Introduction

Intervertebral disc (IVD) degeneration (IVDD), a widely recognised musculoskeletal disease with remarkable socioeconomic burdens worldwide, is one of the leading causes of low back pain [1]. IVDD is caused by numerous factors, including ageing; heredity; mechanical stress related factors such as obesity, multifidus and psoas major deterioration (sarcopenia), vertebral bone mineral density (osteoporosis); smoking; hypertension; diabetes mellitus; metabolic syndrome; anaemia; oestrogen; inflammation; and oxidative stress [1,2,3,4,5,6,7,8].

An IVD consists of three structures: a central jelly-like nucleus pulposus (NP), an outer rigid annulus fibrosus (AF), and upper and lower cartilaginous endplates. NPs and AFs synthesise a water-rich extracellular matrix (ECM) that significantly contributes to normal trunk functioning [1]. Elastic NPs consist primarily of water (70–90%), NP cells, proteoglycans, and type-II collagen, which has an essential part in weight bearing and relieving pressure loads associated with spinal column movement [6]. A tight AF seals the NPs and distributes pressure and forces on IVD, preventing herniation of NP [6, 7]. Due to various external and internal stimuli, dysfunction of NPs and AFs and subsequent excessive catabolism of ECM can lead to degeneration and calcification of the cartilage endplates, resulting in IVDD [6]. Apart from ageing and mechanical overload, oxidative stress and increased secretion of inflammatory cytokines may induce the progressive structural destruction of IVD and accelerate IVDD [7, 8].

Periostin, a type of ECM, has been reported to be closely linked to mechanical stress, inflammation, and ageing as well as to the development and progression of IVDD [9, 10]. Periostin binds to ECM molecules in IVDs to initiate, maintain, and repair IVD, and excessive ECM catabolism is involved in the initiation and progression of IVDD [9, 10]. Periostin is expressed predominantly in fibrous collagen-rich connective tissue, which is constantly under mechanical stress. Periostin-positive cells are most abundant in the lateral AF of the disc and least abundant in NP [10]. Histopathological examination revealed significantly elevated periostin expression in IVDs from IVDD patients and increased fibrosis, including periostin in IVDs, along with structural destruction and fragmentation in human IVDs compared to non-denervated IVDs [10]. Regarding inflammation, periostin is implicated in the aetiology and progression of many chronic inflammatory diseases through its association with inflammatory cytokines [9,10,11]. Elevated levels of inflammatory cytokines including tumour necrosis factor (TNF)-α, interleukin (IL)-6, and IL-17 released by NP cells induce IVDD [8]. Metabolic syndrome, a risk factor for IVDD, induces persistent low-grade systemic inflammation that causes biochemical stress in the systemic tissues [12]. A correlation between periostin and metabolic syndrome has been demonstrated through inflammation [13].

Periostin may influence IVDD via mechanical stress and inflammation; therefore, serum periostin levels may be elevated in patients with severe IVDD. However, to the best of our knowledge, there are no studies on serum periostin levels in patients with IVDD. We evaluated the correlation between IVDD severity and serum periostin concentration and analysed the potential relationship between IVDD and clinical and demographic factors.

Methods

Study design and subjects

This retrospective cohort study was authorised by an Institutional Review Board and conformed to the ethical guidelines of the 1975 Declaration of Helsinki. The study population included 198 consecutive patients with degenerative lumbar diseases (lumbar disc herniation and lumbar spinal canal stenosis) who underwent spinal surgery in our department between January 2020 and December 2022. Because ageing may potentially affect serum periostin levels, this study included both lumbar disc herniation, prevalent among young people, and lumbar spinal canal stenosis, common in middle-aged and older populations.

Magnetic resonance imaging (MRI) and computed tomography (CT) were used to identify patients with lumbar disc herniation and lumbar spinal canal stenosis at levels corresponding to their symptoms [5, 14]. After excluding patients lacking data (n = 2), 196 patients were finally included.

IVDD

The Pfirrmann grade (1–5) was used to evaluate IVDD severity by MRI [5, 14]. To assess the severity of IVDD, the Pfirrmann grades of all lumbar discs were summed to calculate a cumulative grade. Two experienced spine surgeons, who reviewed imaging findings and were blinded to the clinical and analytical data, determined the degree of degeneration. Two experienced spine surgeons, who reviewed the imaging findings and were blinded to the clinical and analytical data, determined the degree of degeneration. If the assessments did not agree, a third spine surgeon established consensus.

