After the pioneering study of Stirling [1], the presence of bacteria in the human non-pyogenic intervertebral discs has drawn increasing attention. The potential significance of this finding may change people’s understanding of intervertebral disc diseases. Then, Albert’s randomized controlled trial demonstrated that their antibiotic treatment was more effective for the patients of Modic type 1 changes and chronic low back pain [2]. Their study supported the existence of bacteria and their significant role in intervertebral disc diseases.

There have been an increasing number of studies on this topic over the recent years. Different methods with more cases were used to prove the existence of the bacteria. However, the results of these studies failed to connect the correlation between existence and proportion of the bacteria in human intervertebral discs. The bacterial culture positive rate ranged from 2% [3] to 53% [1]. The limited sample size of these studies and the divided opinion of the presence of bacteria made the results insufficient to be applied to clinical decision making. Therefore, a systematic review and meta-analysis of the cultural positive rate could be of importance to serve as a reference for further research.

In this study, we aim to find the bacterial culture positive proportion in non-pyogenic human intervertebral disc samples from surgery and to explain the reasons for the different results.


This systematic review was conducted according to the PRISMA statement [4] (see Supplementary Table 1).

Literature search strategy

We searched the PubMed, Cochrane Library, Embase for studies of intervertebral disc infections published between December 31, 2000 and May 31, 2018. The search was limited to studies in English language. We used the keywords of intervertebral disc, Modic changes, disc herniation, Propionibacterium acnes and infections (see supplementary). We also searched and reviewed the reference of retrieved articles for any potentially relevant studies.

Selection criteria

Studies were selected according to the following inclusion criteria: (1) during surgery, intervertebral disc samples were collected from patients without artificial implant in the intervertebral discs nor intervertebral discitis before the surgery and (2) the presence of a bacterial culture to test the existence of the bacteria in these samples. Studies without reporting the bacteria culture results were excluded. Case reports, conference presentations, abstracts, expert opinions, reviews and editorials were also excluded. Only the most complete studies would be included if there were series studies with accumulating samples from the same institutions.

Quality assessment

An 11-item checklist from the Agency for Healthcare Research and Quality (AHRQ) was used to assess the methodological quality of the studies. The item would be given a score of “1” with the answer “Yes” and a score of “0” with the answer “NO” or “UNCLEAR”. Article quality will be defined as: low quality = 0–3; moderate quality = 4–7; and high quality = 8–11.

Data extraction

Data were extracted from texts, figures and tables. Two researchers independently reviewed and extracted data from the articles, specifically, publication year, type of patient, the method of culturing, days of culturing, number of included cases, number of P. acnes-positive cultures, number of total bacteria-positive cultures and number of Modic changes. We collected the data about the positive number of the patients or discs when it is mentioned. A consensus discussion was held when there were disagreements about data extraction.

Statistical analysis

Meta-analysis of proportions was performed for the variable. The analysis was performed by using the meta-package (version 4.9-0) for R (version 3.3.2). The Freeman–Tukey transformation was used [5]. If the sample size was less than 10, the Arcsine transformation would be used as a replacement. Proportions were combined with random effects models, and the odds ratios were combined with fixed effects model. We used the Wilson method to calculate the 95% confidence interval (CI). Heterogeneity was evaluated with I2. P values of less than 0.05 were defined as statistically significant. Publication bias was tested by Egger’s test and Peter’s test and funnel plot. A subgroup analysis was performed on the basis of studies supporting or opposing the existence of bacteria. To investigate potential sources of heterogeneity, we performed meta-regression with bacteria culture days (by comparing studies with culture period more than 14 days with less 14 days), sample size (by comparing studies of more than 100 samples with smaller studies), culture method (by comparing studies with multiple culture media with single media), patient type (by comparing studies with herniation-related patients with others), study quality (by comparing high-quality studies with moderate-quality studies).


Search results

A total of 2882 records were found (see supplementary Figure 1). After the initial screening, we reviewed 25 papers in full. Zhou [6], Yuan [7], Yuan [8] are overlapped with Lin [9]. Rollason [10], Aghazadeh [11] are overlapped with Albert [12], Naghmeh Javanshir [13], respectively. After removal of the accumulated reports and exclusion of ineligible reports, 16 studies published between June 2001 and May 2018 with a total of 2084 reported cases were included in the process for pooling the proportion (Table 1).

Table 1 Study characteristics

Study characteristics

All the included studies reported the existence of P. acnes. Every study reported more than one kind of bacteria except the studies by Javanshir [13] and Lin [9] (See supplementary Table 2). Two studies included more than 300 cases [3, 14]. One study [15] had more than 200 cases, and four studies reported more than 100 cases [9, 13, 16, 17]. All the studies used the bacteria culture, and two studies also detected bacteria DNA in the tissues [12, 15]. Different culture methods and different culture times were used in these studies (Table 2).

Table 2 Study characteristics

Methodological quality assessment

The quality of the studies is presented in Table 1. Scores ranged from 4 to 8. Three included studies were of high quality. The rest of the studies were of moderate quality. Within studies with moderate quality, five studies had a score of 7 which is near to the high quality.

