Modic changes in patients with lumbar disc herniation followed more than 1 year after lumbar discectomy: a systematic review and meta-analysis

Background

Modic changes (MCs) (Modic & Ross, 2007; Modic et al., 1988b) are commonly observed lesions in the vertebral endplates of the spine during radiological examinations. Scholars have proposed that Modic changes are independent risk factors for episodes of low back pain (Hopayian, Raslan & Soliman, 2023; Määttä et al., 2015). These changes manifest distinct pathological features and can be identified on magnetic resonance imaging (MRI). Modic type 0 (MC0) indicates no Modic changes. Modic type 1 (MC1) is characterized by bone edema, leading to reduced signal intensity on T1-weighted images and significant enhancement on T2-weighted images. In contrast, Modic type 2 (MC2) displays features of yellow fat with signal enhancement on both T1- and T2-weighted MRI images. Modic type 3 (MC3) represents sclerotic bone and demonstrates low signal intensity on both T1- and T2-weighted MRI images (Dudli et al., 2016; Modic et al., 1988a, 1988b).

Until the 2010s, follow-up studies of patients with MCs primarily focused on observing the natural progression of these changes, with limited emphasis on evaluating the impact of surgical interventions (Albert & Manniche, 2007; Jensen et al., 2009; Kuisma et al., 2006). Discectomy, a common medical procedure for severe disc herniation, has shown efficacy in patients who do not respond to long-term conservative treatment (Cenic & Kachur, 2009), especially in terms of reducing or eliminating low back pain in patients with various MC types (Nian et al., 2023). However, some articles have reported surgical promotion of Modic type transformation. For instance, Albert & Manniche (2007) noted a higher incidence of MCs among patients with low back pain who had undergone surgery within 1 year of follow-up than among those who did not undergo surgery. A systematic evaluation with higher-level evidence has shown that patients with preoperative MC1 tend to have poorer functional outcomes after discectomy, and MC after lumbar discectomy can significantly affect prognosis (Nian et al., 2023). While some studies have suggested a trend towards the conversion of MC0 and MC1 to MC2 (Albert & Manniche, 2007; Kuisma et al., 2006), a prospective cohort study pointed out a more complex pattern of MC transformation after limited discectomy in the lumbar spine (Bostelmann et al., 2019).

Therefore, understanding the impact of the type of MC alteration on pain and functional prognosis after discectomy is crucial. However, there is still a lack of high-level evidence regarding the postoperative transformation of MC alterations. This study aims to investigate the incidence and trend of MC transformation post-lumbar discectomy.

Methods

Risk of bias assessment

Two investigators evaluated all included studies. The quality of observational studies was assessed using the Newcastle-Ottawa Scale (NOS) (Bostelmann et al., 2019; Kawaguchi et al., 2023; Ohtori et al., 2010; Rahme et al., 2010; Takahashi et al., 2021; Yaman et al., 2017). The NOS scale (total score, 8), commonly employed as an evaluation tool for cohort studies, is composed of three parts: Selection of the study population, comparability between groups, and outcome measures. In the study population selection, one point was rated for meeting the representativeness of the exposed group, the method of selecting the nonexposed group, the method of determining exposure factors, and the determination of outcome indicators at the beginning of the study; two points for controlling for confounding factors during design and statistical analysis; and one point each for adequacy of the assessment of study outcomes, follow-up time, and follow-up in the outcome measures section. Table 1 presents detailed components of the NOS scale. Figure 1 illustrates the Risk of Bias (RoB2) tool used for the risk of bias evaluation for RCTs (el Barzouhi et al., 2014; Lee & Bae, 2015). Risk of bias may arise from various factors, including the country of study, selection of the study population, choice of procedures, and duration of follow-up. Disagreements, if any, were resolved through consultation with an independent third party.

Table 1:

The detailed description of newcastle-ottawa scale (NOS) used to evaluate included non-RCT articles.

	Selection				Comparability	Outcomes
Study	A	B	C	D	E	F	G	H	Score
Takahashi et al. (2021)	▲	▲	▲	▲	▲	▲	▲	▲	8
Bostelmann et al. (2019)	▲	▲	▲	▲	▲▲	▲	▲	▲	9
Yaman et al. (2017)	▲	▲	▲		▲	▲	▲	▲	7
Rahme et al. (2010)	▲	▲	▲	▲	▲▲	▲	▲	▲	9
Ohtori et al. (2010)	▲	▲	▲			▲	▲	▲	6
Kawaguchi et al. (2023)	▲	▲	▲	▲	▲▲	▲	▲	▲	9

DOI: 10.7717/peerj.17851/table-1

Note:

A, Representativeness of the exposed cohort; B, Selection of the non-exposed cohort; C, Ascertainment of exposure; D, Demonstration that outcome of interest was not present at start of study; E, Comparability of cohorts on the basis of the design or analysis; F, Assessment of outcome; G, Follow-up long enough for outcomes to occur; H, Adequacy of follow up of cohort.

