The Association Between Running Injuries and Training Parameters: A Systematic Review

This systematic review was registered in PROSPERO (No. CRD42018112913) and is reported according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2009 guidelines.¹² 

We conducted bibliographical searches in 4 databases (MEDLINE/Ovid, CINAHL, Embase, SPORTDiscus) in collaboration with experienced research librarians. All databases were searched from their inception to July 7, 2020, for 3 concepts: (1) running; (2) musculoskeletal lower limb injury, pain, or pathology; and (3) training parameters. The terminology used for the search strategy is presented in the Supplemental Table (see Supplemental Table, available online at http://dx.doi.org/10.4085/1062-6050-0195.21.S1). The search strategy was tailored for each database, using the appropriate truncation and medical subject heading.

To be included in the review, articles had to (1) be published in English or French, (2) involve humans engaging in running (defined by the authors as novice, recreational, competitive, or mixed cohorts), (3) include any running program (structured or unstructured), (4) report on prospective cohort studies or randomized clinical trials (RCTs), and (5) provide data on RRIs and training parameters (distance, duration, frequency, intensity) or any measure of change in training parameters. Systematic reviews, conference abstracts, and articles from the gray literature were excluded. No restrictions were placed on age, sex, or gender.

After conducting the database search, we removed duplicate articles. All articles were then independently screened for eligibility by 2 blinded reviewers. The same reviewers independently screened the publications' titles and abstracts to identify research eligible for full-text review. Thereafter, eligible articles were retrieved in full text, which was independently scrutinized by 2 raters (A.F., J.F.E.) to confirm inclusion. In case of disagreement about final inclusion, a third reviewer (J.S.R.) was available to make the decision. The bibliographical references of the retrieved studies were also searched to identify additional relevant publications.

A first reader extracted the data from all included studies before a second reader corroborated and completed the extraction, if necessary. Data were extracted according to a predefined standardized form that consisted of authors, study design, population, methods, definition of RRI, running program, study variables (training parameters, changes in training parameters), outcomes (incidence of RRIs), results, and conclusions.

We assessed the methodological quality and risk of bias of included studies using a structured and validated quality appraisal tool, the Quality Assessment for Evaluating Primary Research Studies (QualSyst).¹³  The evaluation checklist contains 14 items pertaining to the study question and design, outcome measurements, interventions, analyses, results, and conclusions. The items were scored depending on the degree to which the specific criteria were met (yes = 2, partial = 1, no = 0). Items not applicable to a particular study design were marked N/A (not applicable) and excluded from the total score.

To ensure consistency in scoring, all team members involved in critical appraisal (A.F., F.D., J.S.R.) met after independently assessing one of the included studies. Next, 2 raters (A.F., F.D.) independently evaluated each study using the QualSyst criteria checklist. A preconsensus, interrater reliability score using an intraclass correlation coefficient (ICC) with 95% CI was calculated for the total quality score. The ratings were then compared to determine the scoring consensus between raters. When consensus was not reached, a third rater (J.S.R.) was available to contribute to the final decision. The following classifications were used to categorize methodological quality: high quality (HQ) > 80.0%; good quality (GQ) = 65.0% to 80.0%; moderate quality (MQ) = 50.0% to 64.9%; and low quality (LQ) < 50.0%.¹⁴ 

We conducted a qualitative review of the evidence. The level of evidence was adapted from the recommendations of the Cochrane Group Collaboration¹⁵  and classified as strong, moderate, limited, very limited, or conflicting (Table 1) after we considered the following domains: imprecision (number of studies and participants), risk of bias (methodological quality), indirectness (methodological and outcomes similarities), and inconsistency (direction of results). To obtain quantitative data about the incidence of RRIs, the total sample size and number of injured runners in each study were extracted. A weighted average (%) was calculated for each category of runners (novice, recreational, competitive, mixed) by dividing the number of injured runners by the total number of runners.

The search strategy identified a total of 9299 articles. After removing duplicates and screening the titles and abstracts, we identified 91 articles eligible for full-text review. Nine additional articles were included after the selected studies' reference lists were hand searched. After full-text screening, 36 studies were included in this review (Figure).

