REF-1454 · Röhrken et al., 2020 · Frontiers in Physiology · Paper · PubMed · Free full text

Polarized or a moderate mix for triathletes?

15 moderately trained, 6 weeks: no significant difference, with effect sizes leaning polarized. Told 35% between thresholds, they logged 20%. Pilot, too small to show equality.

Main caveat: Very small pilot; 7 and 8 per arm; underpowered for any likely difference.
limited, mixed evidenceMeasured on trained athletes

Röhrken, Held and Donath (German Sport University Cologne) allocated triathletes by minimisation to a polarized plan (INT, prescribed 92/0/8 by zones 1/2/3; high-intensity sessions as 30 s work, 15 s rest intervals) or a moderate plan (CON, prescribed 65/35/0), for 6 weeks with similar hours.

Incremental running and cycling tests were done before and after, after tapering 7 ± 2 days. Completed distributions were 75.2/11.1/13.7 (INT) and 77.8/20.3/2.0 (CON); only zone 3 differed significantly. No time × group interaction was significant.

Combining both groups, running performance at the second lactate threshold rose 2.8 ± 5.1% (p = 0.026) and peak running 5.4 ± 6.2% (p = 0.002). Across all athletes, more time in zone 2 went with smaller changes in cycling LT1 (r = −0.61) and peak running (r = −0.52). The authors conclude polarized "did not prove" better but note small effects in its favour.

Our reading: a null in a very small sample, whose point estimates all favour polarized, so it does not show the moderate plan was as good. It is useful for its compliance data: athletes given 35% between the thresholds logged 20%, and athletes given none logged 11%.

Those figures come from a heart-rate time-in-zone method that the authors note lags at high intensity, so part of the gap may be measurement. The cohort correlations run against time in zone 2 and are hypothesis-level.

Protocol

Participants: 35 enrolled; 15 analysed, INT n = 7 (2 women, 29.1 ± 7.6 y) and CON n = 8 (2 women, 30.3 ± 6.1 y); prior 11.7-12.0 h a week; end of season, one short-distance race allowed.
Allocation: minimisation by gender, age, training volume and a fitness index; not blinded; single centre.
Prescribed: INT 92/0/8 at 788 TRIMP a week (easy sessions plus 3 high-intensity sessions, one per discipline; 30 s work, 15 s rest on bike and run). CON 65/35/0 at 941 TRIMP a week (easy sessions plus sessions between LT1 and LT2: 4 × 8 min at 88-92% of LT2 power, a 120-min tempo ride at 80-90% of LT2 power, 4 × 1,600 m at 95-98% of LT2 pace, a progressive run, a 20-min tempo brick run, 20 × 100 m swimming). Anchored on lactate thresholds, not FTP.
Monitoring: heart-rate time in zone, filled with power or pace if missing, session RPE for swimming.
Tests: incremental running and cycling to exhaustion; LT1, LT2 and peak performance.
Analysis: repeated-measures ANOVA with baseline as covariate; SMD; Pearson correlations.
Funding and interests: no funding statement printed; authors declare no commercial or financial relationships.

Key Results

Completed trainingINT (n = 7)CON (n = 8)p
Hours a week10.8 ± 2.410.0 ± 2.70.580
Zone 1 (%)75.2 ± 14.477.8 ± 11.90.704
Zone 2 (%)11.1 ± 10.920.3 ± 10.80.118
Zone 3 (%)13.7 ± 4.12.0 ± 1.5<0.001
TRIMP a week882.0 ± 155739.0 ± 1620.106
Completed to prescribed load (%)112.0 ± 14.986.2 ± 10.00.001
ChangeINT (%)CON (%)SMDGroup comparison p (Table 4)
Run LT14.2 ± 9.5−1.1 ± 6.60.470.160
Run LT24.2 ± 4.21.5 ± 5.80.500.416
Run peak7.2 ± 3.83.8 ± 7.60.650.346
Bike LT17.4 ± 19.5−2.2 ± 10.70.710.234
Bike LT24.5 ± 12.0−0.7 ± 8.20.490.197
Bike peak2.7 ± 9.6−5.7 ± 8.20.890.125
Individual response: SDs of change are large in both arms; no control arm without training, and some athletes in each arm gained more than 10%. Not evidence of responders.
Numbers that do not add up: dropouts are 3 INT and 2 CON in Participants and 3 INT and 1 CON in the limitations, while 35 enrolled and 15 were analysed, leaving most losses unexplained (our arithmetic); session counts differ between Table 2 and the text; Practical Implications opens "a polarized TID may be superior" and then says it gave no benefit; the Conclusion says LT1 running improved where the Results say LT2.

