Number of articles
For the first two research questions, 496 records were identified: 277 in PubMed, 170 in Embase and 49 in SportDiscus. Of these, 40 full texts were selected, 13 of which were duplicate studies resulting in 27 relevant full texts that were assessed for eligibility. Ultimately eight articles were included in the review, four of them relating to question 1 [24–27] and four to question 2 [28–31]. A hand search resulted in the identification of three potential texts, none of which were suitable for inclusion.
For the third research question, 270 records were identified: 156 in PubMed, 62 in Embase and 52 in SportDiscus. Eleven relevant full texts were selected, three of which were duplicate studies, resulting in eight full texts that were assessed for eligibility. Three of these were suitable for inclusion [28, 30, 32]. A hand search resulted in the identification of four potential texts, none of which were included. The reasons why these articles were excluded are listed in Fig. 1.
Two articles [28, 30] included information pertinent to two of the three objectives. In the tables and results section, articles were listed in alphabetic order.
A. Is muscular strength in trunk extension associated with BP?
Description of studies or BMS
Four articles were included to answer the question of whether BMS in trunk extension is associated with BP. The first article, Balagué et al. [24], presents a cross-sectional study in which 117 children aged 11 to 15 years (response rate: 97 %) were included. Its purpose was “to evaluate the relationship between the dynamic strength profile of the trunk, anthropometric parameters, BP and frequency of sport activities performed”. BMS was evaluated using an isokinetic Cybex II dynamometer. Information on BP was obtained through interview and defined in relation to location, cumulative life prevalence and point prevalence. The association between BMS and presence of past history of BP was studied controlling for self-reported frequency of sport. Univariate, multivariate and correlation analyses were used to determine the association between BMS and BP.
The second article, Balagué et al. [25], reports the results from a cross-sectional and a prospective study, in which 95 children aged 13 to 14 years remained at follow-up (response rate: not reported). The objective was to examine if trunk performance capacity has an association with LBP in adolescent boys. The trunk muscle performance was evaluated with standard dynamometer testing protocols. The presence of LBP was determined using a brief semi-structured interview with questions about medical attention and the time of the last episode. The difference in BMS was studied for the groups with and without LBP. The unpaired t-test was used to compare BMS in those with and without BP.
The third article, Merati et al. [26], presents a cross-sectional study in which 144 12-year olds were included (response rate: not reported). The goal of this study was to assess if a deficit in trunk muscular strength plays a role in BP occurrence in pre-pubertal subjects. BMS was measured with a modular-components isokinetic dynamometer. A questionnaire was used to determine the presence of BP. The student t-test was used to compare BMS in those with and without BP.
The fourth article, Newcomer and Sinaki [27], presents a prospective study with a four-year follow-up in which 96 study subjects, aged from 10 to 19 years, remained in the final study group (response rate: 39 %). The main purpose was to determine the occurrence of LBP and its relationship to back strength and physical activity in children. Back strength was tested at baseline by an iso-dynamometer. LBP at follow-up was determined in an interview, based on a list of five questions about LBP-ever, the age at the first episode, one-year prevalence, the consequences on school and sport activities and medical attention. To evaluate this association, logistic regression was used.
Quality assessment of articles on BMS
The quality scores in the four reviewed articles were 56 %, 56 %, 78 % and 78 % (Table 1).
Studied factor: back muscle strength
All the articles assessed BMS with a dynamometer. In one study [25], reference was made to a previous study having shown the measurement to be reliable whereas in two of the studies, reliability was tested and shown to be acceptable in one [24], but results were unreported in the other [26]. In the fourth study [27], the dynamometer was calibrated and it was reported that the method had been previously shown to be reliable and valid. In general, these data can therefore probably be trusted.
Outcome measure: back pain
BP was clearly defined in all the articles as well as the description of the BP assessment. However, only one article reported a recall period of one month or less [24], which was considered suitable in young people. The other articles reported in their results section a history of LBP [25], a recall period of six months [26], and recall periods of one year and a lifetime [27].