Of the 36 included studies, 33 were prospective cohort studies and 3 were RCTs. In all, they involved 23 047 runners (44% females) aged 17 years and older. Seven studies were conducted specifically with novice runners (n = 3315 participants, 63% females),^16–22  14 with recreational runners (n = 7905 participants, 44% females),^23–35,50  5 with competitive runners (n = 414 participants, 48% females),^36–40  and 10 with runners of mixed levels of ability and experience (n = 11 430 participants, 39% females; Table 2).^4,41–49 

Methodological quality scores ranged from 27.3% to 96.2% (Table 3). On average, the studies were of high methodological quality, with a median score of 86.4% (interquartile range [IQR] = 19.3%). Twenty-two studies were classified as high quality, 5 as good quality, 5 as moderate quality, and 4 as low quality. The preconsensus, interrater reliability score for quality assessment was excellent with an ICC of 0.97 (95% CI = 0.95, 0.99).

The most common reasons for deducting points on the methodological quality assessment were the outcomes being self-reported or not measured objectively (item 8, n = 28; see Table 3 for description of items), inappropriate or unjustified sample size (item 9, n = 20), not reporting the participants' selection methods or inclusion and exclusion criteria (item 3, n = 18), and inappropriate consideration of confounding variables (item 12, n = 14).

An overview of the incidence of injuries is presented in Table 2. Only 14 of 36 studies used definitions of RRI similar to the consensus definition developed by Yamato et al⁷  in terms of reducing or missing consecutive training sessions because of pain or the need to consult a health care professional. In total, 6043 (26.2%) runners sustained an injury among all 23 047 runners enrolled in the included studies. The incidence of RRI varied between 8.8% and 91.3% depending on the study population, length of follow-up, and definition of an RRI. More specifically, the incidence of RRI was 14.9% in novice runners (703 RRIs in 4720 runners; range = 9.4%–94.9%; follow-up = 6 weeks–18 months),^16–22,48  26.1% in recreational runners (2057 RRIs in 7888 runners; range = 17.9%–79.3%; follow-up = 1–24 months),^23–35  62.6% in competitive runners (259 RRIs in 414 runners; range = 52.6%–91.3%; follow-up = 3–24 months),^36,40  and 27.6% in runners of mixed levels (3158 RRIs in 11 430 runners; range = 8.8%–51.3%; follow-up = 3–60 months).^4,41–49 

Based on diagnosis details, the most frequently injured body parts were the knee, foot/ankle, and lower leg with 25.8%, 24.4%, and 20.9% of all injuries, respectively. Novice runners sustained RRIs mostly to their knee (30.8%), lower leg (29.7%), and foot/ankle (18.1%).^{16–18,20,22,48}  In recreational runners, the knee (26.5%), foot/ankle (20.5%), and lower leg (20.2%) were the most frequent sites injured,^{23,25–32,34,35,48}  whereas competitive runners most often injured their foot/ankle (39.4%), knee (25.4%), and lower leg (20.8%).^36,37,39  In cohorts of runners of mixed levels, the most commonly injured body parts were the foot/ankle (26.3%), knee (26.1%), and lower leg (23.1%).

An overview of the running programs and of the variables of interest for this review is presented in Table 2. Overall, we found conflicting evidence about the association between RRIs and running distance (20 studies), duration (12 studies), frequency (10 studies), and intensity (14 studies). High heterogeneity among studies (definition of RRI, categories of runners, length of follow-up) prevented us from performing meta-analyses on the association between RRIs and training parameters.