Mechanisms

Measured: shared within-sample running gains; no significant interaction; inverse correlations between %zone 2 and some changes (r = −0.61, r = −0.52).
Authors' interpretation: both programmes may benefit; the dropouts in INT (two for muscular fatigue and the onset of illness symptoms) suggest that 2-3 high-intensity sessions a week may be a substantial stressor.
Inference: with zone 1 about equal in the two arms, this tested mainly 14% vs 2% high intensity on top of a similar easy base, more than it tested tempo.

Practical Application

Polarized or a moderate mix for triathletes? 15 moderately trained, 6 weeks: no significant difference, with effect sizes leaning polarized. Told 35% between thresholds, they logged 20%. Pilot, too small to show equality.
Prescribed is not completed: check what was actually done before judging a distribution; both arms here missed by a wide margin. Not from this paper; the base's own practice: keep the easy easy (SYNTH-26), and use the middle zone as a tool, not a diet (SYNTH-27).
What it does not show: that moderate training is as good as polarized, a difference in swimming or racing, or drift as a settled size (heart-rate lag may inflate or deflate it). It tested no sweet-spot format.

Limitations

Very small pilot; 7 and 8 per arm; underpowered for any likely difference.
Heavy attrition: 35 enrolled, 15 analysed.
Arms overlapped: zone 1 and zone 2 not significantly different as completed.
Load differed: prescribed load favoured CON and completed load favoured INT, neither significantly.
Measurement: heart-rate time in zone can underreport high-intensity time, as the authors note.
No swimming test; end of season with one race allowed.
Reporting inconsistencies, noted under Key Results.
Funding: not printed; no commercial or financial conflict declared.
How this entry was checked and changed
Why this level

the rubric's B for a parallel RCT under 20, lowered a step for limits beyond design: 35 enrolled and 15 analysed with most losses unexplained, no blinding, and arms whose completed distributions overlap

Full text (checked 2026-10-30)
Thirty-five participants who fulfilled the inclusion criteria were enrolled; Three athletes of INT and two athletes of CON needed to be excluded; the high dropout rate (INT n = 3, CON n = 1) (full text)
Prescribed TID (%LIT/%MIT/%HIT) 92/0/8 and 65/35/0 (full text)
Completed Z1 75.2 ± 14.4 and 77.8 ± 11.9; Z2 11.1 ± 10.9 and 20.3 ± 10.8 (p 0.118); Z3 13.7 ± 4.1 and 2.0 ± 1.5 (<0.001) (full text)
Completed to Prescribed Load (%) 112.0 ± 14.9 and 86.2 ± 10.0 (full text)
Run LT2 (%) 4.2 ± 4.2 and 1.5 ± 5.8, SMD 0.50; Bike Peak (%) 2.7 ± 9.6 and −5.7 ± 8.2, SMD 0.89 (full text)
All participants, i.e., taking both groups as one group, improved their Peak and LT2 performance in running (Peak: + 5.4 ± 6.2 % p = 0.002 LT2: + 2.8 ± 5.1 % p = 0.026) (full text)
An inverse correlation between %MIT to cycling LT1 performance changes (r = −0.61, p = 0.009) and %MIT to running Peak performance changes (r = −0.52, p = 0.049) (full text)
A time-in-zone method displays delayed heart-rate elevations and underreports time in HIT compared to a sessions-goal method (full text)