Protocol and registration

The protocol for this systematic review was prospectively registered in the international PROSPERO database. Protocol details can be accessed via http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42017053258. A protocol revision was performed in July 2017 from midportion AT as site of injury to AT in general (insertional and midportion AT combined). This was done as we observed during data extraction that a substantial number of publications did not specify the specific location.

Eligibility criteria

Publications were eligible for inclusion when there was: (1) a potential risk factor investigated in relation to AT and (2) a diagnosis of AT based on clinical findings (local pain and impaired load-bearing capacity). We restricted our selection to prospective and retrospective cohort studies written in English. A determinant was considered to be a potential clinical risk factor if no extensive examination (eg, biopsies) had been performed. Publications were excluded if there was: (A) no adequate control group (eg, contralateral Achilles tendon), (B) a preclinical study design or (C) an imaging study design (eg, potential risk factors derived from MRI or ultrasound examinations). Imaging studies were excluded, since these are regularly not directly available in the sports physician consultation room. For an overview of imaging, we refer to a recent systematic review.9

We also aimed to identify potential novel risk factors as secondary outcome measure by including cross-sectional studies. While the level of evidence from cross-sectional studies is lower than that from cohort studies, they might identify interesting factors to explore in future research.

Literature search strategy and information sources

We conducted a sensitive search strategy for multiple databases with the assistance of a medical librarian (W M Bramer). The following databases were searched up to 12 February 2018: Embase, MEDLINE Ovid, Web of Science, Cochrane Library and Google Scholar. The search strategy is shown in online supplementary appendix 1.

Study selection and data extraction

Titles and abstracts were screened by two independent reviewers (ACvdV and RJdV) to identify eligible publications. Disagreements were solved by consensus, with the involvement of a third review author (EHGO) if necessary. Data extraction was performed by one author (ACvdV), and a data check was performed by a second author (SJB) for 100% of the primary outcomes and for 20% of the other data. This has been shown to be a methodologically sound procedure.10 The potential risk factors were extracted and grouped into patient characteristics (modifiable and non-modifiable), biomechanical factors, pre-existing diseases, medication and training factors. AT subgroup analysis results are presented in case subgroup analyses were performed in studies describing associations for multiple injuries. If multiple populations with AT were assessed in a single publication, only combined results are presented.

Risk of bias assessment

Two reviewers (ACvdV and SJB) independently assessed the methodological quality of all included prospective (level of evidence II) and retrospective cohort studies (level of evidence III). No risk of bias assessment was performed for cross-sectional studies (level of evidence IV), since these studies are considered to be of high risk of bias for the purpose of this review.

To assess risk of bias, we used a standardised set of criteria based on modified questions of existing quality assessment tools (table 1).11–13 This tool has previously been used in a systematic review on risk factors for Achilles tendon rupture.14 If a criterion was met, one point was given. No points were given if the criterion was not met or when it was unclear if the specific criterion was met. A maximum score of 10 points could be obtained. Publications were considered to be of low risk of bias if: (1) a total score of 6 points or more was given and (2) 1 point was given to criteria 6, 7, 8 and 10.14

Data synthesis

A subgroup analysis was initially planned for insertional and midportion AT; however, we revised the PROSPERO protocol (revision date 20 July 2017) because a substantial number of publications did not specify the AT location. Homogeneity of the data was evaluated, and if data could not be pooled because of heterogeneity, a best evidence synthesis based on the study of van Tulder et al was carried out for each potential risk factor.15

1. Strong evidence: ≥2 studies with high quality and generally consistent findings in all studies (≥75% of the studies reported consistent findings).

2. Moderate evidence: one high-quality study and ≥2 low-quality studies and generally consistent results (≥75% of the studies reported consistent findings).

3. Limited evidence: generally consistent findings in ≥1 low-quality study (≥75% of the studies reported consistent findings).

4. Conflicting evidence: <75% of the studies reporting consistent findings.

5. No evidence: no studies could be found.

Study selection

We identified 5111 potentially relevant publications, and after removing duplicates, 3225 remained. After screening title and abstract, we assessed 109 publications in full text. Fifty-four publications were excluded for different reasons after full-text evaluation, as shown in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow chart (figure 1). We included the remaining 55 publications for analysis.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

PRISMA 2009 flow diagram of study selection process. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses. 

Characteristics of the included publications

We included eight prospective cohort studies16–23 and two retrospective cohort studies.24 25 Additionally, 45 cross-sectional studies were included.5 26–69 The characteristics and main findings of the included studies are summarised in table 2 for the cohort studies and online supplementary appendix 2 for the cross-sectional studies.

