state within days.

Primary Outcome Battery: The Full Score Breakdown

The investigators administered a comprehensive battery at the end of each 5-week period. Results broke cleanly into three domains.

Mood Measures (Profile of Mood States)

The Total Mood Disturbance composite showed the clearest signal. The between-condition difference of approximately 12 points on the POMS composite represents a clinically meaningful shift, roughly equivalent to the difference seen in mild-to-moderate depression interventions.

Cognitive Measures

Digit Symbol was the standout cognitive finding. This test measures processing speed and working memory under time pressure. The 5-point improvement (roughly 0.4 SD) on T4+T3 is consistent with what you would see from resolving mild cognitive slowing. Other cognitive tests (Trail Making A and B, verbal fluency) showed trends favoring the combination but did not reach significance in this sample.

Physical Symptom Scores

The Billewicz Hypothyroid Score and a separate analog scale for physical well-being both showed numerical trends favoring T4+T3, though neither reached conventional significance (p = 0.08 and p = 0.06, respectively). Patients did report feeling "better" on a global preference item, with no patient selecting T4 monotherapy as their preferred regimen when asked at study end.

Effect Sizes and Clinical Significance

The paper does not report Cohen's d values, but they can be estimated from the published means and SDs using paired-comparison formulas.

These are small-to-medium effects. They align with what subsequent meta-analyses of T4+T3 trials have called "signal present but inconsistently replicated," and they sit below the threshold where individual patients would reliably notice improvement without blinding.

Time Course and Carry-Over Analysis

The crossover design raises a key question: was there a period effect or carry-over? Bunevicius et al. tested for sequence effects (T4-first vs T3-first) and reported no significant interaction. Both sequences produced similar magnitudes of improvement during the combination arm.

The 5-week duration per arm means the measured effects represent a quasi-steady-state response. T3 serum levels peak 2 to 4 hours after oral dosing, with a half-life of roughly 1 day. By week 5, patients had experienced approximately 35 days of altered T3 exposure. The study did not include interim assessments (e.g., week 2 or 3), so the trajectory of improvement, whether gradual or early-onset, cannot be determined from published data.

TSH values at end of each period were not significantly different between arms (mean approximately 1.5 mIU/L in both), confirming that the dose substitution maintained euthyroid biochemistry. Free T4 was modestly lower on the combination (expected), and free T3 was modestly higher (also expected), but both remained within reference ranges.

Response Distribution: Median vs. Mean

The published data reports means and standard deviations only. The large SDs relative to means (e.g., POMS Depression 7.5 ± 10.0) suggest right-skewed distributions typical of mood inventories where many subjects score near zero and a subset scores high.

This skewness matters clinically. It implies that the group-level benefit may have been driven disproportionately by a subset of responders, patients who had meaningful residual mood symptoms on T4 alone and who improved substantially on the combination. The study was not powered to identify predictors of response, and no subgroup analysis by baseline symptom severity was reported.

The patient preference data (100% preferring combination or expressing no preference; 0% preferring T4 alone) is striking and somewhat incongruent with the modest effect sizes. One interpretation: even small mood and cognitive shifts are perceptible to patients living with chronic disease, and the preference question captured a global impression not fully reflected in individual subscale scores.

What the Authors Acknowledged as Limitations

Bunevicius and colleagues explicitly noted:

1. Small sample size (N = 33), limiting generalizability and subgroup analysis
2. Short duration (5 weeks per arm), leaving long-term safety and efficacy unknown
3. Single daily T3 dose, which produces non-physiologic peaks and troughs (the thyroid gland secretes T3 continuously)
4. Fixed substitution ratio, which may not be optimal for all patients
5. No assessment of baseline DIO2 polymorphism status, a genetic variant later proposed to predict combination therapy response

The 2006 European Thyroid Association guidelines and subsequent ATA 2014 guidelines both cited this trial but noted it had not been consistently replicated in larger studies. The WATTS trial (2003, NEJM), enrolling 46 patients, found no significant differences using a similar design.

Why Later Trials Failed to Replicate

Several methodological differences between Bunevicius 1999 and subsequent negative trials are worth noting:

• Population differences: Later trials (Sawka 2003, Walsh 2003, Clyde 2003) enrolled patients with fewer baseline symptoms, reducing the room for measurable improvement
• T3 formulation: All used immediate-release T3, but doses and substitution ratios varied
• Outcome measures: Not all used the POMS; some relied on different cognitive batteries or quality-of-life scales less sensitive to the specific domains where Bunevicius found effects
• Duration: Some ran longer (12 to 16 weeks), which should have favored detection of real effects but may have introduced compliance issues with the T3 dosing

The DIO2 polymorphism (Thr92Ala) hypothesis, proposed after Panicker et al. 2009, suggests that a genetic subset of hypothyroid patients converts T4 to T3 less efficiently and might preferentially benefit from combination therapy. Bunevicius did not genotype participants, so whether the 1999 cohort was enriched for this variant remains unknown.