Four HQ,^32,35,37,42  2 MQ,^16,43  and 2 LQ^27,36  studies suggested that greater running distance could increase the injury risk. Three of the HQ studies reported that weekly distances of >30 km (hazard ratio [HR] = 3.28; 95% CI = 1.23, 8.75; P = .02),³⁵  >64 km (adjusted relative risk = 2.88),⁴²  or between 60 km and 70 km (99.7% and 94.7% increased risk versus 40 km to 50 km and 50 km to 60 km, respectively)³⁷  increased the risk of RRI. Another HQ study stated that running <40 km per week was a strong protective factor against calf injuries in recreational runners.³²  The incidence of RRI was linked with greater distance during the first (r² = 0.36, P = .001) and third (r² = 0.16, P = .015) phases of the running program for novice runners that was used in the first MQ study,¹⁶  whereas the second study reported a greater average weekly distance in injured compared with uninjured male recreational and competitive runners (effect size: r = 0.32, P = .046).⁴³  The LQ studies demonstrated more RRIs in a study group that ran a greater weekly distance²⁷  and a significant correlation between the injury rate during any 1 month and the distance covered by long-distance marathon runners during the preceding month (r = 0.59).³⁶ 

On the other hand, 2 HQ studies,^45,49  1 GQ study,³³  and 1 LQ²⁴  study described a tendency for a greater running distance to protect against RRIs. Specifically, the first HQ study showed a 10% reduced risk of knee injury (relative risk = 0.901, 95% CI = 0.820, 0.991) in novice and recreational athletes running greater distances,⁴⁹  whereas the other observed fewer injuries in those running >15 km per week compared with those who ran less (risk difference [RD] = −11.3%, 95% CI = −27.2%, 4.6%).⁴⁵  The authors of the GQ study concluded that overall, a greater weekly distance protected against RRIs (odds ratio [OR] = 0.99, β = 0.012).³³  The LQ study²⁴  indicated a trend for more injuries in those running <20 miles (32.2 km) per week (80% injury incidence) compared with >40 miles (64.4 km) per week (50% injury incidence).

Three HQ studies,^20,29,46  2 GQ studies,^25,40  1 MQ study,²⁸  and 1 LQ⁵⁰  study reported no association between weekly running distance and the risk of RRI in cohorts of novice,²⁰  recreational,^25,28,29,50  and competitive⁴⁰  runners as well as runners at mixed levels.⁴⁶  One HQ study⁴⁷  noted conflicting results, associating a greater distance with an increased risk of RRI at the 5-year time point but not after 1 year.

One HQ study²⁶  suggested that recreational runners were more at risk of RRI when running for sessions of longer duration (OR = 1.01, 95% CI = 1.00, 1.02), as did 1 MQ study,²²  which revealed RRI incidences in novice runners of 22.0%, 24.0%, and 54.0% in the 15-, 30-, and 45-minute duration groups, respectively.

In contrast, 3 HQ studies^19,34,49  demonstrated that longer running duration could lead to fewer injuries. Specifically, running >60 minutes in the previous 7 days was a protective factor against the occurrence of RRI (HR = 0.41, 95% CI = 0.20, 0.86).¹⁹  Greater mean session distance was also a significant protective factor among recreational runners (HR = 0.795, 95% CI = 0.725, 0.872),³⁴  and more weekly training hours were associated with fewer injuries, especially to the knee and foot, in a mixed cohort of novice and recreational runners (relative risk = 0.575, 95% CI = 0.451, 0.731).⁴⁹ 

Five HQ^{4,17,23,46,48}  and 2 GQ^25,38  studies found no association between running duration and the incidence of RRIs in cohorts of novice,¹⁷  recreational,^23,25  and competitive³⁸  runners as well as in runners at mixed levels.^4,46,48 

Two HQ studies^19,42  and 1 MQ²²  study showed that more frequent running tended to yield more injuries. Among recreational and competitive runners, greater weekly frequency was linked with more RRIs (7 days versus 0–2 days).⁴²  Running 3 times per week displayed a trend toward more injuries compared with 2 weekly sessions in novice runners (HR = 1.42, 95% CI = 0.97, 2.08), although it did not reach statistical significance.¹⁹  The investigators in the MQ research²²  observed a 39% prevalence of injuries in those running 5 days per week, 12% in runners running 3 days per week, and 0% in those running 1 day per week.