Publication bias

We assessed all 16 included studies with total case number and cultural positive number. Egger’s test did not show risk of publication bias. (t = 0.057, p value = 0.9551). Peter’s test also did not show risk of publication bias. (t = − 1.1756, p value = 0.2594). Funnel plot did not show publication bias. (Supplementary Figure 2).

Pooled proportion of disc infection

From the 16 studies, 2084 cases were included. The pooled proportion of bacteria in the discs was 25.3% (95% confidence interval CI, 15.2–37.0%; I2 = 96.8%; P < 0.0001). Based on the results of the studies supporting or opposing the existence of the bacteria in the intervertebral discs, subgroup analysis was performed. The pooled proportion of the bacteria in the discs from the studies that supported the existence of bacteria in intervertebral discs was 33.7% (95% confidence interval CI, 26.4–41.4%). The pooled proportion of the bacteria in discs from the other group studies was 5.0% (95% confidence interval CI, 0.7–11.9%) (Fig. 1).

Fig. 1
figure 1

Forest plot for pooled proportion of disc-positive culture. Subgroup 1 represents studies that support the existence of bacterial infection. Subgroup 2 represents studies that do not support the existence of bacterial infection. CI: confidence interval

Pooled proportion of P. acnes infection in disc samples

From these 16 articles, the pooled proportion of P. acnes-positive proportion was 15.5% (95% confidence interval CI, 7.7–25.3%, I2 = 96.5%, P < 0.0001). From the studies that support the existence of the bacteria, the pooled proportion of the P. acnes infection in discs was 23.3% (95% confidence interval CI, 16.3–31.2%). For the four studies that disputed the existence of the bacteria, the pooled proportion was 0.3% (95% confidence interval CI, 0.0–1.5%) (Fig. 2).

Fig. 2
figure 2

Forest plot of pooled proportion of P. acnes-positive culture in disc samples. Subgroup 1 represents studies that support the existence of bacterial infection. Subgroup 2 represents studies that do not support the existence of bacterial infection. CI: confidence interval

Pooled proportion of P. acnes infection in the bacteria-positive discs

The most common bacteria are the P. acnes. The pooled proportion of P. acnes in the infected discs was 56.4% (95% confidence interval CI, 42.3–70.0%, I2 = 87.2%, P < 0.0001) (Fig. 3).

Fig. 3
figure 3

Forest plot of pooled proportion of P. acnes-positive culture in the bacteria-positive discs

Comparison of Modic change between positive cultures and negative cultures

Among the included papers that were analysed for the infection proportion of P. acnes, we found Arndt [33], Rigal [3], Drago [34], Chen [30] analysed the presence of Modic changes. The papers of Aghazadeh [11] and Zhou [6] which were excluded due to overlapped data in analysis of infection proportion were also included for assessment of Modic changes (Supplemental Table 3). Overall, the percentages of Modic changes in positive cultures versus non-positive cultures among the six papers were: 77.5 versus 65%, 55.6 versus 38.2%, 100 versus 93.7%, 61.8 versus 86.5%, 85.7 versus 31.3%, 44.4 versus 46.4%, respectively. Meta-analysis found the pooled odds ratio to be 1.27 (95% CI: 0.44–3.64, I2= 68%, p < 0.01) with random effects model; however, heterogeneity was significant (Fig. 4). Thus, we did the sensitivity analysis and found the result of Aghazadeh [11] was different from others which may be the main source of heterogeneity, even though they also found significant relationship between P. acnes-positive samples and presence of MCs in the sample. We re-evaluated meta-analysis without that article and got the reformative result with fixed effects model as 1.97 (95% CI: 1.04–3.73, I2 = 4%, p = 0.39, Supplemental Figure 3).

Fig. 4
figure 4

Forest plot of Modic changes between culture-positive samples and culture-negative samples

Heterogeneity analysis

The pooled proportion of positive rates were significantly different between subgroups that support or dispute the existence of infection. However, meta-regression results showed that the difference of bacteria culture days, multi-culture methods, sample sizes, disc herniation or study quality did not contribute to heterogeneity (Table 3).

Table 3 Meta-regression


Our study found that the pooled proportion of the positive bacterial culture was 25.3% from the studies published between January 2001 and May 2018. Among the bacteria, 56.4% were P. acnes. And the pooled infection of P. acnes in these studies was 15.5%. This demonstrated that P. acnes was the major bacterium in the culture-positive discs. The odds ratio of Modic change in culture-positive samples is 1.27 (95% CI: 0.44–3.64, I2= 68%, p < 0.01). This indicated that the presence of bacteria may contribute to the development of Modic changes; however, the result is not significant.

Propionibacterium acnes (P. acnes) is an aerotolerant, anaerobic, Gram-positive bacterium known as a skin commensal. P. acnes is also an opportunistic pathogen that causes implant-related infection as well as infections of bones and joints [18, 19]. The relatively avascular space of the intervertebral disc is an ideal environment for P. acnes to survive [20]. Animal models also proved that the disc environment was optimal for the bacteria [21] and inoculation of P. acnes can cause degenerative-like changes in the intervertebral discs [9, 21,22,23,24,25]. This may account for the fact that P. acnes is the major bacteria in the positive samples and suggests that P. acnes could be the culprit of some intervertebral disc diseases.

