2021 — Cricket Fast Bowling Technique and Lumbar Bone Stress Injury

Research theme

This is the single most important paper in this cluster for a coach. It is the first study ever to prospectively link full-body fast bowling technique to MRI-confirmed lumbar bone stress injury (LBSI). Every previous attempt was retrospective, lumped all lower-back injuries together, or relied on 2D “action classification” (side-on / front-on / mixed).

Method: 50 elite male fast bowlers on an international performance pathway, each bowling a minimum of six maximum-velocity deliveries on a good length, captured by an 18-camera Vicon system at 300 Hz with a synchronised Kistler force plate at 1500 Hz under the front foot. 47 reflective markers; 95 anthropometric measurements for subject-specific inertia. All bowlers underwent MRI lumbar spine screening (with CT/CT-SPECT follow-up where uncertain) as part of the ongoing ECB screening programme.

The group definitions are what make this study strong. Injured group = sustained an LBSI within 2 years after their biomechanical assessment. Uninjured group = never had an LBSI, and were assessed before age 22, and were at least 23 years old with ≥150 match days of professional cricket by the end of 2019 — i.e. they had survived the high-risk window with real workload, not merely avoided injury by not playing.

Independent samples t-tests to find group differences, then significant variables with medium-or-greater effect sizes fed into a binary logistic regression to build a predictive model.

What they measured

• Whether the bowler subsequently got a lumbar bone stress injury (MRI/CT-confirmed, stress reaction or stress fracture).
• How bent the back leg is at back foot contact (rear knee angle, rear hip angle at BFC).
• How bent the front leg is when it lands and at release (front knee angle, front hip angle at FFC and BR; front leg plant angle; front foot plant angle).
• How the pelvis is tipped, dropped and turned (pelvis orientation — tilt, drop, twist — at BFC, FFC, BR, and min/max between).
• The hinge between the pelvis and the low back (lumbopelvic angle: flexion-extension, side flexion, rotation) — the mid-point of the PSIS markers defines this junction.
• The hinge between the ribcage and the low back (thoracolumbar angle: flexion-extension, side flexion, rotation) — mid-point of xiphoid process and L1 spinous process.
• The lower thorax relative to the pelvis (lower thoraco-pelvic angle), replicating Ranson et al.’s 2008 definition.
• The classic mixed-action measures — shoulder counter-rotation and pelvis–shoulder separation at BFC (Portus et al., 2004 definitions).
• Ground reaction forces at the front foot — peak vertical and horizontal force, loading rates, impulses, in kN and in bodyweights.
• Run-up speed and ball release speed.
• Convention: anatomical position = 180°; below 180° = flexion, contralateral side flexion, contralateral rotation.

Findings

1. 39 of 50 bowlers (78%) sustained a prospective LBSI. Injured mean age 19.3 ± 2.0 at assessment, injured at 19.9 ± 2.1. Uninjured 19.8 ± 1.5. Comparable height and mass.

2. Injury location confirms the unilateral pattern. Of 39 LBSI (26 stress fractures, 13 stress reactions): 82% contralateral to the bowling arm, 3% ipsilateral, 10% bilateral. 72% at the pars interarticularis, 13% pedicle. Level: L4 28%, L5 28%, multilevel 23%, L3 13%.

3. Injured bowlers had played FEWER days of professional cricket, not more. Median ± IQR: 104 ± 236 days (injured) vs 268 ± 264 days (uninjured), ES > 0.5. This is a survivorship effect built into the uninjured group definition, but it is worth stating plainly: high career workload was not what marked out the injured group here.

4. Significant technique differences (bold = the paper’s own significant findings). Group means ± SD, LBSI vs non-LBSI:

At back foot contact:

Injured bowlers arrive at BFC with a 10° more flexed rear hip and a 10° more flexed rear knee. In the frontal/transverse plane they are side-flexed towards the bowling arm (ipsilateral, 182°) and rotated away from it (contralateral, 177°), whereas uninjured bowlers are the mirror image: side-flexed away (contralateral, 179°) and rotated towards (ipsilateral, 182°).

At front foot contact:

At ball release:

Between BFC and BR: injured bowlers had less extension of the front hip (max front hip angle 131 ± 9 vs 137 ± 6, medium effect) and more ipsilateral pelvic drop (max pelvis drop orientation 195 ± 8 vs 190 ± 6, medium effect).