However, in 2 HQ studies,^31,34  running more often led to fewer injuries in recreational runners. In the first, compared with running 2 to 5 days per week, running only 1 day per week was a significant risk factor overall and in female runners (OR = 3.6, 95% CI = 1.1, 12.3), although the trend was not significant in males.³¹  In the second study, greater frequency was associated with a lower risk of RRI in the unadjusted statistical model (HR = 0.707, P = .002).³⁴ 

Two HQ studies,^46,47  2 GQ studies,^25,33  and 1 MQ²⁸  study described no association between running frequency and injury rates in cohorts of recreational runners^25,28,33  and runners at various levels.^46,47 

One HQ study¹⁹  indicated that greater running intensity in novice runners was more hazardous (HR = 1.28, 95% CI = 1.18, 1.40). One MQ study⁴¹  noticed more injuries in runners at mixed levels taking part in tempo runs during the first 6 weeks of a training program (OR = 3.96, 95% CI = 1.35, 11.61) and a trend toward more injuries in those practicing speed intervals (P = .06). The authors of the latter study reported that almost all runners who sustained RRIs (96.4%) performed tempo or interval runs as part of their marathon training.⁴¹ 

Investigators in 2 HQ^32,45  studies suggested that greater running intensity could yield fewer RRIs. Performing speed intervals regularly was a protective factor against the occurrence of knee injuries among recreational marathon runners (OR = 0.49, 95% CI = 0.26, 0.93).³²  In a cohort of runners of mixed abilities, there was a tendency toward fewer injuries among runners with a faster pace (>10 km/h, RD = –17.4%, 95% CI = −39.0%, 4.5%).⁴⁵ 

Four HQ studies,^29,34,42,46  3 GQ studies,^25,33,38  and 1 LQ⁵⁰  study found no association between running intensity and RRIs in cohorts involving mostly recreational runners^{25,29,33,34,50}  but also competitive runners³⁸  and mixed cohorts.^42,46  Two HQ studies identified conflicting results.^26,47  The first one found that speed training was associated with more injuries and interval training with fewer injuries.²⁶  The other reported that greater intensity was linked with more injuries at the 1-year time point (only in men) but not after 5 years.⁴⁷ 

We found conflicting evidence about the association between RRIs and specific changes in training parameters based on 11 studies. Five^{16,20,21,39,43}  stated that a recent increase in running distance was associated with an increased risk of RRIs. Specifically, 1 HQ study of 58 novice runners noted that over a 10-week follow-up, injured runners had a greater distance progression the week before the onset of injury compared with other weeks (86% difference, 95% CI = 12.9%, 159.9%; P = .026).²⁰  Injured runners also had an average increase in weekly running distance of 31.6% ± 3.1%, compared with 22.1% ± 2.1% in uninjured runners, although the difference did not reach statistical difference (P = .07).²⁰  According to another HQ study²¹  of 873 novice runners (12 months), those progressing weekly distance by >30% seemed more vulnerable to “distance-based injuries” than those who progressed <10% (HR = 1.59, 95% CI = 0.96, 2.66; P = .07). However, that same study found no association between overall incidence of injuries and distance progression. Another MQ study¹⁶  of 73 novice runners (over 18 months) associated a greater risk of RRI with a recent increase in distance between different training phases of a program leading to the completion of a marathon, especially when increasing from a mean weekly distance of 34.9 km to 43.6 km and duration from 3.2 hours to 4.0 hours. One GQ study³⁹  described a trend for university-level competitive runners (n = 97) to sustain more injuries when they had larger differences in running distance between regular and high-mileage weeks during a 3-month cross-country season (P = .06). One MQ study⁴³  with a mixed cohort of 76 recreational and competitive runners reported that, over the course of a 12-month follow-up, injured runners often increased their weekly distance by >30% and even >50% in the 4 weeks before sustaining an injury.

Five other studies (3 HQ,^4,18,30  1 GQ,³⁸  1 LQ⁵⁰) showed no association between recent changes in training parameters and the incidence of RRIs. Week-to-week changes in frequency or duration of running were not linked with RRIs in a mixed cohort of novice and recreational runners.⁴  An RCT¹⁸  from the same research group revealed that a 10% average increase in weekly running distance had no preventive effect against RRIs in novice runners when compared with a weekly distance progression >10% (OR = 0.8, 95% CI = 0.6, 1.3).

