R EVI EW ARTI CLE
Sports Medicine 13 (6): 408-422, 1992 0112.1642/92/0006-0408/$07.50/0 © Adis International Limited . All rights reserved. SPOll38a
Factors Related to the Incidence of Running Injuries
A Review
Jean H. Hoeberigs Instituut Sportgeneeskunde Limburg, University of Limburg, Maastricht, The Netherlands
Contents 408 409 4/0
4/0 4// 4// 4/2 4/2 4/3 4/5 4/7 4/7 4/9
Summary
Summary I. Basics of Epidemiology 2. Running Injury Epidemiology 2.1 Denominator-Based Incidence Rates 2.2 Period of Observation 2.3 Runners: Who is in the Denominator? 2.3.1 Origin of the Study Population with Reference to the Total Running Population 2.3.2 Sampled Population 2.3.3 Response Population 2.4 Running Injuries: What is in the Numerator? 2.5 Aetiological Factors in Running Injuries 2.6 Developments in Running Injury Epidemiology 3. Conclusions and Comments
The term incidence is interpreted in many different ways in the literature. Running injury epidemiology should include denominator-based incidence rates, in which the number of new injuries observed during I year is related to the population of runners at risk. In IO studies with denominator-based incidences selected from the literature, the annual incidence rates of injured runners vary from 24 to 65%. Comparison of denominator-based incidence rates from different studies requires a discussion of the denominator and of the numerator; i.e. the study population and the definition of running injury. Injury definitions differ from one study to another. Study populations are particular subgroups of the total running population and they differ from one study to another. Subgroups may differ in origin: volunteers, runners from a mailing list or entrants of a road race. Incidence rates are higher among supervised volunteers than among listed runners, and higher among both these groups than among race-entrants.' The choice from the universe of the running population and the used injury definition are methodological issues. Incidence is dependently associated with the prevalence of predisposing running injury factors. There is consistent epidemiological support for the role of a few aetiological factors; i.e. higher mileage per week, previous running injury, higher running speed and lesser running experience. Higher mileage per week is probably the strongest predictor. In the selected injury studies, mileage per week differs from one study population to another. Differences in mileage per week do not explain differences in incidence rate between these studies. In conclusion, caution must be taken when comparing annual incidence rates of different studies. Methodological issues are at least as important as aetiological factors. Study populations
The Incidence of Running Injuries
409
may refer to different selections of the universe of the running population. The lengths of observation periods and 'running injury' definitions may differ from one study to another.
Participation in running activities rapidly grew in the mid-seventies in many countries and so did the number of injuries. The incidence of running injuries seen by physicians increased dramatically (Clement et al. 1981). This increase, however, is a numerator-based incidence, in which the number of running injuries is related to the number of other medical problems for which physicians are consulted. To assess the magnitude of a disorder or a phenomenon (e.g. birth rate) in the population, epidemiologists prefer to work with a denominatorbased rather than with a numerator-based incidence (Caspersen 1989; Krauss & Conroy 1984; Koplan et al. 1985; MacMahon & Pugh 1970; Walter & Hart 1990). Denominator-based incidence refers to the number of individuals in the population at risk. In theory, a disease can affect any individual in the open population. For instance, lung cancer may affect any individual, although special categories (older men, smokers) are more at risk. Running injuries, however, will occur only among people who participate in running activities. With regard to these activities, the open population consists of 2 categories of individuals: runners and non runners. Runners are at risk and nonrunners are not at risk for running injuries. A clear appreciation of the manner in which a 'runner' is defined is important in order to interpret adequately the phenomenon of running injuries in the population of runners at risk. In sports injury epidemiology, Garrick and Requa (1978) were the first to relate the number of injured athletes to the population at risk. In order to investigate the degree to which one can compare the incidence rate of different studies, this review article deals with the denominator-based incidence of running injuries. First, some theoretical principles and basic terminology from epidemiology will be introduced. Secondly, incidence values from a selected set of studies concerning running injuries will be given. Finally, a step-by-
step analysis of study populations, running injuries and study methods in the selected studies will be undertaken, in order to show factors which may explain the variety in resultant incidence values.
1. Basics of Epidemiology Epidemiology can be regarded as a scientific method to establish the magnitude of a disorder in the population and to find reasons for the occurrence of this disorder in this population. Descriptive epidemiology starts with the count of the number of cases ('running injuries' or 'injured runners') and the count of the population at risk ('runners'). Counting is done during a specified period of observation (retrospective or prospective). In many fields of epidemiology the specified period of observation is 1 year. For instance, annual mortality due to lung cancer is 70 per 100 000 population in some European countries. Such an annual denominator based mortality rate gives a clear description of the magnitude of a problem. Caspersen (1989) advocated the adoption of clear descriptions of incidence rates in physical activity epidemiology, for example, by adopting similar periods of observation. In epidemiology, 2 basic terms are distinguished: incidence and prevalence. Incidence of injuries is defined as the number of new injuries during the period of observation. Prevalence of injuries pertains to the total number of injuries in this period; prevalence also includes injuries which developed before, but are still present at the start of the observation period. In cross-sectional or 'snapshot' (Paffenberger 1988) studies the term point prevalence is in use. In these studies the length of observation is limited to one moment only. Most epidemiological studies about running injuries present incidence rather than prevalence numbers. Having counted running injuries incurred in 1 year and the population involved, the annual in-
Sports Medicine 13 (6) 1992
410
cidence rate can be assessed by dividing the first number by the second. The incidence of running injuries per year, a, is defined as: a
No. of new running injuries per year No. of runners involved
From this formula it is clear that special attention must be given to the definition of 'running injuries' and the definition of 'runners'. In addition to the measurement of the number of running injuries, studies may measure the number of 'injured runners'. The incidence of injured runners per year, {3, is defined as: {3
is a determinant of the numerator in the incidence rate). 4. The presence or absence of aetiological factors (since this increases or reduces the incidence of running injures).