However, the culture-positive proportion varied within studies and some studies did not support the existence of bacteria in the intervertebral discs [3, 21, 26, 27]. The heterogeneity was significant. Our study also demonstrated that the significantly different culture-positive proportions between groups support or oppose the existence of bacteria. The main controversy is whether positive results are related to contamination. P. acnes, a common skin commensal, is considered to be related to orthopaedic surgery contamination [28, 29]. To distinguish contamination from the real infection, some studies used muscle or other related tissue biopsy cultures to serve as a control group [6, 11, 27, 30]. However, most studies did not set control groups to eliminate the possibility of contamination. Capoor’s [15] study also suggested the significant bacterial counts as distinction from the contamination.

On the other hand, P. acnes culture requires a relatively strict anaerobic condition. It grows slower than other anaerobic bacteria [31], and the efficiency of culture methods is important. Suan’s study suggested that more than a 13-day culture period and both anaerobic and aerobic culture methods were needed for P. acnes recovery [32]. Among our included studies, adequate culture time period up to 14 days or more is used in many of the studies [1, 3, 9, 15, 16, 26, 27, 30]. Multiple culture mediums for the tissues were also used to prevent false-negative result [16, 26, 27, 33, 34]. However, according to the meta-regression results, these differences among studies did not contribute to the heterogeneity of the pooled proportions. Attention should still be paid to the culture method. Additionally, Capoor et al. [15] suggested homogenizing the disc samples before culture, concerning the biofilm form of P. acnes, would improve the culture efficiency.

Although the culture result is controversial, many other methods like HE and modified Brown–Brenn staining [7], confocal scanning laser microscopy [14], fluorescence in situ hybridization [35], liquid chromatography–tandem mass spectrometry (LC–MS/MS) [36] and PCR assay [12, 15, 36, 37] have provided the evidence for the presence of bacteria. Studies have also demonstrated the differences between genotypes of P. acnes from intervertebral discs and skin. This is evidence that the culture-positive results are not just the contamination from surgery [10, 38].

Although study has found intervertebral herniation is associated with positive bacterial culture [17], our meta-regression result did not find that patient type, especially herniation, contributed to the heterogeneity. This may be explained by the fact that the presence of bacteria can cause the degenerate changes and its related symptoms not just herniation. However, we cannot ignore that the great variety of included studies and the limited study numbers prevent further analysis of the possible correlation between development of low-virulent infection and its corresponding intervertebral disc diseases and symptoms. Previous systematic reviews had indicated moderate evidence for the relationship between bacteria presence and low back pain with herniation, and Modic change associated with disc herniation [39] or disc degeneration [40]. In the included studies, the highest culture-positive proportions were from studies of severe sciatica patients with confirmed discogenic radiculitis [1], confirmed herniation and its related symptoms and signs [12, 13, 15]. However, with these same kinds of patients, studies also reported cultural results did not support the infection in discs [26, 27]. In order to further investigate the correlation between the development of infection and the type of degeneration, more studies with specified group patients or control groups are needed in the future.

Great interests have been aroused in the field about the relation between Modic change and disc infection. Many researches have been done [2, 3, 12, 23,24,25, 34, 41, 42]. Our meta-analysis result indicated that the presence of bacteria may contribute to the development of Modic changes but this result is not significant, with the result of 1.27 (95% CI: 0.44–3.640). The sensitivity analysis found this result is significant after exclusion of the paper of Aghazadeh [11] which contributes main heterogeneity to this result. The reformative result was 1.97 (95% CI: 1.04–3.73). In Aghazadeh, Javad’s study, their culture period is 7 days which is the lowest. The other studies included in odds ratio meta-analysis all have at least 10-day culture periods. Although their study also found the relationship between Modic change and the presence of P. acnes, we suppose that longer culture period may make this meta-analysis result more significant. And further study should be done to explore this question.

Because of the individual differences, large sample size is still needed to ensure the clinical feature of this non-pyogenic infection in the future. To fulfil this, a sensitive, reliable, practical or noninvasive clinical test method is required. Stirling’s [1] study introduced a serological test to diagnose deep-seated infection by detecting extracellular antigen from bacteria. Magnitsky’s [43] study introduced a noninvasive method to detect the existence of P. acnes in intervertebral discs in an animal model by NMR spectroscopy. These types of technologies may help in future clinical study.

There are some limitations to the study. First, the heterogeneity between studies is significant. The variety in methodology may contribute to it but the limited number and the variety of included studies prevent further analysis. Secondly, most of the studies were in a moderate quality and lack of culture control group makes the positive results less convincing. Thus, the results may have moderate contribution to the literature due to the heterogeneity of the articles.

Although an increasing number of studies and evidences demonstrate the existence of the bacteria in the intervertebral discs, the proportion of bacteria in discs still varied from studies. Our study performed a systematic review of the previous study and provided a positive cultural proportion for reference. In order to confirm the results, more cases should be done with efficient method of culture techniques, control groups, specified patients’ types, advanced test technologies and so on.