Results

Study selection

Initially, we identified 100 articles, which included 13 articles from PubMed, 18 articles from Embase, 63 articles from Web of Science, four articles from Cochrane, and two articles from manual search. After removing duplicates, 58 articles remained. Through title and abstract screening, 39 articles were further excluded, leaving 19 articles for full-text assessment. Subsequently, six studies were excluded due to no Modic type illustration, four for incomplete outcomes, and one for no clear surgical procedure, resulting in eight articles (Bostelmann et al., 2019; el Barzouhi et al., 2014; Kawaguchi et al., 2023; Lee & Bae, 2015; Mostofi, Moghaddam & Peyravi, 2018; Ohtori et al., 2010; Rahme et al., 2010; Takahashi et al., 2021; Yaman et al., 2017) included in the meta-analysis (Fig. 2).

Study characteristics

Table 2 provides detailed characteristics of the included studies, including country of origin, patient demographics, procedure types, body mass index, duration of pain, MC groups, follow-up time, and MC diagnosis criteria. All studies utilized similar MC diagnosis criteria and grouping standards. Although some articles mentioned inclusion criteria for MC3 and mixed MC types (Lee & Bae, 2015), these subtypes were excluded due to limited data. All eight studies presented outcome indicators at both preoperative and postoperative time points. In total, 728 patients diagnosed with LDH were identified, with 689 patients meeting the inclusion criteria (445 with MC0 (64.6%), 70 with MC1 (10.2%), and 174 with MC2 (25.2%)), after excluding cases with preoperative MC3, mixed MC types, or lost follow-up. The incidence of MC conversion in all groups was 27.7%. Patients with MC0 exhibited a higher incidence of conversion to any other MC types (37.0%), compared to those with MC1 and MC2 (20.5% and 19.1%, respectively; Figs. 3–6).

Table 2:

Summary of eight included articles.