No. of newly injured runners per year No. of runners involved
The analytical part of epidemiology is more complex than the descriptive part and can be considered as the second step in epidemiology, following the descriptive first step. With the incidence rate of injured runners, the analytical step tries somehow to explain why the incidence rate has reached the value as measured. The analysis tries to show why some runners were injured and others were not. If the incidence rate is high, it is most interesting to know the factors that contributed to the development of these injuries, i.e. aetiological factors. If, on the other hand, the incidence rate is low, it is probably more interesting to know the factors that protected this population against a higher incidence rate, i.e. protective factors. In order to explain why running injuries occur, knowledge of the protective and aetiological factors is necessary. Whether the incidence is 'high' or 'low', the development of running injuries is dependent on the presence of protective and aetiological factors. In summary, 4 issues are important in the analysis of factors related to the incidence of running injuries: 1. The length of the observation period. 2. The definition of 'runners' (since this is a determinant of the denominator value in the incidence rate). 3. The definition of 'running injuries' (since this
2. Running Injury Epidemiology The number of running injury studies in the scientific literature is overwhelming. Denominatorbased incidence studies of running injuries are scarce in comparison to the large number of studies of which most include only clinical data (Clement et al. 1981; Glick & Katch, 1970; James et al. 1978). Running injury studies differ from each other in various ways. Definitions of 'runners' and 'running injuries' may differ and postulated aetiological factors (such as bad weather conditions) may differ in magnitude from one study to another. As a consequence of these differences, different incidence rates may result. 2.1 Denominator-Based Incidence Rates Based upon the length of their observation periods, 10 studies have been selected (see table I). These studies have a denominator-based incidence which was obtained during a l-year period of observation. Annual rates have been roughly calculated from the original data in 2 studies (Bovens et al. 1989; Jacobs & Berson 1986). The annual incidence rates of running injuries (a) and of injured runners ({3) are shown in table I. In 6 studies the number of injured runners was given but not the number of injuries (Holmich et al. 1988, 1989; Jacobs & Berson 1986; Koplan et al. 1982; Macera et al. 1989; Marti 1988; Marti et al. 1988). In these cases only {3-values could be calculated. Consequently, the a value is given in a few studies and the {3 value in all of the selected studies (see table I). In a study among 106 runners we observed an 80% a value (Hoeberigs et al. 1990) and a concomitant 49% {3 value (previously unpublished). Focusing upon {3 values, these incidence rates vary from 0.24 to 0.65. In other words, in the 10
The Incidence of Running Injuries
411
Table I. Denominator-based annual incidence rates of running injuries (a) and injured runners (Il) in 10 selected studies Reference
Koplan et al. (1982) Jacobs & Berson (1986) Blair et al. (1987) [study 1] Lysholm & Wiklander
Sex
Walter et al, (1989)
Bovens et al. (1989)
A (injuries/N)
B (injured athletes/N)
Duration of study (years)
a (A/year)
Il (B/year)
M+F M+F M+F
1168 451 438
? ? >0.33
0.37 0.47 0.24
1 2 1
? ? >0.33
0.37 0.24 0.24
M+F
All 60 LDR 28 4358 428 60 1426 485 98 1265
0.92 0.64 ? ? ? ? ? ? ?
1 1 1 1 1 1 1 1 1
0.92 0.64 ? ? ? ? ? ? ?
980t 301 73
0.58
0.65 M7 0.46 0.40 0.43 0.31 0.52 0.49 0.50 0.49 0.46
0.58
2.38
0.83
~t
0.65 0.57 0.46 0.40 0.43 0.31 0.52 0.49 0.50 0.49 0.46
1.6
1.49
0.52
(1987) Marti at al, (1988); Marti (1988) Holmich et al. (1988) Holmich et al, (1989) Macera et al. (1989)
N
M F M M+F M F M+F M F M+F
Abbreviations: M = male, F = female; LDR = long distance runners.
selected studies 24 to 65% of the runners sustained 1 or more injuries in the year of observation. The highest l3-value is 2.7 times the lowest l3-value. In order to understand these differences in incidence rates, it is necessary to analyse study populations, injuries and study methods in these studies. 2.2 Period 'of Observation In 2 of the 10 selected studies the period of observation was more than 1 year (Bovens et al. 1989; Jacobs & Berson 1986). If the incidence rates in these studies had not been corrected for the time of observation, the 83% value in the study of Bovens et al. (1989) would have been the highest value observed so far. An observation period of 1 year is of paramount importance in the interpretation of running injury epidemiology, because of the effect that weather conditions or seasonal changes can have upon the occurrence of sports injuries (Dick 1990; Ekstrand et al. 1990; Engstrom et al. 1990; Porter 1984; Richards & Richards 1984;Ridley et al. 1990;Smith 1980).