Another RCT³⁰  reported a similar incidence of injuries in 447 recreational runners whose schedules focused on increasing intensity or distance, although the average weekly increase in volume was only 3.25%. One more study in 17 recreational runners indicated no difference in RRIs after increasing running distance by 103% or intensity by 152% over a 4-week period.⁵⁰  As for competitive runners, the acute-to-chronic workload ratio was not associated with the onset of RRIs in a small sample of 23 runners.³⁸ 

One HQ study⁴⁴  reported conflicting results in 261 runners of mixed levels. Indeed, more RRIs were recorded during the first 3 weeks of a 14-week training program in those increasing weekly running distance between 20% and 60% compared with those increasing their weekly distance by <20%. However, this association became nonsignificant after 7 weeks and 14 weeks, and those increasing by >60% did not sustain more injuries.⁴⁴ 

Given the high volume of research on running injuries, this systematic review provided a much-needed update on the current state of the literature. Overall, the incidence of RRI across studies was 26.2%. Whereas previous data suggested greater injury rates in novice compared with recreational runners,⁵¹  our findings suggested the opposite, with a greater incidence in recreational (26.1%) than in novice runners (14.9%). In contrast with that previous review,⁵¹  we reported the number of events rather than the injuries per 1000 hours of running. Analyses based on exposure are interesting for putting numbers into perspective. However, we believe that considering injuries as events happening during a running program of any duration can better inform injury-prevention strategies, especially in novice and recreational runners. Almost half (48%) of novice runners who abandoned a running program did so because of an injury.^4,52  Thus, avoiding RRIs appeared essential to maintaining participation in running and its associated health benefits.⁵³  Yet it is possible that injuries per exposure time may better apply to the training reality of competitive runners, in whom the incidence was reported to be as high as 62.6% during follow-up periods of up to 24 months that likely included much more running than done by novice and recreational runners.

Overall, our findings showed conflicting evidence about the role of specific training parameters (distance, duration, frequency, intensity) as well as the influence of recent training changes in the onset of RRIs. In 2007, based on 17 studies (13 prospective, 4 retrospective), van Gent et al⁶  had already outlined the conflicting levels of evidence linking training parameters and RRIs. In 2012, based on 31 studies (13 prospective, 9 retrospective, 6 case-control series, and 3 RCTs), Nielsen et al⁵  also reported conflicting results between distance, duration, frequency of training, intensity, and RRIs. Finally, the authors of a more recent and smaller systematic review¹¹  based on only 4 studies concluded that very limited evidence associated sudden changes in training loads with the onset of RRIs. In comparison, we reviewed 33 prospective cohort studies and 3 RCTs. The sole intervention in all studies was a running program and combined training parameters at large and recent changes in training. In our opinion, the persistence of conflicting results over the years speaks to a lack of consistent definitions and reporting guidelines in the field of RRIs. In addition, assessing the relationship between training parameters and RRIs is certainly more complex than looking at training parameters in isolation.

Conflicting evidence for the association between running distance, duration, frequency, and intensity, as well as recent changes in training, and RRIs outlines the complexity of running injuries. Factors related to movement biomechanics, load capacity, and lifestyle factors⁸  are but a few elements that can contribute to RRIs. Importantly, these need to be measured and reported objectively to provide meaningful insights. Unfortunately, researchers in only 5 studies (14%) of the 36 included in this review collected global positioning system–based data to improve the accuracy of actual training loads.^{20,21,30,44,45}  Moving forward, investigators should consider using wearable devices to provide a better picture of actual training parameters, instead of prescribed parameters, and to minimize reporting errors from participating runners.⁵⁴  They should also consider describing multiple components of workload (eg, intensity and frequency)—not just distance⁵⁵—as well as patterns of variation experienced during the running program and around the timing of injuries to provide a better picture of training loads. Recent examples of studies^56,57  involving different running populations emphasized the substantial underestimation of changes in week-to-week training loads when only distance or duration was considered.