5. Ground reaction forces did NOT differ between groups — at all. Peak vertical force 5.6 ± 1.3 vs 5.5 ± 0.8 kN (6.9 ± 1.6 vs 6.8 ± 1.0 BW). Peak horizontal, loading rates, impulses — none significant, in absolute or bodyweight-normalised terms.

6. The predictive model. Two parameters — rear hip angle at BFC and lumbopelvic angle at FFC — correctly classified 88% of bowlers by injury history (model 2; Cox & Snell r² = 0.27, Nagelkerke r² = 0.41, model χ²(2) = 15.5, p < 0.01). Rear hip angle alone got 76% (Nagelkerke r² = 0.24).

• Each 1° increase (straighter) in rear hip angle at BFC → odds of LBSI × 0.88 (95% CI 0.80–0.97).
• Each 1° increase (more extended) in lumbopelvic angle at FFC → odds of LBSI × 1.25 (95% CI 1.04–1.52).

7. Worked odds ratios — these are the numbers to quote to a coach. All relative to the uninjured group mean (rear hip 156°, lumbopelvic 172° = odds ratio 1.0):

Note the confidence intervals on the extremes are enormous. The 4.8 and 11.5 figures are the defensible ones.

8. Status of the evidence. This is a prospective correlational study with a predictive model. It is much stronger than the retrospective work before it — technique was measured before injury for the injured group, on a cohort where the uninjured comparators had genuinely survived heavy workload. But it is not causal. No intervention was performed. The authors are careful: poor lumbo-pelvi-femoral control is “a potential cause”; the cause of the injurious rear-hip position “was not identified within this study.”

What a coach should look for on video

This paper supports four genuine cues, and it kills one that coaches have been taught for thirty years.

Cue 1 — The back leg at back foot contact (the single strongest predictor)

• The cue: How tall and straight the bowler is over the back leg the moment the back foot lands. Look at the angle at the hip — the line from the trunk down through the thigh.
• Camera view + frame: Side-on, scrub to the frame where the back foot first contacts the ground. This is the frame before the bound into the delivery stride finishes — the back foot landing, not the front.
• What “good” looks like: A relatively straight, extended rear hip — around 156° in the uninjured group, and the lowest-risk bowler in the study was at 170°. The bowler stands tall over the back foot and passes over it.
• What the fault looks like: A collapsed, sat-down back leg — hip and knee both folded, the bowler dropping into the back foot rather than riding over it. The injured group averaged 146° at the hip and 146° at the knee — a full 10° more bent at each than the uninjured group. The worst bowler in the study was at 123°.
• Why it matters: Injury risk, and it is quantified. Every 1° of extra rear-hip flexion multiplies the odds of LBSI by 1/0.88. Going from the uninjured mean (156°) to the injured mean (146°) — ten degrees, which is visible on video — raises the odds ratio from 1.0 to 4.8. This one parameter alone correctly classified 76% of bowlers.
• Coaching caution: The paper explicitly does not say “tell them to straighten the back leg.” It argues the flexed position is likely a consequence — a bowler flexes to get into the mid-range of the joint where more torque is available, because they need that torque to stabilise the pelvis and redirect momentum from the run-up. The fix may be strength, or run-up speed/alignment, not a cue about the leg. See Cue 4.

Cue 2 — The pelvis and low back at front foot contact

• The cue: Whether the pelvis tips forward and the low back hollows out as the front foot lands.
• Camera view + frame: Side-on, scrub to front foot contact (the frame the front foot first loads).
• What “good” looks like: Pelvis relatively upright — uninjured group 175° pelvic tilt — and the lumbopelvic junction not driven into extension: uninjured 172°.
• What the fault looks like: Pelvis tipped anteriorly (injured mean 170°, i.e. 5° more anterior tilt), front hip more closed/flexed (injured 130° vs uninjured 137°), and the low back arching to compensate at the lumbopelvic junction (injured 176° vs 172°). Visually: the belt line tips forward, the bum sticks out, and the small of the back hollows sharply just above the waistband while the front thigh comes up towards the chest.
• Why it matters: Injury risk. Each 1° of extra lumbopelvic extension at FFC multiplies LBSI odds by 1.25. Combined with a flexed rear hip at BFC, the odds ratio goes from 4.8 to 11.5. This is the second half of the 88%-accurate model.
• Mechanism the authors propose: The bowler with an anteriorly tilted pelvis at FFC compensates by extending at the lumbopelvic junction, which positions the upper spine to maximise trunk flexion to release — a known ball-speed enhancer. So this may be a bowler buying speed with lumbar extension. Alternatively it is simply the pelvis losing control.

