The interference effect: what the meta-analyses show

For most people, running doesn't measurably blunt maximal strength or muscle size. The biggest recent meta-analysis, 43 studies and 1,090 people, found effects of -0.06 for maximal strength and -0.01 for muscle size, both indistinguishable from zero.¹ What does take a hit is explosive strength, and in one analysis, lower-body strength in already-trained lifters who ran and lifted in the same session.²

So "cardio kills gains" is wrong as a slogan. "Some cardio, done carelessly, costs some of some gains" is closer, and less catchy. The details matter, so here they are.

Where the idea came from#

In 1980 Robert Hickson had one group lift five days a week, one do endurance work six days a week, and one do both.³ For seven weeks the combined group gained leg strength as fast as the lifters. Then their strength levelled off and fell in weeks nine and ten. Their VO2max rose as much as the endurance group's.

It's a good study and it named the effect. It's also one study, of roughly eleven sessions a week for ten weeks. That's a heavy load to generalise from.

What the effect sizes mean#

Most numbers below are standardised mean differences (SMD), sometimes written ES. They compare a group that lifted and did aerobic work against a group that did the identical lifting alone. Zero means no difference. Negative means the concurrent group gained less.

The unit is the spread between people, not kilos. An SMD of -0.28 means the concurrent group's average gain sat about a quarter of a standard deviation below the lifting-only group's. That's a shift in a group average. It says little about what happens to any one person. When the 95% confidence interval crosses zero, the data are compatible with no effect at all, and the honest reading is "not shown", not "shown to be zero".

One more thing that changes how to read them: the comparison is always against people doing the same lifting with no aerobic work. Nobody in these studies was asked to choose between a stronger squat and a faster 10K. In studies that measured it, like Hickson's, the concurrent groups got much fitter aerobically, and that gain isn't in these numbers.

The four meta-analyses at a glance#

They overlap. Several trials appear in more than one, so four papers don't mean four independent confirmations. Schumann 2022 and Lundberg 2022 come from the same group and the same literature search.¹⁴

The 2012 meta-analysis#

Wilson and colleagues pooled 21 studies and 422 effect sizes.⁵ They compared lifting only, endurance only and concurrent training on their own pre-to-post gains.

Wilson 2012: training effect by group

	Value (effect size)
Strength: lifting only	1.76
Strength: concurrent	1.44
Hypertrophy: lifting only	1.23
Hypertrophy: concurrent	0.85
Power: lifting only	0.91
Power: concurrent	0.55

Mean pre-to-post effect sizes from Wilson et al. 2012 (21 studies). Larger is a bigger gain. Abstract figures.

Power was the only outcome where all three groups differed significantly from each other. For strength and size, lifting only and concurrent both beat endurance only. Two moderator findings from this paper still get quoted everywhere. First, lifting alongside running cost both size and strength, and lifting alongside cycling didn't. Second, more frequent and longer endurance sessions went with smaller gains: correlations of -0.26 to -0.35 for frequency and -0.29 to -0.75 for duration.

Correlations across studies aren't the same as a dose test inside one study. Keep that in mind when you read "keep your runs short".

The 2022 updates#

Strength, power and whole-muscle size#

Schumann and colleagues reran the question in 2022 with stricter rules: supervised training, at least four weeks, and a comparison group doing the identical lifting plan with no aerobic work.¹ Maximal strength: 37 studies, SMD -0.06 (95% CI -0.20 to 0.09). Muscle hypertrophy: 15 studies, -0.01 (-0.16 to 0.18). Explosive strength, meaning jumps, rate of force development and leg-press power: 18 studies, -0.28 (-0.48 to -0.08).

The explosive-strength cost was clearer when both were done in the same session, defined as 20 minutes or less apart (-0.31, -0.62 to -0.01). When sessions were at least 3 hours apart it wasn't significant. None of the other moderators mattered: running versus cycling, more or fewer than five weekly sessions, untrained versus active, under or over 40.

Muscle fibres#

Lundberg and the same group then looked one level down, at individual fibre size, across 15 studies and 300 people.⁴ Overall fibre growth was a little lower with concurrent training: -0.23 (-0.46 to -0.00), right on the line. Split by fibre type, neither result was significant on its own: -0.34 (-0.72 to 0.04) for type I and -0.13 (-0.39 to 0.12) for type II. Type I fibres, the slow-twitch ones, took a bigger hit when the aerobic work was running (-0.81, -1.26 to -0.36). That running result rests on just three studies; twelve of the fifteen used cycling. The authors call it preliminary, and they're right to.

Training status#

Petré and colleagues asked whether experience changes the answer, using lower-body 1RM (squat and leg press) from 27 studies and 750 people.² They counted someone as trained if they were an athlete or had done a structured programme for at least three months.

Effect of adding endurance work on strength, with 95% CIs

	Low (SMD)	High (SMD)	Mark (SMD)
Max strength, all (Schumann)	-0.2	0.09	-0.06
Muscle size, all (Schumann)	-0.16	0.18	-0.01
Explosive strength (Schumann)	-0.48	-0.08	-0.28
Fibre size (Lundberg)	-0.46	0	-0.23
Leg 1RM, untrained (Petré)	-0.29	0.35	0.03
Leg 1RM, trained (Petré)	-0.59	-0.11	-0.35
Trained, same session (Petré)	-1.08	-0.25	-0.66
Trained, sessions over 2 h apart (Petré)	-0.43	0.23	-0.1

Point estimate (mark) and 95% confidence interval. Schumann 2022 (43 studies), Lundberg 2022 (15 studies), Petré 2021 (27 studies). Below zero means smaller gains than lifting alone.