2.3 Runners: Who is in the Denominator? How many kilometres a week must somebody run, before he or she is called a runner? Catching a bus or a train may involve running several metres. Although the person involved may hurt herself or himself while doing this run, a brief increase in walking speed or running speed does not classify them as a runner. The question is: If mileage per week is the distinctive criterion between runners and nonrunners, do runners also differ from nonrunners with regards to health status or injury proneness? Runners seem to differ from nonrunners with regard to health beliefs and health-related behaviours (Walsh 1985). Runners also seem to prefer health professionals other than general practitioners as their initial contact to seek advice for a running injury (Furman 1986). However, the main question, whether runners are healthier than nonrunners, has not yet been answered. The answer can be important with regard to the risk of sustaining an injury. Those who belong to a less healthy part of the population ('current population
412
of nonrunners'?) might have a higher injury risk compared with healthier people (those who already participate in running?). If running popularity increases, 'the future population of runners' is less healthy than 'the current population of runners', under the hypothesis that 'current runners' are healthier than 'current nonrunners'. Consequently, running injury incidence rate in 'the future population of runners' will exceed 'current' incidence rate, under the hypothesis that healthier people have lower injury risk. This lack of basic knowledge should be noted before focusing upon the running population. The source of the population at risk is the universe of the running population. There is a wide spectrum of runners, for example, between sprinters and ultramarathon runners, between recreational joggers and elite runners, between novices and experienced runners, between track runners and cross-country runners. It is not feasible to capture the total running population in the denominator of a single study. However, in order to study the incidence of running injuries, epidemiologists need a study population of runners. Study populations are derived from some subgroup of the total running population. Each runner in a study meets the same criteria, which are more or less specific for the study: for instance, entrants of a l6km race, men, older than 16 (Marti et al. 1988). The final study population differs from one study to another (see table II). 2.3.1 Origin of the Study Population with Reference to the Total Running Population Subgroups of the running population may differ in origin. Many running injury studies deal with a population consisting of race entrants (Holmich et al. 1988, 1989; Jacobs & Berson 1986; Koplan et aI. 1982; Marti et al. 1988;Walter et al. 1988, 1989). Some populations consist of members of a club (Blair et al. 1987; Lysholm & Wiklander 1987) or of volunteers entering a supervised training programme (Bovens et al. 1989). Occasionally, a population consists of persons on a mailing list who wished to be notified of road races (Macera et al. 1989) or a population consists of running members
Sports Medicine 13 (6) 1992
of a national track and field association (Hoeberigs et al. 1990). Epidemiologists do not choose a subgroup population at random. The choice from the universe of the running population is determined by given circumstances, like the availability of mailing lists or the organisation of road races in the neighbourhood of the institute that carries out the research. Mailing lists include injured and noninjured runners. Entrants in a road race may exclude injured or unfit runners (Clough et al. 1987, 1989). In theory, the incidence rate among entrants in a road race may be lower than among runners on a mailing list (fig. 1). Except for the 24% incidence rate from Blair et al. (1987), the 52 and 49% incidence rates in the study of Macera et al. (1989) and the 49% value in the study of Hoeberigs et al. (1990) were higher than in 5 studies conducted among race entrants (Holmich et al. 1988, 1989;Jacobs & Berson 1986; Koplan et al. 1982; Marti et al. 1988a, b). In one study (Walter et al. 1989)the incidence among race entrants is the same as in the study of Macera et al. (1989) and Hoeberigs et al. (1990). However, it should be noted that Macera et al. (1989) focused upon lower extremity injuries only. Runners who are directly supervised (Bovens et al. 1989; Lysholm & Wiklander 1987) showed higher incidence rates than race entrants or persons from a mailing list (see fig. 1). A more continuous monitoring system may be the main reason for the higher incidence. Supervised runners have more occasions to report their injuries than runners who only once receive a questionnaire. Obviously the origin of the study population can be regarded as a predictor of the incidence rate, i.e. the incidence rate is higher among directly supervised runners than among runners on a mailing list and higher among runners on a mailing list than among entrants in a road race. 2.3.2 Sampled Population A subgroup from the total running population may be too large for a study. The researchers may be urged to take a sample from this subgroup before acquiring injury data. Some studies include all members on a list (Blair et al. 1987; Lysholm &
413
