Search strategy and inclusion criteria
The Cochrane database was searched, revealing no systematic reviews about training characteristics and RRI. A search of the Pubmed, Web of Science, Embase, and Sportdiscus databases was conducted October 11th 2011 to identify studies that met the inclusion criteria using the search strategy presented in Appendix 1. The search was limited to studies of humans, published in English, and included only original articles.
Prospective cohort studies, cross-sectional studies, case-control studies, and randomized controlled trials were included in the current systematic review if a relationship between training characteristics and RRI was investigated. Studies with novice, recreational, and elite runners between the ages of 18 to 65 were included. Articles were excluded if participants were sprinters or middle distance runners, or were predominantly exposed to types of sporting activity other than running such as triathlons, and military training programs. Articles on cadavers, computer modeling/simulation studies were excluded.
The exposure variables of interest were training characteristics including volume, distance, mileage, time, duration, frequency, intensity, speed, and pace. Different methods for analyzing or reporting these characteristics were accepted. For instance, volume could be measured as kilometers or miles per day, per week, per month or as the gradual increase in mileage per week over a given period of time. The outcome of interest was RRI in general or specific RRI of the lower extremity or spine. Muscle cramps, corns, blisters, and calluses were not included as RRI.
Data collection and analysis
Each study identified as a result of the electronic search was initially evaluated independently by two authors (RON and IB) by screening the title and abstract. Articles without an abstract were excluded. All articles of interest were retrieved and evaluated for eligibility. Articles were excluded if no information was provided on injuries during follow up, in case of overview articles, or articles about degenerative diseases only.
Methodological quality assessment
The methodological quality of the cross sectional studies, case-control studies, and prospective cohort studies was assessed by means of a methodological quality assessment list developed and used by van der Worp et al,26
which was based on a list developed by van der Windt et al.27
The list was adapted slightly to make it specific for training and RRI. The assessment contains items on information and validity and/or precision in five categories: study objective, study population, outcome measurements, assessment of the outcome, and analysis and data presentation. Separate quality assessment lists were constructed for cross-sectional studies, case-control studies, and prospective cohort studies. The items of the quality assessment list are presented in . Each item was evaluated as either positive (+) or negative (−) by two reviewers independently. In cases where it was unclear whether a study did or did not meet an item, or if no clear information regarding the item was stated, the item was scored as negative (−). Results of the quality assessment made by the two reviewers were compared, and any disagreement concerning an item was resolved in a consensus meeting. The total quality of each study was calculated by counting the number of items being positive (+) from item 3 to 16 divided by the total number of items for the study type (11 for case-control studies, 9 for prospective cohort studies, and 8 for cross sectional studies).
Summary of quality scoring criteria for cross-sectional studies, case control studies, and prospective cohort studies.
The methodological quality of the randomized controlled trials included was rated using the PEDro rating scale which is based on the Delphi list developed by Verhagen and colleagues.28
The total methodological quality score was found by evaluating the internal validity and statistical reporting using an 11 criteria list. The total quality of each randomized controlled trial was calculated by counting the number of items being positive (+) from item 2 to 11 divided by 10. Previously, the PEDro scale has demonstrated an inter-rater agreement of [k
] = 0.73–0.82.29
After examining 4561 titles and abstracts, 62 articles were identified as potentially relevant. After reference checking, one additional study was identified.30
The full texts of all 63 articles were retrieved and were subsequently evaluated by both RON and IB. Review of the complete texts excluded 32 articles. Of the excluded articles, four were overview articles,31–34
four included persons less than 18 years of age,35–38
three included persons with degenerative diseases only,39–41
eight articles did not describe the relationship between training characteristics and RRI,42–47
three had no control group,48–50
two were modeling articles,51,52
seven had a faulty injury definition or none at all,53–59
and one was a design article.60
Finally, 30 articles were included in the review.