Data collection
In two articles [24, 25], the data collection for BP was made through semi-structured interview but it was not clear if the person who made the interview and the person in charge of strength measurement were the same. In the other two [26, 27], questionnaires were used, thus ensuring separate data collection of these two variables, necessary to prevent reporting bias.
Study sample
All studies recruited at least some of their study subjects from schools, one having to resort to additional assistance from medical practitioners for recruitment [24]. In only one of the studies [26],participants were reported to have been randomly selected. Whether study participants were representative of the general population is therefore doubtful.
Results for research question 1
None of the four relevant articles demonstrated an association between BMS in extension and BP. Therefore no meta-analysis was performed for this research question.
B. Is muscular endurance in trunk extension associated with BP?
Description of studies on BME
The first article exploring the association between BME in trunk extension and LBP reports on a cross-sectional study written by Andersen et al. [28], in which 9413 17-year olds were included (response rate: 41 %). The aim was “to examine the association between physical fitness and self-reported BP in adolescents”. The BME was assessed with the Biering-Sorensen test. BP was self-reported and focused on the presence of pain in the past month, prior experience of BP and the location of the pain. Logistic regression was used to assess the association between BME and BP adjusting for sex, height and smoking.
The second article, Bernard et al. [29], describes a retrospective study in which 327 individuals aged 10 to 18 were included (response rate: 50 %). The main aim was “to compare muscle endurance of back flexors and extensors between a control group of 276 teenagers and a group of 51 teenagers from a pediatric unit, who suffered from chronic LBP”. The BME was assessed with the Biering-Sorensen test. LBP information was assessed with a visual analogue scale in a specific questionnaire for the chronic LBP group. The relevant analysis was performed by comparing the BME in the clinical group to the control group. Nevertheless, some of the participants in the control group also reported some LBP (n = 47 according to the method section and n = 48 according to the results section). How this information was obtained was not explained. The association between BME and BP was tested using linear regression.
The third article, Johnson et al. [30], is a cross-sectional study including 625 youngsters aged 11-19 (response rate: not reported). The aim was “to establish reference data and pattern of back extensor strength in school-aged Nigerian adolescents”. The BME was assessed with the Biering-Sorensen test. The history of LBP and present LBP was assessed by questionnaire. The difference in BME was tested for those with or without a history of LBP using a t-test. The same was done for present LBP.
The fourth relevant article, Perry et al. [31], also describes a cross-sectional study in which 1608 adolescents, all aged 14, were included (response rate: 69 %). The aim of this study was to determine if physical fitness is related to increased risk of BP. The BME was assessed with the Biering-Sorensen test. Information on BP was obtained by a questionnaire including lifetime prevalence of pain, pain in the last month, chronic pain and also pain diagnosis. Results were reported separately for boys and girls, in which the lower 25 % and the higher 25 % were compared to the middle 50 %. The association between BME and BP was tested with multivariate logistic regression.
Quality assessment of articles on BME
The methodological quality scores were 67 %, 78 %, 78 %, and 89 % (Table 2).
Studied factor: back muscle endurance
All the authors assessed BME with the Biering-Sorensen test, which has been reported to be a reliable and valid tool [13, 33].
Outcome measure: back pain
The definition of BP and method of assessment were always clearly defined. In all articles except one [29], the recall period was appropriate for at least one variable concerning BP. However, the aim of that article was in fact to compare a clinically affected group of children with chronic LBP against a group of “normal” children. The recall period, therefore, did not appear to be of importance in this case.
Data collection
The data on BME and BP were collected independently (blindly) by two different persons or at least by using a questionnaire in all studies.
Study sample
Although attempts were made to access children from the general population, in three of the studies representativeness was not explicitly addressed [29–31]. In the fourth study [28], although the target population was not representative of the general population, their sample was compared to another representative group and no difference was found in the physical fitness test between these two groups, meaning that their study sample had external validity, at least on this key variable.