The Incidence of Running Injuries
Table II. Designs and populations in 10 selected running injury studies. In studies marked with an asterisk, there was a check for
nonrespondents Reference
Design
Koplan et al. (1982)
Retrospective, 1 year
Jacobs & Berson (1986) Blair et al. (1987) [study 1]
Retrospective, 2 years Retrospective, 1 year
Lysholm and Wiklander (1987)
Prospective, 1 year
Marti et al. (1988); Marti (1988) Holmich et al. (1988)
Retrospective, 1 year Cross-sectional and 1 year restrospective Cross-sectional and 1 year retrospective Prospective, 1 year Retrospective (written): 1 year Prospective (by phone): 1 year Prospective, 18-20 months
Helmich et al. (1989) Macera et al. (1989) Walter et al. (1989)
Bovens et al. (1989)
Age (years)
Definition of athlete
693 M 730 F
M: 33.4 F: 29.9
355 M 96 F 438 M+F
M: 33.9 (14-64) F: 32.4 (8-57) 43.8±10.5
1 year after 1Okm race random selection of raceentrants; 55% of 1250 M and 58% of 1250 F responded'; 1 year after race 89% (614) of men and 79% (580) of women were still running; a current runner is someone who runs or jogs at least 6 miles/week 21% of entrants on 10km-race (2664); 82% of sample (550) responded' Members of fitness club, who reported at least 10 miles (16km) running during at least 1 week within a 3-month period (response: 61%) Members of 2 clubs, including 19 sprinters (S), 13 middle distance (MDR) and 28 long distance runners (LOR). All responded Entrants in a 16km race. 84% response"
Study population
44 M 16 F 4358 M 428 F 60 M
S: 20.6±3.8 MDR: 18.6±2.4 LOR: 34.5 ± 7.4 35 (>16 only) 28 (18-51)
Elite marathon runners «2h 40 min) entrants in marathon. All responded
M: 34 (11-77)
68% response of entrants in marathon race; focus upon men
485 M 98 F 985 M 303 F
41.6 (13-75) 36.1 (22-64) >14
58 M 15 F
M: 35.2±7.9 F: 33.5±6.4
61% response' of persons on a mailing list, who wished to be notified of road races. Entrants in 2 pairs of running events 4, 16, 5.6 and 22.4km (67% responded initially; of these respondents 88% were followed prospectively, which is the study population') 63% of 115 volunteers (mainly novices in running) who followed a supervised training programme and who had adequate diaries"
1310 M 116 F
Wiklander 1987; Macera et al. 1989), others select by gender (Holmich et al. 1989), age (Walter et al. 1989) or both gender and age (Marti et al. 1988a, b). Random sampling (Koplan et al. 1982) should be preferred to nonrandom or less random sampling (Jacobs & Berson 1986) if the total subgroup population cannot be studied. If sampling occurs, the sampled population can be regarded as the base of the final study population. Such a study should include a check for randomness. Methods of obtaining data (injuries, runningrelated data and demographical variables) from runners in the sampled populations is the concern
of epidemiologicalinvestigators. Unfortunately, not every runner in a chosen sample will respond.
2.3.3 Response Population The highest response rate in a running injury survey was 84%, achieved in a study by Marti et al. (1988). Other response rates were below 84%, except for 2 survey studies in smaller populations (Holmich et al. 1988; Lysholm & Wiklander 1987). The response population is the actual study population from which the injury data are obtained. With reference to the injury effect, the response population should not be a selection of the sam-
Sports Medicine 13 (6) 1992
414
pled population. It is important to check for absence of response bias by comparing nonrespondents with respondents. This has been done in most of the selected studies (see asterisks in table II). Until now, epidemiological studies could not find clear differences in incidence rate between respondents and nonrespondents (Koplan et al. 1982; Macera et al. 1989). However, there are other differences between these categories. Younger runners are probably less compliant in responding to questionnaires than older runners. Respondents have consistently been somewhat older than nonrespondents (Bovens et al. 1989; Koplan et al. 1982; Macera et al. 1989). The numerator definition in some studies
showed another remarkable finding. Macera et al. (1989) stated that 47% of the nonrespondents were not eligible to participate in the study. This became clear after these authors tried to match the running activities of nonrespondents with the criteria that these authors accepted for their definition of a runner. Koplan et al. (1982) reported that 60% of the nonrespondents continued to run, compared to 88% of the respondents. These remarkable findings are relevant to the basics of denominator definition. An individual does not stay 'a runner' forever after she or he has met the criteria of the definition of 'a runner'. In conclusion, respondents are older than nonrespondents and are more likely to continue running.
0.65
• Lysholm (sprinters)
0.60 • Lysholm (long distance runners)
0.55 D Macera
• Bovens
III
~
l: l:
0.50
2
"0
l!! :::l
:5"
• Walter D Walter
• 0 Hoeberigs; Macera
D 0 Marti; Walter
0.45
'0
D Holmich (elite)
Ql
U l: Ql
"0
'0 .!: OJ
:::l l: l:
-c
0.40
o
Marti
• Koplan
0.35
0.30
0.24
0= Males 0= Females Both sexes
• Holmich
•=
• Jacobs
• Blair
Race entrants
Listed runners
Supervised runners
Fig. 1. Annual incidence of injured runners by origin of subgroup as a selection from the universe of the running population. For details of references see table II.