Untrained people showed no loss (0.03, -0.29 to 0.35). Moderately trained people showed a trend that didn't reach significance (-0.20, -0.42 to 0.02). Trained people gained less (-0.35, -0.59 to -0.11), and the effect showed up only in studies where both were done within 20 minutes of each other (-0.66). When sessions were more than two hours apart, trained people showed no significant difference (-0.10, -0.43 to 0.23).

Schumann's group is more cautious about it. They note that Petré's analysis counted some studies more than once when several training groups shared one control group, which can inflate power. They also note their own training-status subgroup leaned the same direction without reaching significance. So: plausible, worth respecting if you're trained and chasing a lower-body number, not proven.

Legs pay, arms mostly don't#

Look at where the costs land. Petré only measured lower-body 1RM, and that's where the trained-lifter effect showed up.² Sabag's meta-analysis of interval training added to lifting split it out directly: muscle size and upper-body strength changed about the same as with lifting alone, while lower-body strength gained a little less (ES -0.248, p = 0.049).⁶ Schumann's explosive-strength outcomes were jumps, rate of force development and leg-press power, mostly legs.¹

That fits common sense: the legs do the running, so the legs carry the fatigue into the next lifting session. If your heavy training days are upper body, interference is close to a non-issue on the current evidence.

Running or cycling?#

This is where the meta-analyses openly contradict each other.

• Wilson 2012: running cost strength and size, cycling didn't.⁵
• Schumann 2022: no difference between running and cycling on any outcome.¹
• Lundberg 2022: running hurt type I fibre growth, from three studies.⁴
• Sabag 2018, interval training only: lower-body strength was a little lower overall (-0.248, p = 0.049). The trend pointed at cycling intervals (-0.377, p = 0.074), not running intervals (-0.176, p = 0.261), though neither was significant on its own.⁶

The honest summary is that nobody knows yet. The explanation people usually give for blaming running, its extra pounding, sounds reasonable. These pooled numbers don't establish it.

Why it would happen at all#

The appealing story is molecular: endurance signalling switching off the growth signalling that lifting switches on. It shows up in cell and animal models. Human studies of single concurrent sessions haven't found that switch in a way that explains blunted growth.⁷ Coffey and Hawley's view is that interference is real in chronic training studies, but its mechanism is unknown and probably several things at once.⁸ Schumann's group speculates about something plainer for explosive strength: residual fatigue from the aerobic work lowering the quality of the lifting, not its quantity.¹

Practically, it doesn't matter much which explanation wins. The levers are the same: how close together the two sessions are, how much endurance work there is, and how trained you are.

What the evidence supports, and what it doesn't#

The other direction: does lifting hurt your running?#

Most of this research measures what aerobic work does to lifting. The reverse looks milder. In Hickson's original study the combined group's VO2max rose as much as the endurance group's, even while their strength stalled.³ In the sequencing meta-analysis by Murlasits, lifting first or running first made no difference to VO2max (0.39 ml/kg/min, 95% CI -1.03 to 1.81).⁹ And strength training done well can make endurance athletes faster, with a lower energy cost of moving as one likely reason.¹⁰

The real cost runs through fatigue. A heavy lifting session can lower the quality of endurance sessions for several hours to days afterwards.¹¹ That's not interference in the molecular sense. It's a scheduling problem, and it's fixed by keeping heavy legs away from the run that matters.

The limits of all of this#

Most studies are short. Petré lists the lack of studies longer than 12 weeks as a gap, along with the lack of research on women, trained women especially.² Schumann's included studies averaged 4.3 on the PEDro quality scale, which they class as medium, and seven were rated low quality.¹ Many participants were untrained or recreationally active, and almost nobody in these studies trained like a hybrid athlete doing both seriously for years. Effect sizes also hide individual spread. A pooled -0.06 doesn't mean nobody lost anything.

If you've done both for years#

There's almost no research on people like you. The "trained" groups in these studies were usually lifters or athletes in a single sport, adding the other quality for a few months. If you've run and lifted together for years, nobody has measured what an extra block of concurrent training costs you. The cautious read is Petré's: the more trained you are, the more a small interference matters, because there's less easy progress left to absorb it.²

What to do with it#

If you're new to one half, add it. The data say the cost to the other half is small or zero, and the untrained groups in Petré's analysis showed no loss (ES 0.03).

If you're a trained lifter with a lower-body strength goal, the one fairly consistent signal is distance between the sessions. Keep hard runs and heavy leg work in separate sessions, at least a few hours apart. In Petré's trained group the effect went from -0.66 within 20 minutes to -0.10 at more than two hours. Run before or after lifting covers order and gaps in detail.

If power matters to you, like jumps or sprint starts, protect it the same way. Explosive strength is the quality with the clearest cost, and the cost was tied to same-session training.

If you're a runner adding lifting, interference barely applies to you in this direction. Your risk is the heavy leg day sitting the day before your intervals. Move the leg day, not the order.

If you train upper body on the days you run, you're already doing the low-risk version.

If you're short on time, the bigger risk isn't interference. It's dropping a quality completely. The minimum effective dose covers how little keeps each half alive, and zone 2 when you also lift covers the easy-running side. The sources behind hybridAF's own scoring are on the science page.