Periostin

Blood samples were obtained on admission to undergo spine surgery and immediately sent for testing. Patient serum samples were maintained at −80 °C prior to measuring serum periostin levels. Serum periostin levels were quantified using ELISA kits (Shino Test, Tokyo, Japan) for human periostin. The values were calculated by subtracting the absorbance at 550 nm (secondary wavelength) from the absorbance at 450 nm (primary wavelength), measured using an EnVision multilabel plate reader (Perkin Elmer, Massachusetts, USA).

Preoperative clinical characteristics

Collected demographic data, including age, sex, body mass index (BMI), psoas muscle index, smoking habits, comorbidities (Charlson comorbidity index [CCI], anaemia, and osteoporosis), and spinal diseases (lumbar disc herniation and lumbar canal stenosis), were collected. These factors were selected because IVDD is associated with numerous factors, including ageing, obesity, psoas major deterioration (sarcopenia), smoking, and comorbidities such as hypertension [1,2,3,4,5,6,7,8]; these factors could also potentially serve as confounding factors.

The psoas muscle index (cm2/m2) was calculated by dividing the area of the cross section of the psoas major muscle at L3 by height [15]. This measurement has been documented to provide a strong correlation with total body muscle mass, making it a suitable surrogate for assessing sarcopenia.

CCI was calculated to estimate the burden of comorbidities, with higher scores indicating greater comorbidity [16]. The following classifications were used: 1 point (congestive heart failure, myocardial infarction, peripheral artery disease, cerebrovascular disease, dementia, chronic lung disease, collagen disease, peptic ulcer, mild liver disease, diabetes without organ damage), 2 points (hemiplegia, moderate-to-severe renal impairment, localised solid cancer, diabetes mellitus with organ damage, leukaemia, lymphoma), 3 points (moderate-to-severe liver impairment), and 6 points (metastatic solid tumour, AIDS). In addition to these comorbidities, anaemia and osteoporosis were assessed. Anaemia was defined by World Health Organization criteria as haemoglobin levels < 12 g/dL for female participants and 13 g/dL for male participants [5]. Osteoporosis was defined using CT values in this study, as not every patient underwent bone mineral density examination. Recently, lumbar vertebral body CT values have been used to diagnose osteoporosis (L1 CT value ≤ 90 Hounsfield Unit [HU]) [17]. The HU value for the oval region of interest was the central anterior third of the L1 vertebral body [17]. Region of interest should include as much trabecular bone as possible, excluding heterogeneous areas such as cortical bone and posterior venous plexus. For patients with moderate-to-severe vertebral fractures, HU values were assessed at L2 instead of L1.

Statistical analysis

The Shapiro–Wilk test was performed to assess the distribution normality of the quantitative data. In the comparison analyses, quantitative data were compared using Student’s t-test or Mann–Whitney’s U test, and qualitative data were compared using the Chi-square test. To examine the relationship between the Pfirrmann grade (cumulative: L1/2–L5/S1) and periostin, we conducted a logistic regression analysis to calculate odds ratio. The regression models were constructed in multiple steps. First, the univariate models were created. Second, we created multivariate models adjusted for age, sex, and BMI. Next, we created multivariable models adjusted for factors with p < 0.10 in the comparison analyses, in addition to age, sex, and BMI. These factors accounted for potential confounding, recognising that apparent associations may exist even when there are no significant differences in comparison analysis. To evaluate the relationship between cumulative Pfirrmann grade (L1/2-L5/S1) and periostin, Spearman’s correlation coefficient (ρ) was evaluated for variables without normal distribution. The correlation coefficients were categorised as follows: negligible (0.00–0.10), weak (0.10–0.39), moderate (0.40–0.69), strong (0.70–0.89), and very strong (0.90–1.00) [18]. The significance level (p-value) was adjusted using Bonferroni correction as needed. Statistical analyses were conducted using JMP® Pro 16 (SAS Institute, Cary, North Carolina, USA).