Risk of bias in included studies
The quality of included studies is presented in . The overall methodological quality of the included prospective studies, case-control studies, and cross sectional studies was 44.1% ranging from 9 to 89%. The most problematic areas were 1) the main purpose of many of the studies was different than the relation between training and RRI, 2) description of the demographic characteristics (gender, age, body mass index) of the participants was lacking, and 3) lack of adjustment for the effect of multiple training variables. The overall quality of the three randomized controlled trials was 43%.
Summary of quality scoring for all included studies. Scores given for the items of the quality assessment list for prospective cohort studies, cross sectional studies, and case-control studies and the PEDro scale for randomized controlled trials.
Description of studies and injury definition
The year of publication for the included studies ranged from 1977 to 2008. The studies represented populations in USA, Canada, New Zealand, The Netherlands, Denmark, Switzerland, Germany, and Sweden. The total sample size of included participants was 24,066, ranging from 28 to 4,335 subjects in each study. Of the 30 included studies, nine were retrospective cohort studies, 12 were prospective cohort studies, six were case-control studies, and three randomized controlled trials. The study characteristics of the selected studies were described to obtain insight into the homogeneity of the study populations (). The types of participants (novice, recreational, and elite), and the injury definition used varied considerably between the studies. For instance, Lysholm et al16
used “all injuries that markedly hampered training or competition for at least 1 week were noted” while Valliant61
used “injury was defined as physiological damage or bodily pain which interfered with one×s ability to run”. The mean age of all participants in the 30 studies varied from 19.5 years to 44 years with an average of 35.4 years. Mean body mass index was 22, ranging from 20.97 to 25.86. Four studies included only males while two included only females. For the remaining studies, an average of 67.6% of the participants included were males. presents summary data from each study regarding the type of runner, demographic characteristics, and injury definition as quoted verbatim from the article.
Descriptions of included studies characteristics. Injury definitions are quoted verbatim unless stated otherwise.
Description of training characteristics
In 22 studies, the training characteristics were assessed retrospectively by a questionnaire. The recall period varied from two weeks to 10 years. In eight studies, daily running diaries9,16,17,30,62–64
or an internet based log15
were used. In five studies, training interventions were used.5,9,15,17,30
Odds Ratio (OR), Hazard Ratio (HR), and Relative Risk (RR) were the most common measures of association. The unit of measurement in this review is miles. However, some articles used kilometers. In these cases, kilometers were converted into miles using 0.62137 as conversion factor. Different definitions were used in the reviewed studies for training volume, duration, intensity, and frequency.
In 28 articles out of 30 articles, the link between training volume and RRI was investigated. The most commonly used approach to define exposure was to measure the average weekly miles4,13,14,16,22,61–63,65–69
of running over a period of time. In other studies, weekly distance per weekly frequency7
or total running distance64,71
were used as the measure of exposure.
In three articles, average hours9,17,69
spent running per week were used as the exposure variable. In another study, the weekly progressive increase in duration during a graded training program was used,15
while two other studies used minutes per day3,30
as their measure of exposure.