Of the 10 cohort studies, five studies included only participants with midportion AT and five studies did not specify the AT location. Sample sizes of the included cohort studies ranged from 69 to 80 106 participants (median 285) with the number of AT cases ranging from 5 to 450 (median 18). Mean age ranged from 18 years old to 59 years old (median 21). This relatively young median age was caused by the profession of most of the populations investigated (military recruits in five studies, students in one study). There was a greater proportion of male participants in seven cohort studies, compared with two studies in which there was a greater proportion of female participants (median percentage males 78%). Data regarding the number of participants, mean age or sex was incomplete in two studies. The follow-up period for the prospective cohort studies ranged from 6 weeks to 2 years (median 39 weeks).

Of the 45 cross-sectional studies, 15 studies included only participants with midportion AT, one study included patients with insertional AT and 29 studies did not specify the AT location. Sample sizes ranged from 20 to 57 725 participants (median 201).

Risk of bias assessment

All 10 cohort studies were considered to be of high risk of bias according to the predefined criteria (tables 1 and 3). Seven cohort studies scored six points or higher; however, they were lacking clinical examination as valid outcome (two studies), a valid and reliable determinant measure (four studies), an unbiased assessment of the study outcome (one study) and/or adequate statistical analyses (three studies). As a result, at best, a limited evidence for the association between the potential risk factor and tendinopathy could be detected. Results of the best evidence synthesis are presented in table 4.

Risk factors

Summary of main findings

This is the first high-quality systematic review of clinical risk factors for Achilles tendinopathy. We identified 10 cohort studies, all of which had a high risk of bias and 45 cross-sectional studies.

There is limited evidence for the following nine risk factors: (1) prior lower limb tendinopathy or fracture, (2) use of ofloxacin antibiotics, (3) increased time between heart transplantation and initiation of quinolone treatment for infectious disease, (4) moderate alcohol use, (5) training during cold weather, (6) decreased isokinetic plantar flexor strength, (7) abnormal gait pattern with decreased forward progression of propulsion, (8) more lateral foot-roll over at the forefoot flat phase and (9) a creatinine clearance of <60 mL/min in heart transplant patients.16 17 19 22 24 25

Although other potential risk factors such as body weight or BMI, static foot posture measurements and physical activity level are often said to be risk factors in clinical practice, there is currently no scientific evidence that they are associated with AT.16–19 21–24 70

Clinical implications

Our systematic review indicates that for AT prevention and treatment, the advice to patients might include: (1) to reduce the use of alcohol to less than 7 units per week for men and less than 4 units for women, (2) to avoid the use of ofloxacin if alternatives are available and (3) to improve plantar flexor strength by performing strengthening exercises of the calf muscles.16 17 25

Whether these interventions will be effective is unknown. For example, calf muscle exercises seem a plausible preventive intervention as decreased plantar flexor strength is a risk factor. However, eccentric calf muscle exercises did not decrease AT incidence in soccer players in a randomised trial.71 Further research is needed before we can state whether the logical interventions work or not.

Abnormal gait pattern with decreased forward progression of the propulsion and a more lateral foot roll-over at the forefoot flat phase were found to be risk factors for AT in novice runners. Van Ginckel et al stated that more research is needed to confirm these findings, since this gait pattern with a decreased forward progression might be commonly used in well-trained athletes to improve gait economy.19 72 Since the gait pattern was determined barefoot, it is also not known whether these findings can be extrapolated to the running population, as running shoes might alter running gait.19 Biomechanical characteristics in AT are discussed in more detail in a recent systematic review.73

Previous research showed the relationship between BMI, body weight or waist circumference and tendon pathology.74 The hypothesis of this relationship is primarily based on the fact that the absolute tendon load is increased and that increased cytokine levels (Prostaglandin E2, tumor necrosis factor-α and Leukotriene B4) cause low-grade inflammation in obese individuals.75 76 In our systematic review, we were not able to find an association between being overweight and AT.16–19 22 23 It should be noted that more than half of the cohort studies that investigated BMI as a risk factor were in adolescent populations, in which being overweight is less common.77 Arnoczky and colleagues hypothesised based on an animal model that being underweight is associated with AT, as a consequence of a catabolic state causing a decreased collagen production.78 This could lead to a U-shaped relationship for the BMI and AT, making it less likely that an association be found.79 More cohort studies are needed in heterogeneous populations.