Results for research question 2
In all four articles, an association was found between BME and BP. In three of these [28–30], it was reported that those with BP had a weaker BME compared to those without BP (Table 5). In the fourth study [30], many associations were tested. In this study, only one (diagnosed BP) of four outcome variables (BP ever, one month prevalence, chronic back pain (CBP), diagnosed BP) was statistically significant for girls and not for boys. In the text, multivariate analysis is reported to have resulted in an increased likelihood of diagnosed BP in those with reduced BME as compared to the middle group. On the other hand, those with the greater BME, when compared to the middle group, were also found to be more likely to report diagnosed BP, i.e. indicating a U-curve for diagnosed BP.
On this topic, in all articles, the data allowed us to perform a meta-analysis (Fig. 2). A negative association was found between the BME and BP (OR = 0.75, 95 % CI 0.58-0.98). The I2 was 66.1 % indicating a high heterogeneity between the studies. This can be explained by the fact that some articles divided their sample according the sex of the participant and by the differences in the definition of back pain.
C. Is aerobic capacity associated with BP?
Description of studies of AC
The first of the three relevant articles for the third research question, relating to AC in general and BP, reports the results from a cross-sectional study performed by Andersen et al. [28], in which 9413 17-year olds were included (response rate: 41 %). The aim was “to examine the association between physical fitness and self reported BP in adolescents”. AC was assessed by VO2max measured with the help of a cycle ergometer. BP was self-reported and defined as the presence of pain in the last month, prior experience of BP and the location of the pain. Logistic regression was used to assess the association between BME and BP adjusting for sex, height and smoking.
The second article, Cardon et al. [32], describes a cross-sectional study in which 749 children aged 8 to 12 years were included (response rate: not reported). The aim of this study was “to examine whether physical fitness, physical activity, and psychosocial determinants of physical activity are associated with reports of back or neck pain”. AC was assessed by a 20-m endurance shuttle run protocol. BP was evaluated with a questionnaire that defined BP based on the location of pain in the past week. It also included severity and frequency. Analysis of variance was used with age as a covariate to determine the association between AC and BP.
The third article for this research question, Perry et al. [30], also describes a cross-sectional study in which 1608 14-year olds were included (response rate: 69 %). The aim of this study was to determine if physical fitness is related to increased risk of BP. The AC was assessed by sub-maximal cycle ergometry, using a specific protocol (PWC 170). Information on BP was obtained with a questionnaire that included lifetime prevalence of pain, pain in the last month, chronic pain and also pain diagnosis. Results were reported separately for boys and girls, in which the lower 25 % and the higher 25 % were compared to the middle 50 %. The association between AC and BP was tested with multivariate logistic regression.
Quality assessment of articles on AC
The three articles had quality scores of 67 %, 89 %, and 89 % (Table 3).
Studied factor: aerobic capacity
All the authors clearly defined the tools used to evaluate AC and all but one [28] provided references about validity and reliability of their evaluation test.
Outcome measure: back pain
BP (and the way it was assessed) was clearly defined in all the articles. The recall periods were one month or less in all studies [28, 32], although Perry et al. [30] also used longer recall periods.
Data collection
In all the studies, the data for BP were collected with a questionnaire and therefore the independent and dependent variables were collected separately.
Study sample
Attempts were made to access children from the general population in all three studies. As previously explained, in one of the studies [28] the results on the physical fitness tests were similar to those in a representative sample from another study. For the other two [30, 32] the final representativeness is unknown.
Results for research question 3
In all three articles, at least one association was reported between AC and BP. In two of the studies, results were reported separately for girls and boys with positive findings only for the boys [30, 32]. However, these two studies reported conflicting results as one study found a positive association [30] whilst the association was negative in the other [32]. In the third study [28], the association between AC and BP disappeared after adjustment for BME. Hence, the AC and BP may well be associated but it is unclear how, with the possibility that AC is but a proxy for BME. Meta-analysis was not performed for this research question because it was not possible to obtain the confidence intervals of all the odds ratios in the included articles.