Results

The median age of the 196 patients (female: 82) was 71.0 years (41.8%). The distribution of periostin and the cumulative Pfirrmann grade (L1/2–L5/S1) are shown in Fig. 1a and b. Periostin was not normally distributed (Shapiro–Wilk test, p = 0.010), and its median value (interquartile range [IQR]) was 31.0 (26.0–37.0). The cumulative Pfirrmann grade (L1/2–L5/S1) was not normally distributed (Shapiro–Wilk test, p = 0.002; median (IQR), 17.0 [14.0–18.0]). We categorised periostin based on the median value of 31 (Fig. 1a). The cutoff value for cumulative Pfirrmann grade (L1/2–L5/S1), which discriminates the severity of IVDD from the distribution in Fig. 1b, was 17. This was similar to that reported by Guo et al. [6], who divided the patients into the following two groups: cumulative Pfirrmann grade (L1/2–L5/S1) ≤ 17 (n = 129) and >17 (n = 67). Factors with p < 0.10 in the comparison analyses between patients with the cumulative Pfirrmann grade (L1/2–L5/S1) of ≤17 and >17 were age (median, 70.0 years vs. 73.0 years, p = 0.001), CCI score (median, 4.0 vs. 5,0, p = 0.001), anaemia (31/129 patients [24.0%] vs. 27/67 patients [40.3%], p = 0.018), and lumbar disc herniation (49/129 patients [38.0%] vs. 9/67 patients [13.4%], p < 0.001). Therefore, we developed five regression models: (1) unadjusted; (2) adjusted for age, sex, and BMI; (3) adjusted for age, sex, BMI, and CCI score; (4) adjusted for age, sex, BMI, CCI score, and anaemia; and (5) adjusted for age, sex, BMI, CCI score, anaemia, and spinal disease. In all regression models, serum periostin (>31 ng/mL) was associated with the cumulative Pfirrmann grade (Table 1). In the adjusted multivariate regression models, age, sex, BMI, CCI score, anaemia, and spinal disease (lumbar disc herniation and lumbar spinal canal stenosis) did not affect the cumulative Pfirrmann grade (L1/2–L5/S1) (Table 2).

Fig. 1
figure 1

Distribution of periostin (a) and the cumulative Pfirrmann grade L1/2–L5/S1 (b)

Full size image
Table 1 Participants’ characteristics of this cohort and their comparisons between the cumulative Pfirrmann grade (total from L1/2 to L5/S1) ≤ 17 and  >17
Full size table
Table 2 Relationship between the cumulative Pfirrmann grade from L1/2 to L5/S1 (>17) and periostin (≤31 ng/mL and >31 ng/mL)
Full size table

The correlations between periostin levels and Pfirrmann grades are shown in Table 3. Periostin showed moderate correlations with the Pfirrmann grade at levels L1/2 (ρ = 0.550, p < 0.001), L2/3 (ρ = 0.527, p < 0.001), L3/4 (ρ = 0.458, p < 0.001), L4/5 (ρ = 0.460, p < 0.001), L5/S1 (ρ = 0.483, p < 0.001), and cumulatively (ρ = 0.690, p < 0.001).

Table 3 Spearman’s correlation coefficient (ρ) of periostin with the Pfirrmann grade
Full size table

Discussion

The results of our study suggest that (1) higher serum periostin level is correlated with a higher cumulative Pfirrmann score (L1/2–L5/S1), indicating a higher severity of IVDD; and (2) serum periostin (>31 ng/mL) is associated with the cumulative Pfirrmann score but not with age, sex, BMI, CCI score, anaemia, or spinal disease in adjusted multivariate regression models.

In this study, serum periostin was identified as an independent risk factor for IVDD according to multivariate analysis in a model. This model reduces confounding because serum periostin level is elevated by various factors. Serum periostin levels exceed 100 ng/mL during childhood and adolescence and have been reported to decrease to ~ 50 ng/mL once bone growth ceases [19]. Herein, serum periostin (>31 ng/mL) was associated with the cumulative Pfirrmann score but not with age, sex, BMI, CCI score, anaemia, or spinal disease in adjusted multivariate regression models. However, because periostin is easily transferred or released from inflammatory areas into many different body fluids, its expression increases in several diseases characterised by allergic inflammation, fibrosis, atherosclerosis, and tumour formation [11]. Therefore, the use of a median cutoff value (>31 ng/mL) should be evaluated with caution because of the influence of comorbidities; however, periostin was a significant factor in the adjusted multivariate analysis. Hence, periostin may serve as a clinically relevant and useful biomarker that can aid in the diagnosis, estimating disease progression, activity, and prognosis and allowing for appropriate IVDD treatment modalities selection [11].