The Incidence of Running Injuries
415
2.4 Running Injuries: What is in the Numerator?
Sports Medicine 13 (6): 408-422, 1992 0112.1642/92/0006-0408/$07.50/0 © Adis International Limited . All rights reserved. SPOll38a
Factors Related to the Incidence of Running Injuries
A Review
Jean H. Hoeberigs Instituut Sportgeneeskunde Limburg, University of Limburg, Maastricht, The Netherlands
Contents 408 409 4/0
4/0 4// 4// 4/2 4/2 4/3 4/5 4/7 4/7 4/9
Summary
Summary I. Basics of Epidemiology 2. Running Injury Epidemiology 2.1 Denominator-Based Incidence Rates 2.2 Period of Observation 2.3 Runners: Who is in the Denominator? 2.3.1 Origin of the Study Population with Reference to the Total Running Population 2.3.2 Sampled Population 2.3.3 Response Population 2.4 Running Injuries: What is in the Numerator? 2.5 Aetiological Factors in Running Injuries 2.6 Developments in Running Injury Epidemiology 3. Conclusions and Comments
The term incidence is interpreted in many different ways in the literature. Running injury epidemiology should include denominator-based incidence rates, in which the number of new injuries observed during I year is related to the population of runners at risk. In IO studies with denominator-based incidences selected from the literature, the annual incidence rates of injured runners vary from 24 to 65%. Comparison of denominator-based incidence rates from different studies requires a discussion of the denominator and of the numerator; i.e. the study population and the definition of running injury. Injury definitions differ from one study to another. Study populations are particular subgroups of the total running population and they differ from one study to another. Subgroups may differ in origin: volunteers, runners from a mailing list or entrants of a road race. Incidence rates are higher among supervised volunteers than among listed runners, and higher among both these groups than among race-entrants.' The choice from the universe of the running population and the used injury definition are methodological issues. Incidence is dependently associated with the prevalence of predisposing running injury factors. There is consistent epidemiological support for the role of a few aetiological factors; i.e. higher mileage per week, previous running injury, higher running speed and lesser running experience. Higher mileage per week is probably the strongest predictor. In the selected injury studies, mileage per week differs from one study population to another. Differences in mileage per week do not explain differences in incidence rate between these studies. In conclusion, caution must be taken when comparing annual incidence rates of different studies. Methodological issues are at least as important as aetiological factors. Study populations
The Incidence of Running Injuries
409
may refer to different selections of the universe of the running population. The lengths of observation periods and 'running injury' definitions may differ from one study to another.
Participation in running activities rapidly grew in the mid-seventies in many countries and so did the number of injuries. The incidence of running injuries seen by physicians increased dramatically (Clement et al. 1981). This increase, however, is a numerator-based incidence, in which the number of running injuries is related to the number of other medical problems for which physicians are consulted. To assess the magnitude of a disorder or a phenomenon (e.g. birth rate) in the population, epidemiologists prefer to work with a denominatorbased rather than with a numerator-based incidence (Caspersen 1989; Krauss & Conroy 1984; Koplan et al. 1985; MacMahon & Pugh 1970; Walter & Hart 1990). Denominator-based incidence refers to the number of individuals in the population at risk. In theory, a disease can affect any individual in the open population. For instance, lung cancer may affect any individual, although special categories (older men, smokers) are more at risk. Running injuries, however, will occur only among people who participate in running activities. With regard to these activities, the open population consists of 2 categories of individuals: runners and non runners. Runners are at risk and nonrunners are not at risk for running injuries. A clear appreciation of the manner in which a 'runner' is defined is important in order to interpret adequately the phenomenon of running injuries in the population of runners at risk. In sports injury epidemiology, Garrick and Requa (1978) were the first to relate the number of injured athletes to the population at risk. In order to investigate the degree to which one can compare the incidence rate of different studies, this review article deals with the denominator-based incidence of running injuries. First, some theoretical principles and basic terminology from epidemiology will be introduced. Secondly, incidence values from a selected set of studies concerning running injuries will be given. Finally, a step-by-
step analysis of study populations, running injuries and study methods in the selected studies will be undertaken, in order to show factors which may explain the variety in resultant incidence values.
1. Basics of Epidemiology Epidemiology can be regarded as a scientific method to establish the magnitude of a disorder in the population and to find reasons for the occurrence of this disorder in this population. Descriptive epidemiology starts with the count of the number of cases ('running injuries' or 'injured runners') and the count of the population at risk ('runners'). Counting is done during a specified period of observation (retrospective or prospective). In many fields of epidemiology the specified period of observation is 1 year. For instance, annual mortality due to lung cancer is 70 per 100 000 population in some European countries. Such an annual denominator based mortality rate gives a clear description of the magnitude of a problem. Caspersen (1989) advocated the adoption of clear descriptions of incidence rates in physical activity epidemiology, for example, by adopting similar periods of observation. In epidemiology, 2 basic terms are distinguished: incidence and prevalence. Incidence of injuries is defined as the number of new injuries during the period of observation. Prevalence of injuries pertains to the total number of injuries in this period; prevalence also includes injuries which developed before, but are still present at the start of the observation period. In cross-sectional or 'snapshot' (Paffenberger 1988) studies the term point prevalence is in use. In these studies the length of observation is limited to one moment only. Most epidemiological studies about running injuries present incidence rather than prevalence numbers. Having counted running injuries incurred in 1 year and the population involved, the annual in-
Sports Medicine 13 (6) 1992
410
cidence rate can be assessed by dividing the first number by the second. The incidence of running injuries per year, a, is defined as: a
No. of new running injuries per year No. of runners involved
From this formula it is clear that special attention must be given to the definition of 'running injuries' and the definition of 'runners'. In addition to the measurement of the number of running injuries, studies may measure the number of 'injured runners'. The incidence of injured runners per year, {3, is defined as: {3
is a determinant of the numerator in the incidence rate). 4. The presence or absence of aetiological factors (since this increases or reduces the incidence of running injures).