Attributing the cause of RRIs solely to training loads—or external loads—fails to address the plethora of factors related to the individual—or internal loads—which vary among individuals and even in the same individual over the course of a running program.⁵⁸  The common advice to increase distance by 10% per week for all runners,⁵⁹  in every situation, is too simplistic and should not be recommended based on our results. A 10% increase for a novice runner who runs 10 km per week is likely much safer than for a competitive runner who runs 150 km per week, including high-intensity workouts. In addition to experience and previous adaptations to exercise that involve impact, subjective measures of well-being, such as psychological stress, fatigue, and recovery, can all affect an athlete's response and adaptation to training.⁶⁰  Only 10 studies in this review reported outcomes related to personality and mental health,^{17,24,26,29,41}  diet,^22,39,50  smoking habits,^22,33,42  or alcohol intake.^33,39  Only 2 studies^22,39  asked questions about sleeping habits, and only one³⁹  documented hormonal variations and the use of contraceptives in female runners. Monitoring and reporting these variables and measures of internal loads is crucial to gain a better understanding of the causes of RRIs in both females and males. These factors should be considered especially in competitive runners, among whom the interaction between training parameters and recovery could explain the higher rate of RRIs.

The clinical applicability of this review was limited because of conflicting findings. However, the conflicting findings stemmed from the many limitations of the included studies. Results regarding the incidence of RRIs in different running populations were limited due to the heterogeneity of definitions used in the different articles. Findings about RRIs and training parameters or changes in these parameters were limited because of the lack of objective data about training loads and the lack of data on factors related to the individual. We understand that collecting all of these burdens research teams and might not always be feasible, depending on the budget. Still, unless these factors are addressed in future research, we predict that future reviews will still demonstrate conflicting findings about the causes of RRIs. It is also possible that the wide variability in factors affecting the capacity for adaptation in different runners precludes us from making specific recommendations for novice, recreational, or competitive runners about training parameters to reduce the risk of RRIs. Running biomechanics, footwear, and surface or changes in these variables should also be considered when possible.

Meaningful conclusions and recommendations are made possible with consensus definitions and reporting guidelines. Unfortunately, fewer than 40% (14/36) of included studies used RRI definitions that were in accordance with, or similar to, the consensus definition published by Yamato et al⁷  regarding reduced or missed consecutive training sessions because of pain or consultation of a health care professional. Our results also call for a better standardization of follow-up periods when monitoring injury incidence. Included studies varied between 1 month and 60 months, making the results too heterogeneous to generalize for the running population. A set of reporting guidelines could include, for example, details on injuries sustained during each 2-week period of a running program (eg, during weeks 1–2, 3–4, 5–6, and so on until the end of the study period). This could help to identify trends in injury incidence using a meta-analysis process, which was not possible in this review, and compare data in different levels of runners.

Novice, recreational, and competitive running populations should be more specifically defined. For the purpose of this review, we relied on information provided in the articles, although the details may not have been comparable. Despite a previous classification suggested based on the existing literature,⁶¹  a scientific process to determine consensus definitions could provide clear guidelines that experts in the field agree upon. This would make it easier to improve study designs, tailor research questions to the different populations,⁶²  and unite multiple research teams in their efforts toward a common goal. Standardized guidelines could also translate to better research uptake by the running community, coaches, and health care professionals.

Despite high rates of RRIs, conflicting evidence described the associations between weekly running distance, duration, frequency, and intensity, or recent changes in these parameters, and the incidence of RRIs. Thus, at this time, no specific recommendations related to optimal training parameters or progressions can be issued to guide clinical decision making and program planning. This result was largely due to a lack of consistent definitions and reporting guidelines. Finally, RRIs are multifactorial and likely not explained solely by training parameters. Authors of future prospective studies on the incidence of RRIs should consider how variations in objectively measured training parameters interact with factors related to the individual using psychosocial, hormonal, lifestyle, and recovery outcomes.

Supplemental Table. Search strategy.

Found at DOI: http://dx.doi.org/10.4085/1062-6050-0195.21.S1