In 16 articles, training intensity was described.1,4,9,11,13,14,17,22,63–70
In a majority of these, average pace of workout was used to express intensity during training, measured as minutes per mile (min/mile) or minutes per kilometer (min/km).13,22,63,65–70,72
Other studies used kilometers per hour,17,64
16 km running time,14
or percentage of maximal attainable heart rate.9
The number of weekly training sessions was reported in a variety of ways as number of training sessions,71
per week. Most often the data were analyzed by dividing the weekly amount of days running into different categories. The comparisons vary widely across studies, however. The reference groups were defined as either 1, 1-2, 1-3, 1-4, or 1-5 days per week, and were compared to either one or several exposure groups varying between 3, 4, 5, 6, 7, 4-5, 5-7, 6-7 days per week. In one article, a regression model was used to investigate the risk of RRI as the weekly frequency increased during training prior to a marathon.6
Relationship between training characteristics and RRI
Hootman et al4
found an increased risk of injury among males (HR = 1.66 [1.43, 1.94]) and females (HR = 2.08 [1.45, 2.98]) running more than 20 miles per week. Lysholm et al16
found a significant correlation (r = 0.59) in long-distance/marathon runners between the distance covered in a given month and the number of injury days during the following month. Walter et al11
found no significant difference in relative risk between the reference group who ran less than 10 miles per week and the groups who ran distances between 10 and 39 miles per week. However, the relative risk of injury was significantly higher among males (2.22 [1.30-3.68]) and females (3.42 [1.42-7.85]) running ≥40 miles per week. This was supported by Macera et al22
who found a significantly increased odds ratio for sustaining injury among males running ≥40 miles per week over a period of 3 months (2.9 [1.1-7.5]). In the same study, no association was found between weekly mileages and risk of injury among women.22
Although a majority of studies reported a relationship between weekly mileage and RRI, no significant association between miles per week and likelihood of injury was found in two prospective studies and one retrospective study.7,62,74
In retrospective studies, several authors compared total volume per week between injured and non-injured subjects. Koplan et al12
investigated the proportion of injuries over a 10 year period in different mileage strata. The proportion of women reporting injury was highest in those who ran 40-49 miles per week. For men, the proportion was highest among those who ran 30-39 miles per week. Those running more or less miles per week had a smaller proportion of injuries. Marti et al14
found that runners who sustained injuries during the study period ran greater weekly mileage when compared to non-injured runners (26.3 km [3.2-83.8] versus 22.0 km [2.1-78.6], p 0.01). In a one-way analysis, Valliant61
also indicated that injured runners ran significantly more miles per week than non-injured runners (47.5 ± 20.5 miles versus 29.6 ± 16.7 miles, p < 0.01). This is supported by Jacobs et al13
and Koplan et al1
who found mileage run per week to be highly associated with injury.
In two studies, the RRI per 1000 hours of running in groups running different mileages per week were investigated.9,17
The number of injuries per 1000 hours of running appeared to decrease with increasing weekly mileage ().
Figure 1. Relationship between miles per week and Running Related Injury (RRI) reported as mean [95% confidence interval] for different comparisons. Results from the articles by Bovens and Jakobsen are calculated based on figures in the articles. RRI = Running (more ...)
Walter et al11
investigated the relationship between longest run per week and risk of injury. The relative risk of sustaining an injury when the longest run each week is >5 miles, is 2.49 [1.64-3.71] among males and 1.78 [0.99-3.13] among females compared with a reference group having their longest run below 5 miles. Van Middelkoop et al7
measured weekly distance per weekly frequency. Running an average of 6.8–9.3 miles per training session was not associated with increased or decreased risk of sustaining an injury compared to average runs above or below 6.8–9.3 miles.
Several authors have investigated the relationship between training volume and specific running injuries. Reinking et al10
investigated subjects sustaining exercise related lower leg pain and found no significant difference in injuries between individuals training more or less than 40 miles per week. Satterthwaite et al6
found an increased odds ratio for hamstring (1.07 [1.02, 1.13]) and knee (1.13 [1.04, 1.23]) injuries by a weekly increase in mileage of 6 miles. Wen et al69
found a significant difference in weekly mileage between subjects sustaining hip (18.7 miles per week) or hamstring injuries (22.4 miles per week) compared to controls (13.3 and 13.4 miles per week). Kelsey et al8
found miles run per week in the past year to be non-predictive of stress fractures. Wen et al69
found weekly mileage and hours per week protective against overall injuries, knee injuries, and foot injuries. In case-control studies, no difference in weekly mileage was found between controls and persons with plantar fasciitis,65
or anterior knee pain,68
while patients with patellofemoral pain ran significantly less than healthy controls.66
For iliotibial band friction syndrome, Messier and colleagues found conflicting results in two different studies. In one study, injured participants ran significantly less than healthy controls, and in the other study no significant difference in weekly mileage between injured and healthy participants was reported.65,67
Pollock et al30
found an increasing injury incidence among novice runners who ran in 15, 30, and 45 minute duration groups of 22%, 24%, and 54%, respectively. Jakobsen et al17
reported 7.4 and 6.9 RRI per 1000 hours of running among marathon runners who ran 204 [95% CI: 198-210] and 162 [95% CI: 156-168] minutes per week on average over a one year period. Over a time period of 18 months, Bovens et al9
reported 12.1, 10.0, and 7.0 injuries per 1000 hours of running among marathon runners who ran 162, 192, and 240 minutes per week. Buist et al15
found an average of 33 [95% CI: 27-40] RRI per 1000 hours of running in two groups of novice runners. One group was instructed to run an average of 52 minutes per week over a 13 week period (30 RRI/1000 hours), while the other group were instructed to run an average of 59 minutes per week over a 8 week period (38 RRI/1000 hours). shows the RRI/1000 hours of running in groups running different minutes per week.