Periostin may affect IVDD concerning mechanical stress and inflammation and provide flexibility and load transmission over the entire spine, as affected by mechanical stresses [6, 7]. Periostin is highly expressed in mechanically stressed tissues and promotes injury repair in many tissues [20]. However, its overexpression is associated with various diseases characterised by inflammation, apoptosis, fibrosis, atherosclerosis, and tumour formation [20]. The magnitude and duration of mechanical stress are positively correlated with apoptosis in NP cells [20], and periostin expression is considerably higher in human degenerated NP cells than in non-degenerated NP cells [10]. Therefore, the loss of NP cells by apoptosis decreases IVD function and leads to periostin overexpression. Overexpression of periostin in IVD promotes cartilage ECM degradation and induces chondrocyte apoptosis, leading to IVDD [11, 21]. Therefore, a positive correlation between IVDD and periostin is logical. Although obesity, sarcopenia, and osteoporosis, which affect the mechanical stress on IVD, have been reported as risk factors for IVDD [5], they did not contribute to IVDD in the present study. Further studies are needed to examine the impact of the set of endpoints for obesity, sarcopenia, and osteoporosis and the low patient variability and small number of cases.

Inflammation is another important feature of the IVDD environment [10]. IVDD occurs when the NP, which has no blood vessels, is exposed to circulation, resulting in inflammation and triggering an autoimmune inflammatory response [22]. High levels of TNF-α, IL-6, and other proinflammatory cytokines are released by disc cells, exacerbating inflammation through interaction with periostin and contributing to IVDD progression [22]. Thus, periostin may be involved in the pathogenesis of IVDD via mechanical stress and inflammation. The impact of mechanical stress and inflammation on IVDD may be related to the fact that periostin was identified as an independent risk factor in multivariate analysis. Moreover, mechanical stress factors, including obesity, osteoporosis, and sarcopenia and inflammation-related conditions such as anaemia and CCI may not be sufficient to serve as biomarkers for IVDD.

Periostin has been intensively investigated and clinically used as a biomarker, especially for inflammatory and diverse allergic diseases, including asthma, allergic rhinitis, chronic sinusitis, atopic dermatitis, and allergic conjunctivitis in adults [11]. As a downstream molecule of IL-4 and IL-13, which have important roles in the aetiology of allergic diseases, periostin is involved in fibrosis [11]. Elevated serum periostin levels have been reported in malignancies, including head and neck cancer, breast cancer, non-small cell lung cancer, hepatocellular carcinoma, pancreatic cancer, and colorectal cancer, and they have shown potential as biomarkers for diagnosis, metastasis, and prognosis [23]. Because periostin is produced by osteocytes and osteoblasts and exhibits distinctive features in clinical conditions, it has been reported as a potential biomarker for knee osteoarthritis, osteoporosis, bone fracture repair status, and prediction of knee OA severity [24].

The relationship between periostin and IVDD suggests that it is a promising therapeutic target and biomarker. Knockdown of periostin, which encodes periostin, with siRNA and inactivation of periostin with a neutralising antibody alleviates IVD ageing [5]. Thus, periostin-neutralising antibodies might serve as potential therapeutic agents for IVD owing to their anti-ageing effects. Targeted therapies have been extensively reported in recent years, and the periostin-mediated regulation of myelinuclear cell apoptosis and inflammation, which are important pathological changes in IVDD, may be targeted in the future.

There are some limitations to this study. First, potential confounders, such as physical activity, allergic diseases, occupation, and educational status, were not considered in the baseline characteristics of the recruited patients because of the retrospective nature of the study. Second, because this was a cross-sectional study, the correlation between elevated periostin levels and VDD severity was clear; however, it is unclear whether a causal relationship exists. Conversely, after excluding confounding factors, the model demonstrated a strong correlation between the two variables. Third, given the potential for selection bias, it may have been ideal to include a control group of community dwellers. Fourth, the Pfirmann scale is subjective; in the future, the use of deep learning-based algorithms for automated radiological grading will be the mainstay [25]. Lastly, the mechanism by which periostin affects disc degeneration, particularly the relationship between periostin and inflammatory cytokines in IVD cells, remains unclear.

Therefore, further studies are required. Future prospective longitudinal studies should deal with these limitations. The results of this study should be considered exploratory and hypothesis generating.

Conclusion

Higher serum periostin levels correlated with IVDD severity. Periostin may be a valuable biomarker for the diagnosis of IVDD, estimating disease progression and activity and providing direct prognostic information.