No. of newly injured runners per year No. of runners involved
The analytical part of epidemiology is more complex than the descriptive part and can be considered as the second step in epidemiology, following the descriptive first step. With the incidence rate of injured runners, the analytical step tries somehow to explain why the incidence rate has reached the value as measured. The analysis tries to show why some runners were injured and others were not. If the incidence rate is high, it is most interesting to know the factors that contributed to the development of these injuries, i.e. aetiological factors. If, on the other hand, the incidence rate is low, it is probably more interesting to know the factors that protected this population against a higher incidence rate, i.e. protective factors. In order to explain why running injuries occur, knowledge of the protective and aetiological factors is necessary. Whether the incidence is 'high' or 'low', the development of running injuries is dependent on the presence of protective and aetiological factors. In summary, 4 issues are important in the analysis of factors related to the incidence of running injuries: 1. The length of the observation period. 2. The definition of 'runners' (since this is a determinant of the denominator value in the incidence rate). 3. The definition of 'running injuries' (since this
2. Running Injury Epidemiology The number of running injury studies in the scientific literature is overwhelming. Denominatorbased incidence studies of running injuries are scarce in comparison to the large number of studies of which most include only clinical data (Clement et al. 1981; Glick & Katch, 1970; James et al. 1978). Running injury studies differ from each other in various ways. Definitions of 'runners' and 'running injuries' may differ and postulated aetiological factors (such as bad weather conditions) may differ in magnitude from one study to another. As a consequence of these differences, different incidence rates may result. 2.1 Denominator-Based Incidence Rates Based upon the length of their observation periods, 10 studies have been selected (see table I). These studies have a denominator-based incidence which was obtained during a l-year period of observation. Annual rates have been roughly calculated from the original data in 2 studies (Bovens et al. 1989; Jacobs & Berson 1986). The annual incidence rates of running injuries (a) and of injured runners ({3) are shown in table I. In 6 studies the number of injured runners was given but not the number of injuries (Holmich et al. 1988, 1989; Jacobs & Berson 1986; Koplan et al. 1982; Macera et al. 1989; Marti 1988; Marti et al. 1988). In these cases only {3-values could be calculated. Consequently, the a value is given in a few studies and the {3 value in all of the selected studies (see table I). In a study among 106 runners we observed an 80% a value (Hoeberigs et al. 1990) and a concomitant 49% {3 value (previously unpublished). Focusing upon {3 values, these incidence rates vary from 0.24 to 0.65. In other words, in the 10
The Incidence of Running Injuries
411
Table I. Denominator-based annual incidence rates of running injuries (a) and injured runners (Il) in 10 selected studies Reference
Koplan et al. (1982) Jacobs & Berson (1986) Blair et al. (1987) [study 1] Lysholm & Wiklander
Sex
Walter et al, (1989)
Bovens et al. (1989)
A (injuries/N)
B (injured athletes/N)
Duration of study (years)
a (A/year)
Il (B/year)
M+F M+F M+F
1168 451 438
? ? >0.33
0.37 0.47 0.24
1 2 1
? ? >0.33
0.37 0.24 0.24
M+F
All 60 LDR 28 4358 428 60 1426 485 98 1265
0.92 0.64 ? ? ? ? ? ? ?
1 1 1 1 1 1 1 1 1
0.92 0.64 ? ? ? ? ? ? ?
980t 301 73
0.58
0.65 M7 0.46 0.40 0.43 0.31 0.52 0.49 0.50 0.49 0.46
0.58
2.38
0.83
~t
0.65 0.57 0.46 0.40 0.43 0.31 0.52 0.49 0.50 0.49 0.46
1.6
1.49
0.52
(1987) Marti at al, (1988); Marti (1988) Holmich et al. (1988) Holmich et al, (1989) Macera et al. (1989)
N
M F M M+F M F M+F M F M+F
Abbreviations: M = male, F = female; LDR = long distance runners.
selected studies 24 to 65% of the runners sustained 1 or more injuries in the year of observation. The highest l3-value is 2.7 times the lowest l3-value. In order to understand these differences in incidence rates, it is necessary to analyse study populations, injuries and study methods in these studies. 2.2 Period 'of Observation In 2 of the 10 selected studies the period of observation was more than 1 year (Bovens et al. 1989; Jacobs & Berson 1986). If the incidence rates in these studies had not been corrected for the time of observation, the 83% value in the study of Bovens et al. (1989) would have been the highest value observed so far. An observation period of 1 year is of paramount importance in the interpretation of running injury epidemiology, because of the effect that weather conditions or seasonal changes can have upon the occurrence of sports injuries (Dick 1990; Ekstrand et al. 1990; Engstrom et al. 1990; Porter 1984; Richards & Richards 1984;Ridley et al. 1990;Smith 1980).