Figure 2. Relation between minutes per week and number of Running Related Injury (RRI) per 1000 hours of running. Results from the articles by Bovens and Jakobsen are calculated based on figures in the articles. Int = intervention. Con = controls. RRI = Running (more ...)
Buist et al15
investigated the relationship between weekly progression in running duration and likelihood of injury. There was no significant difference in the incidence of RRI in a group of runners with a 13 week training program with a mean duration increase of 10% per week compared to the incidence of RRI in a group of runners training an 8 week training program with a mean duration increase of 24% per week. However, although not significant, the mean survival time of runners in the 13 week training group was 212 minutes, compared to 167 minutes in runners of the 8 week training group.
In fourteen studies, the relationship between training intensity and development of RRI was investigated.1,4,11,13,14,17,63–70
Jacobs et al13
found a pace above 8 min/mile to increase the risk of injury as compared with a pace below 8 min/mile (p<0.05). Hootman et al4
found a reduced odds ratio (0.51 [0.35, 0.74]) for sustaining an injury among males who ran at above a 15 min/mile pace compared to those who ran at a faster pace (p=0.0004). A similar significant difference was found for female subjects (p≤0.05). However, lack of adjustment for other predictor variables such as weekly mileage weakened this association. This is supported by Marti et al,14
who found that running speed calculated from 10 mile race time was positively related to injury incidence in univariate analysis, but adjustment for mileage clearly weakened this association. In eight studies, no significant relationship between average training pace and likelihood of injury were found.1,11,17,63–66,68
Wen et al63,69
reported that no association was found between running pace and injury. However, it was reported that interval training increased the risk of shin injury (p<0.05).63
In a case-control study, Messier et al67
found runners with iliotibial band friction syndrome to run on average 3 seconds/mile faster than the control group during non-competition training (p=0.05). McCrory et al70
found training pace to be a significant (p≤0.05) discriminator between persons with achilles tendinitis when they examined pace in minutes per kilometer, where the pace of those injured was 4.64 ± 0.08 as compared to controls which was 4.87 ± 0.07.
In eight articles, the relationship between training frequency and development of RRI was investigated.5,6,11,13,22,30,64,71,74
Results are presented in . In six articles, RRIs in general were investigated;5,11,13,22,74
one investigated front thigh injuries,6
and one shin splint.71
In several studies, an increased risk, relative risk, or odds ratio for sustaining an RRI was reported when the weekly running frequency increased.11,13,22,30,69,71,74
Persons running 6-7 times per week had the highest risk. On the contrary, Taunton et al5
found an increased risk of injury among females running one time per week. Males also showed a similar trend, although it was not statistically significant (p=0.064). Satterthwaite et al6
found the odds ratio for sustaining an anterior thigh injury increased by 1.19 [1.05-1.34] per one day increment in running frequency. No significant association was found for hamstring, hip, knee, or calf injuries. Knobloch et al71
reported an increased risk of shin splints among individuals running more than five days per week.
Relationship between running frequency (days per week) and Running Related Injury (RRI). OR = odds ratio; RR = relative risk; CI = Confidence Interval.