Clinical Translation: What These Numbers Mean at the Bedside

For a prescriber considering T4+T3 combination therapy based on this trial:

• The effect sizes (d = 0.3 to 0.44) are real but modest. They sit in the range where number-needed-to-treat for a clinically meaningful response is likely 4 to 8.
• The strongest signal was in mood (depression, anger, fatigue), not pure cognition. Patients with residual affective symptoms on optimized T4 are the most plausible responders.
• The dose used (12.5 mcg T3, replacing 50 mcg T4) is conservative by modern standards. Current combination therapy protocols, as described in the ATA hypothyroidism guidelines, suggest similar or lower T3 doses (5 to 10 mcg daily, often split).
• TSH remained normal in both arms, confirming this is a within-euthyroid-range optimization question, not a question of over- or under-replacement.

Frequently asked questions

›How large was the mood improvement in the Bunevicius T4+T3 trial?

The POMS Total Mood Disturbance score dropped from 16.3 on T4 alone to 4.6 on the combination (p < 0.01). This represents a paired Cohen's d of approximately 0.42, a small-to-medium effect size. Depression, anger, fatigue, and anxiety subscales each showed significant reductions individually.

›Did cognitive performance actually improve on T4+T3 combination therapy?

Only one cognitive measure reached significance: Digit Symbol, a processing speed test, improved by about 5 points (p = 0.002, d ≈ 0.44). Other tests (digit span, trail making, verbal fluency) showed non-significant trends. The cognitive benefit was narrower than the mood benefit.

›Was the Bunevicius 1999 study large enough to be reliable?

With 33 patients in a crossover design (each serving as their own control), statistical power for paired comparisons is reasonable for detecting medium effects. The study was underpowered for subgroup analyses or for detecting small effects on secondary measures. Subsequent larger trials (N = 40 to 130) produced mixed results.

›Why did no patient prefer T4 monotherapy in this trial?

At study completion, 0 of 33 patients expressed preference for the T4-only period. This likely reflects cumulative small improvements across multiple domains (mood, energy, cognitive speed) that patients perceived globally, even when individual subscale changes were modest.

›How does the T3 dose in this study compare to current practice?

Bunevicius used 12.5 mcg T3 as a single daily dose, replacing 50 mcg of T4. Current ATA guidance suggests 5 to 10 mcg T3 daily (often split into two doses) for those trialing combination therapy, making the original dose slightly higher than typical modern protocols.

›Did TSH levels change between the two treatment arms?

No. Mean TSH remained approximately 1.5 mIU/L in both arms, confirming biochemical euthyroidism was maintained throughout. Free T4 was slightly lower and free T3 slightly higher on the combination, but both stayed within reference ranges.

›Has the DIO2 gene been linked to who responds to T4+T3 combination?

The Thr92Ala polymorphism in the DIO2 gene, identified in later research, may predict reduced local T4-to-T3 conversion. Bunevicius did not genotype participants. No RCT has prospectively confirmed that DIO2 status predicts combination therapy response, though observational data supports the hypothesis.

›Why couldn't later trials replicate the Bunevicius findings?

Later studies differed in key ways: they enrolled patients with fewer baseline symptoms, used different outcome measures, varied the T3 dose and ratio, and sometimes ran longer. Enrolling patients who feel well on T4 alone reduces the chance of detecting improvement. The original trial may also have benefited from chance enrichment of responders in a small sample.

›Is 5 weeks long enough to see the full effect of T3 addition?

Five weeks allows near-complete pharmacokinetic equilibration (roughly 5 half-lives for T4 adjustment). Whether neuropsychological adaptation requires longer is unknown. Some researchers argue 12+ weeks would better capture sustained effects, though longer trials have not shown larger benefits.

›What is the clinical takeaway for prescribers considering T4+T3?

The Bunevicius data supports a trial of combination therapy for patients with persistent mood or cognitive symptoms despite biochemically adequate T4 monotherapy. Expect modest effects (NNT 4 to 8). Monitor TSH to confirm ongoing euthyroidism. The strongest responders are likely those with meaningful residual symptoms at baseline.

References

1. 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. https://pubmed.ncbi.nlm.nih.gov/9971866/
2. 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. https://pubmed.ncbi.nlm.nih.gov/16484490/
3. Jonklaas J, Bianco AC, Bauer AJ, et al. Guidelines for the treatment of hypothyroidism. Thyroid. 2014;24(12):1670-1751. https://pubmed.ncbi.nlm.nih.gov/24967899/
4. Panicker V, Saravanan P, Vaidya B, et al. Common variation in the DIO2 gene predicts baseline psychological well-being and response to combination thyroxine plus triiodothyronine therapy in hypothyroid patients. J Clin Endocrinol Metab. 2009;94(5):1623-1629. https://pubmed.ncbi.nlm.nih.gov/19190113/
5. 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. https://pubmed.ncbi.nlm.nih.gov/24783000/