2.3 Runners: Who is in the Denominator? How many kilometres a week must somebody run, before he or she is called a runner? Catching a bus or a train may involve running several metres. Although the person involved may hurt herself or himself while doing this run, a brief increase in walking speed or running speed does not classify them as a runner. The question is: If mileage per week is the distinctive criterion between runners and nonrunners, do runners also differ from nonrunners with regards to health status or injury proneness? Runners seem to differ from nonrunners with regard to health beliefs and health-related behaviours (Walsh 1985). Runners also seem to prefer health professionals other than general practitioners as their initial contact to seek advice for a running injury (Furman 1986). However, the main question, whether runners are healthier than nonrunners, has not yet been answered. The answer can be important with regard to the risk of sustaining an injury. Those who belong to a less healthy part of the population ('current population
412
of nonrunners'?) might have a higher injury risk compared with healthier people (those who already participate in running?). If running popularity increases, 'the future population of runners' is less healthy than 'the current population of runners', under the hypothesis that 'current runners' are healthier than 'current nonrunners'. Consequently, running injury incidence rate in 'the future population of runners' will exceed 'current' incidence rate, under the hypothesis that healthier people have lower injury risk. This lack of basic knowledge should be noted before focusing upon the running population. The source of the population at risk is the universe of the running population. There is a wide spectrum of runners, for example, between sprinters and ultramarathon runners, between recreational joggers and elite runners, between novices and experienced runners, between track runners and cross-country runners. It is not feasible to capture the total running population in the denominator of a single study. However, in order to study the incidence of running injuries, epidemiologists need a study population of runners. Study populations are derived from some subgroup of the total running population. Each runner in a study meets the same criteria, which are more or less specific for the study: for instance, entrants of a l6km race, men, older than 16 (Marti et al. 1988). The final study population differs from one study to another (see table II). 2.3.1 Origin of the Study Population with Reference to the Total Running Population Subgroups of the running population may differ in origin. Many running injury studies deal with a population consisting of race entrants (Holmich et al. 1988, 1989; Jacobs & Berson 1986; Koplan et aI. 1982; Marti et al. 1988;Walter et al. 1988, 1989). Some populations consist of members of a club (Blair et al. 1987; Lysholm & Wiklander 1987) or of volunteers entering a supervised training programme (Bovens et al. 1989). Occasionally, a population consists of persons on a mailing list who wished to be notified of road races (Macera et al. 1989) or a population consists of running members
Sports Medicine 13 (6) 1992
of a national track and field association (Hoeberigs et al. 1990). Epidemiologists do not choose a subgroup population at random. The choice from the universe of the running population is determined by given circumstances, like the availability of mailing lists or the organisation of road races in the neighbourhood of the institute that carries out the research. Mailing lists include injured and noninjured runners. Entrants in a road race may exclude injured or unfit runners (Clough et al. 1987, 1989). In theory, the incidence rate among entrants in a road race may be lower than among runners on a mailing list (fig. 1). Except for the 24% incidence rate from Blair et al. (1987), the 52 and 49% incidence rates in the study of Macera et al. (1989) and the 49% value in the study of Hoeberigs et al. (1990) were higher than in 5 studies conducted among race entrants (Holmich et al. 1988, 1989;Jacobs & Berson 1986; Koplan et al. 1982; Marti et al. 1988a, b). In one study (Walter et al. 1989)the incidence among race entrants is the same as in the study of Macera et al. (1989) and Hoeberigs et al. (1990). However, it should be noted that Macera et al. (1989) focused upon lower extremity injuries only. Runners who are directly supervised (Bovens et al. 1989; Lysholm & Wiklander 1987) showed higher incidence rates than race entrants or persons from a mailing list (see fig. 1). A more continuous monitoring system may be the main reason for the higher incidence. Supervised runners have more occasions to report their injuries than runners who only once receive a questionnaire. Obviously the origin of the study population can be regarded as a predictor of the incidence rate, i.e. the incidence rate is higher among directly supervised runners than among runners on a mailing list and higher among runners on a mailing list than among entrants in a road race. 2.3.2 Sampled Population A subgroup from the total running population may be too large for a study. The researchers may be urged to take a sample from this subgroup before acquiring injury data. Some studies include all members on a list (Blair et al. 1987; Lysholm &
413
The Incidence of Running Injuries