Cue 3 — Ipsilateral pelvic drop through the delivery stride

• The cue: Whether the hip on the bowling-arm side drops as the bowler goes through front foot contact.
• Camera view + frame: Front-on or behind-the-arm, scrubbed through BFC → FFC → BR. Watch the line between the two hip bones.
• What “good” looks like: The pelvis stays relatively level in the frontal plane. Uninjured max pelvis drop orientation 190 ± 6°.
• What the fault looks like: The bowling-arm-side hip visibly sagging/hitching down — injured group max 195 ± 8°, a medium effect size difference.
• Why it matters: Injury risk. The authors read this, together with the anterior pelvic tilt and reduced front hip extension, as the signature of failure to stabilise the pelvis during BFC→FFC — i.e. poor pelvi-femoral control. It is one of the ten significant parameters, though it did not survive into the final model.

Cue 4 — Where the sideways lean comes from (the subtle one)

• The cue: When the bowler leans away from the bowling arm to get the hand high at release, which part of the trunk bends.
• Camera view + frame: Front-on, scrub to ball release.
• What “good” looks like: The lean is achieved high — bending between the ribcage and the low back (thoracolumbar). The uninjured group had more contralateral thoracolumbar side flexion at BR (160 ± 3°, further below 180° = more contralateral lean at that joint).
• What the fault looks like: A relatively stiff ribcage-on-lumbar segment (injured 163 ± 4° — less lean at that joint) with the bend instead happening low, at the waistband (lumbopelvic side flexion at BR: injured 174 ± 5° vs uninjured 176 ± 4° — more contralateral lean, though p = 0.09, d = 0.57). Visually: the bowler gets to the same total lean, but it looks like a sharp kink just above the belt rather than a smooth curve through the mid-back.
• Why it matters: Injury risk. Both groups end up with roughly the same total trunk lean — they achieve it differently. The authors argue the injured bowlers must therefore be doing more of it at the lumbopelvic junction, which sits immediately adjacent to L4/L5 where the fractures occur. Flag this cue as the paper’s weakest number: the lumbopelvic side flexion difference did not reach p < 0.05. It has a medium effect size and it fits the injury location, but it is inferred.

What NOT to look for: shoulder counter-rotation / the mixed action

Do not screen for the mixed action. See the contradiction section below. In this study shoulder counter-rotation (43 ± 14° injured vs 40 ± 20° uninjured) and pelvis–shoulder separation at BFC (21 ± 15° vs 13 ± 24°) were both non-significant. The 30° counter-rotation threshold that has driven UK and Australian coaching for decades did not predict who got injured.

What NOT to look for: front foot impact / “pounding the front foot”

Ground reaction force did not differ between groups on any measure. Peak vertical force was ~6.9 vs 6.8 bodyweights. Coaching a “softer” front foot landing is not supported by this data as an LBSI intervention.

What NOT to look for: pace

Ball release speed 35.1 vs 35.8 m/s, not significant. A fast bowler is not at risk because they are fast.

Caveats and limits

• Badly unbalanced groups: 39 injured vs 11 uninjured. The authors acknowledge this could “skew the technique associated with LBSI and result in a sample size bias.” Eleven uninjured bowlers is a thin comparator for a two-parameter logistic model, and the extreme odds ratios (88.9; 484.0) have confidence intervals spanning three orders of magnitude — they are illustrative arithmetic, not findings.
• Sample: elite male bowlers, mean age 18.9. Not women, not juniors (see the Keylock adolescent papers), not club cricketers.
• The uninjured group is defined by survivorship — ≥23 years old with ≥150 professional match days and never injured. That is a deliberately hard filter, which strengthens the comparison, but it means the uninjured group is by construction a group of durable bowlers who may differ in ways beyond technique.
• Correlational, not causal. No intervention. The paper never claims that changing rear hip angle prevents injury. It repeatedly says the cause of the injurious position was not identified and proposes strength/control as the underlying driver.
• Discrete instants, not continuous curves. Only BFC, FFC, BR and min/max were analysed. Timing and rates of change were not.
• Absolute angles, not normalised to each bowler’s own ROM. The authors list this as a limitation; the 2023 Felton ROM paper starts to address it.
• No multiple-comparison correction. Dozens of t-tests at α = 0.05 with no Bonferroni adjustment (deliberately, to avoid Type 2 errors). The authors state that individual comparisons “should be considered cautiously” but that the logistic regression is not compromised by multiple testing.
• Stepwise logistic regression has known problems (bias in parameter estimation, reliance on a single best model). The authors mitigated this by computing all possible models for comparison.
• One trial per bowler.