Another striking finding is the limited evidence for the absence of an association between physical activity level and AT. Inconclusive results regarding physical activity level were previously also demonstrated in patellar tendinopathy.80 In the majority of the scientific literature, tendinopathy is described as ‘overuse injury’. It could be that ‘overuse’ or ‘physical activity level’ is not measured accurately enough to detect associations. It could also be hypothesised that a sudden change in load is more important than the absolute load that is currently being measured in studies. To date, there is, however, no convincing evidence that AT is a result of overuse.

Moderate alcohol use and increased time between heart transplantation and initiation of quinolone treatment for infectious disease were potential risk factors for AT.17 24 It is hard to hypothesise why these determinants are risk factors for AT. They might be a confounding factor, with lifestyle factors that influence AT risk. Another reason might be that these findings have been detected by chance. When a high number of analyses are performed, the chance of statistical significance findings increases. Adequate statistical methods can prevent possible coincidental findings. Another potential risk factor that is lacking a clear explanation is training during cold weather.22 The direct cause-effect relationship is not known, as the results might be influenced by temperature or type of surface. This particular study did not report information on the temperature or training surface during the training period. These results are therefore difficult to extrapolate.

Research implications

Our review showed that the majority of potential risk factors have only been investigated in cohort studies with a low number of cases (median 18 cases). Professor Roald Bahr and colleagues demonstrated that 20–50 cases are needed to detect moderate to strong associations and even 200 cases to detect small to moderate associations.81 Therefore, most studies in this review are likely to be underpowered to detect associations. Sample sizes in future studies should therefore be considered carefully. Future studies should also distinguish insertional from midportion AT, since these are two separate entities. It has been suggested that compression forces due to the bony prominence of the calcaneus play a role in the development of insertional AT, while this does not occur in midportion AT.4 Combining these entities, as occurred in most studies, is not ideal.

There are several interesting determinants found in the cross-sectional studies for future research. Use of oral contraceptives and hormone replacement therapy were more common in patients suffering from AT.38 Only one research group investigated these factors; therefore, more research is needed to confirm these findings. Regarding lipid profile, Dr Jamie Gaida and colleagues reported that triglyceride level, triglyceride/high-density lipoprotein cholesterol ratio and apolipoprotein B were elevated in patients with AT.34 Hypertension prevalence was found to be increased in females in the study by Holmes and Lin, suggesting a possible relationship between blood flow circulation to the Achilles tendon and AT.38 However, all of these factors should be studied in a longitudinal study design since it is not clear whether there is a cause–effect relationship.

Genetic profiling is also a major topic in AT. Since 2002, 16 cross-sectional studies have evaluated the presence of genetic variations in AT.26 29 33 37 39 44–49 51–55 These genetic variations are linked to collagen structure, tendon or matrix homeostasis, apoptosis or inflammation pathways.8 This type of research provides more information regarding the biological pathways of the disorder. Future therapy strategies could focus on targeting these pathways.

Strengths and limitations

The strength of this systematic review is that we performed this structured analysis according to the PRISMA guidelines.82 Consequently, we were able to include 10 cohort studies, whereas a different systematic review on this topic only included one cohort study.8 That recent systematic review provided an excellent overview of all the literature considering genetic variants in AT, but important publications considering non-genetic risk factors in AT were missing. By including these cohort studies, our methods provide evidence that can be used directly in a clinical setting. We were also able to present an overview of topics on which future research should focus. Despite our robust research design, there are also methodological limitations. First, we only selected publications written in the English language. Second, we were not able to pool data because of the heterogeneity. The strength of the associations could therefore only be evaluated with a best evidence synthesis and not with meta-analysis. Third, we chose to use a standardised set of criteria based on modified questions of existing quality assessment tools. This was recommended by Hayden et al, since no validated tools are available for systematic reviews on risk factors.83 Our tool used strict criteria, primarily considering clinical examination as valid outcome measure (criterion 6). Four cohort studies did not meet this criterion of which three of these studies used International Statistical Classification of Diseases and Related Health Problems (ICD) codes/search terms.17 24 25 The fourth study examined only runners with serious symptoms.20 These approaches to case-finding creates a bias. Furthermore, if this criterion would not be taken into account, the studies would still be considered to be of high risk of bias based on the other criteria. Fourth, the median age of the included cohort studies was 21 years due to the profession of most of the populations investigated. This is a relatively young age, since the mean age to develop AT is 30–60 years, and AT is therefore expected to be less common in these studies.69