Table II. Designs and populations in 10 selected running injury studies. In studies marked with an asterisk, there was a check for
nonrespondents Reference
Design
Koplan et al. (1982)
Retrospective, 1 year
Jacobs & Berson (1986) Blair et al. (1987) [study 1]
Retrospective, 2 years Retrospective, 1 year
Lysholm and Wiklander (1987)
Prospective, 1 year
Marti et al. (1988); Marti (1988) Holmich et al. (1988)
Retrospective, 1 year Cross-sectional and 1 year restrospective Cross-sectional and 1 year retrospective Prospective, 1 year Retrospective (written): 1 year Prospective (by phone): 1 year Prospective, 18-20 months
Helmich et al. (1989) Macera et al. (1989) Walter et al. (1989)
Bovens et al. (1989)
Age (years)
Definition of athlete
693 M 730 F
M: 33.4 F: 29.9
355 M 96 F 438 M+F
M: 33.9 (14-64) F: 32.4 (8-57) 43.8±10.5
1 year after 1Okm race random selection of raceentrants; 55% of 1250 M and 58% of 1250 F responded'; 1 year after race 89% (614) of men and 79% (580) of women were still running; a current runner is someone who runs or jogs at least 6 miles/week 21% of entrants on 10km-race (2664); 82% of sample (550) responded' Members of fitness club, who reported at least 10 miles (16km) running during at least 1 week within a 3-month period (response: 61%) Members of 2 clubs, including 19 sprinters (S), 13 middle distance (MDR) and 28 long distance runners (LOR). All responded Entrants in a 16km race. 84% response"
Study population
44 M 16 F 4358 M 428 F 60 M
S: 20.6±3.8 MDR: 18.6±2.4 LOR: 34.5 ± 7.4 35 (>16 only) 28 (18-51)
Elite marathon runners «2h 40 min) entrants in marathon. All responded
M: 34 (11-77)
68% response of entrants in marathon race; focus upon men
485 M 98 F 985 M 303 F
41.6 (13-75) 36.1 (22-64) >14
58 M 15 F
M: 35.2±7.9 F: 33.5±6.4
61% response' of persons on a mailing list, who wished to be notified of road races. Entrants in 2 pairs of running events 4, 16, 5.6 and 22.4km (67% responded initially; of these respondents 88% were followed prospectively, which is the study population') 63% of 115 volunteers (mainly novices in running) who followed a supervised training programme and who had adequate diaries"
1310 M 116 F
Wiklander 1987; Macera et al. 1989), others select by gender (Holmich et al. 1989), age (Walter et al. 1989) or both gender and age (Marti et al. 1988a, b). Random sampling (Koplan et al. 1982) should be preferred to nonrandom or less random sampling (Jacobs & Berson 1986) if the total subgroup population cannot be studied. If sampling occurs, the sampled population can be regarded as the base of the final study population. Such a study should include a check for randomness. Methods of obtaining data (injuries, runningrelated data and demographical variables) from runners in the sampled populations is the concern
of epidemiologicalinvestigators. Unfortunately, not every runner in a chosen sample will respond.
2.3.3 Response Population The highest response rate in a running injury survey was 84%, achieved in a study by Marti et al. (1988). Other response rates were below 84%, except for 2 survey studies in smaller populations (Holmich et al. 1988; Lysholm & Wiklander 1987). The response population is the actual study population from which the injury data are obtained. With reference to the injury effect, the response population should not be a selection of the sam-
Sports Medicine 13 (6) 1992
414
pled population. It is important to check for absence of response bias by comparing nonrespondents with respondents. This has been done in most of the selected studies (see asterisks in table II). Until now, epidemiological studies could not find clear differences in incidence rate between respondents and nonrespondents (Koplan et al. 1982; Macera et al. 1989). However, there are other differences between these categories. Younger runners are probably less compliant in responding to questionnaires than older runners. Respondents have consistently been somewhat older than nonrespondents (Bovens et al. 1989; Koplan et al. 1982; Macera et al. 1989). The numerator definition in some studies
showed another remarkable finding. Macera et al. (1989) stated that 47% of the nonrespondents were not eligible to participate in the study. This became clear after these authors tried to match the running activities of nonrespondents with the criteria that these authors accepted for their definition of a runner. Koplan et al. (1982) reported that 60% of the nonrespondents continued to run, compared to 88% of the respondents. These remarkable findings are relevant to the basics of denominator definition. An individual does not stay 'a runner' forever after she or he has met the criteria of the definition of 'a runner'. In conclusion, respondents are older than nonrespondents and are more likely to continue running.
0.65
• Lysholm (sprinters)
0.60 • Lysholm (long distance runners)
0.55 D Macera
• Bovens
III
~
l: l:
0.50
2
"0
l!! :::l
:5"
• Walter D Walter
• 0 Hoeberigs; Macera
D 0 Marti; Walter
0.45
'0
D Holmich (elite)
Ql
U l: Ql
"0
'0 .!: OJ
:::l l: l:
-c
0.40
o
Marti
• Koplan
0.35
0.30
0.24
0= Males 0= Females Both sexes
• Holmich
•=
• Jacobs
• Blair
Race entrants
Listed runners
Supervised runners
Fig. 1. Annual incidence of injured runners by origin of subgroup as a selection from the universe of the running population. For details of references see table II.
The Incidence of Running Injuries
415
2.4 Running Injuries: What is in the Numerator?