Relationship to other Felton work

• Builds on and supersedes the 2020 ISBS paper (2020 Alway — Does Lumbar Spine Kinematics Contribute to Lumbar Bone Stress Injury?), which found no differences in intra-lumbar (L1–L3 vs L3–L5) joint angles between injured and uninjured bowlers and concluded “other joint rotations or other risk factors may have a greater contribution.” This paper found them — at the hip and the lumbopelvic junction.
• Is the reference point for everything after it in this cluster. The 2022 Keylock BMD paper, the 2022 Keylock adolescent risk-factor paper, and the 2023 Felton ROM paper all cite it as the established technique-injury model.
• The 2023 Felton ROM paper (2023 Felton — Range of Motion and Key Performance and Injury Technique Characteristics) takes this paper’s seven injury-related technique parameters and asks which off-field flexibility restrictions predict them — turning this paper’s findings into a non-video screening tool.
• Uses the Worthington et al. (2013) marker set and analysis pipeline that runs through the whole Loughborough/ECB body of work.

CONTRADICTION: The mixed action does not predict lumbar bone stress injury. This study measured shoulder counter-rotation (43 ± 14° injured vs 40 ± 20° uninjured) and pelvis–shoulder separation at BFC (21 ± 15° vs 13 ± 24°) and found no link to LBSI — despite the injured group averaging 43° of counter-rotation, well above the 30° “mixed action” threshold from Portus et al. (2004) that has driven coaching practice since the 1990s. The paper states directly: “Mixed bowling actions have previously been widely considered as the cause of LBSI… This study found no link between shoulder-counter rotation, pelvis-shoulder separation at BFC or the shoulder twist orientation at BFC, and LBSI.” Its conclusion is blunt: “Coach education should incorporate these findings and move away from using far removed derivatives to inform practice which are not consistent with predicting LBSI injury.” For a cricket coach this is the headline finding of the entire cluster.

CONTRADICTION: Excessive contralateral trunk side flexion at ball release is NOT the fault it was thought to be — the injured bowlers had LESS of it at the thoracolumbar joint. Bayne et al. (2016) and Ranson et al. (2008) linked excessive contralateral thoracolumbar side flexion to LBSI. This study found the opposite direction: injured bowlers were significantly less contralaterally side flexed at the thoracolumbar joint at BR (163 ± 4° vs 160 ± 3°, large effect size). The paper says so explicitly: “While this contradicts previous findings…”. The reinterpretation is that both groups achieve similar total trunk lean, but the injured bowlers source more of it from the lumbopelvic junction (p = 0.09, d = 0.57) — one joint lower, adjacent to where the fractures occur.

CONTRADICTION: High ground reaction forces do not independently cause LBSI. Ranson et al. (2008) proposed that extreme lower-thoracic side flexion combined with large ground reaction forces was “the most significant stressor of the contralateral side lumbar neural arch.” This study found no difference in any GRF parameter between injured and uninjured bowlers, in absolute or bodyweight-normalised terms. The authors conclude GRF “may not independently contribute to LBSI but may contribute in combination with lumbar kinematics.”

CONTRADICTION (partial, with Ranson et al. 2008): Ranson concluded the mechanical aetiology was motion of the lower thorax relative to the pelvis. This paper agrees with Ranson that action classification is not the answer, but relocates the mechanism: “The findings of this study highlight however that it is the motion at the lumbopelvic junction, which is adjacent to the site of typical LBSI, that is the likely mechanical aetiology. Future biomechanical analysis on fast bowlers should focus on the lumbopelvic junction.”

TENSION (with the 2020 paper in this cluster): The 2020 study, on 45 bowlers from the same programme, found no significant lumbar angle differences of any kind. This study, on 50 bowlers with cleaner prospective grouping, found ten significant differences. The difference is partly the group definitions (the 2020 uninjured group had not survived the risk window) and partly which joints were segmented — the 2020 paper split within the lumbar spine, this one looked at the joints above and below it.