Author (Publish year)	Country	Journal	Study period	Procedure type	Age	BMI	Position of Modic change	Female (%)	Modic type before surgery	Modic type after surgery	Clinical symptoms	Duration of pain	Follow-up period (baseline)	Inclusion criteria	Exclusion criteria	MC diagnosis
Prospective study
Takahashi et al. (2021)	Japan	BMC Musculoskeletal Disorders	2010–2018	Microdiscectomy	43.6 + 14.7 (<75)	/	L3/4:6	31 (48%)	MC0:48	MC0:40	Persistent and unremitting LBP	>3 months	1 year	(1) The presence of persistent and unremitting LBP for more than 3 months, X-ray images, and magnetic resonance imaging (MRI).	(1) Patients with LSS or degenerative spondylolisthesis (DS) comorbidities.	Type 1 showed decreased signal intensity on T1-weighted images and increased signal intensity on T2-wcighted images.
									MC1:3	MC1:8					(2) Patients who were diagnosed with a lateral herniation treated with fusion surgery
							L4/5:28		MC2:10	MC2:12				(2) Those who were not improved by sufficient conservative treatment and wished to undergo surgical treatment.	(3) Patients who were with thoracic myelopathy and hip osteoarthritis.	Type 2 showed increased signal intensity on T1-weighted images and isointense or slightly increased signal intensity on T2-weightcd images (Disc degeneration was graded using the Pfirrmann scale.)
							L5/S:31		MC3:1	MC3:2					(4) Patients with recurrent herniation.
									Unable:3
Rahme et al. (2010)	Lebanon	Journal of Neurosurgery-Spine	2004–2008	Microdiscectomy	Mean:54 (24–78)	/	L4-L5 and L5-S1	14 (34%)	MC0:22	MC0:9	/	/	3–5 years	Patients with a virgin (previously nonoperated) lumbar spine.	/	1) Type 1 changes, which are hypointense on T1-weighted imaging and hyperintense on T2-weighted imaging and represent bone marrow edema and inflammation;
									MC1:5	MC1:6						2) Type 2 changes, which are hyperintense on T1-weighted imaging and isointense or slightly hyperintense on T2-weighted imaging and are associated with fatty degeneration of the bone marrow;
									MC2:14	MC2:26
Ohtori et al. (2010)	Japan	Spine (Phila Pa 1976)	/	Discectomy	35.5 ± 5	/	L4-L5:27	19 (42.2%)	MC0:22	MC0:19	LBP and leg pain	>3 months	2 years	(1) Patients had low back and leg pain, continuing for at least 3 months.	(1) Patients who had previously undergone spinal surgery	Modic type 1 signals (low intensity on T1weighted spin-echo images and high intensity on T2-weighted spin-echo images), those with Modic type 2 signals (high intensity on both T1- and T2-weighted spin-echo images), and those with
										MC1:21
							L5-S1:18		MC1:23	MC2:6				(2) Patients were diagnosed with lumbar disc herniation on MRI.	(2) Spinal tumor, infection, and trauma
										MC3:0
Retrospective study
Kawaguchi et al. (2023)	Japan	Global Spine Journal	Between January 2013 and December 2018	Microscopic discectomy	47.4 ± 10.2	23.2 ± 2.7	L3/4: 6	40 (41%)	MC0:96	MC0:67	Redicular pain	11.6 ± 5.3 wks	18 ± 4.6 months	The inclusion criteria were lumbar disc herniation on MRI with corresponding radicular pain, failure of conservative treatment for 6 weeks or more, no spinal instability as assessed on functional radiographs, and a minimum 1-year follow-up following discectomy.	The exclusion criteria were coexisting lumbar spinal canal stenosis, history of lumbar surgery, cauda equina syndrome, and insufficient collection of tissue samples	Modic change at the operated level was assessed based on Modic classification which was divided into 6 groups; type 0, type 1, type 2, type 3, and mixed type 1/2 and 2/3
										MC1:6
							L4/5:49			MC2:14
							L5/S:41			MC2:14
Bostelmann et al. (2019)	Germany	European Spine Journal	2010–2014	Discectomy	44.0 + 7.0 (21–75)	25.78 (23.23–29.22)	/	107(38%)	MC0:115	MC0:69	Back pain/Leg pain/Combined	>6 wks	2 year	(1) Age between 21 and 75 years	(1) Prior index level surgery.	Type 0: No oedematous reaction or vascular congestion induced in the adjacent bone marrow of the endplates
									MC1:30	MC1:64					(2) Spondylolisthesis with >25% slip.	Type 1: Hypointense reaction and vascular congestion in the adjacent marrows on T1-weighted MR imaging; hyperintense on T2-weighted images, new or increased relative to the previous time point
									MC2:121	MC2:96				(2) The presence of lumbar radiculopathy due to a primary, single-level disc herniation between L1 and S1	(3) Scoliosis of greater than 10 degrees (both angular and rotational).	Type 2: Bone marrow converted to a predominantly fatty marrow. Hyperintense on T1 and isointense to hypointense on T2. The exact signal intensity depends on the degree of T2 weighting
									MC3:1	MC3:2				(3) Failure of at least 6 weeks of nonsurgical treatment.	(4) Insulin-dependent diabetes.
										Unable:36					(5) Contraindication for MRI.
Yaman et al. (2017)	Turkey	Hong Kong Med J	2004–2009	Discectomy	53.2 ± 12.3	24.8 ± 0.8	L4-L5	8 (32.0%)	MC0:7	MC0:5	Radicular pain with or without neurological deficit, numbness in the lumbar spine, buttock, and/or lower extremity	>3 months	Average 320 days	(1) Radicular pain for at least 3 months that was refractory to 6 weeks of conservative treatment with or without neurological deficit,	(1) Prior lumbar surgery at another institution	Modic endplate changes were also classified with the help of their radiological department on T1/T2weighted sagittal MRI sequences.
									MC1: 4	MC1:6				(2) Numbness in the lumbar spine, buttock, and/ or lower extremity,	(2) Segmental instability,
															(3) Vertebral fractures and spinal infections,
									MC2:11	MC2:11				(3) Age between 21 and 75 years,	(4) Other types of degenerative disc disease
															(5) Tumours
									MC3:3	MC3:3				(4) Magnetic resonance imaging (MRI) and/or computed tomography demonstrating anatomical unilateral LDH correlating with symptoms.	(6) Pregnancy,
															(7) Age over 75 years
Random controlled trial
el Barzouhi et al. (2014)	Netherlands	The Spine Journal : Official Journal of the North American Spine Society	/	Discectomy	41.9 + 9.8	26.2 ± 3.9	L3-L4:7	57 (34%)	MC0:99	MC0:32	Persistent radicular pain in the L4-S1 dermatome with or without neurological deficit/Severe disabling leg pain	6-12 wks	1 year	(1) Age 18–65 yr	(1) Cauda equina syndrome or severe paresis (MRC < 3)	Type 1 lesions, hypointense on T1-weighted images and hyperintense on T2-weighted images, represent marrow edema, and are associated with an acute process. Type 2 lesions, the most common type, have increased signal on T1-weighted images and isointense or slightly hyperintense signal on T2-weighted images, and represent fatty degeneration of subchondral marrow and are associated with a chronic process.
										MC1:82				(2) Persistent radicular pain in the L4, L5 or S1 dermatome with or without mild neurological deficit	(2) Complaints of a lumbosacral radicular syndrome in the same dermatome within the past 12 months
									MC1:1	MC2:31				(3) Severe disabling leg pain of 6–12 weeks duration	(3) A history of unilateral disc surgery on the same level
							L4-L5:73			Mixed MC1 and MC2:20				(4) Evidence of a unilateral disc herniation confirmed on MRI	(4) Spinal canal stenosis
									MC2:67	Otherwise:3				(5) Sufficient knowledge of Dutch language	(5) Degenerative or lytic spondylolisthesis
														(6) Informed consent	(6) Pregnancy
							L5-S1:88		Mixed MC1 and MC2:1						(7) “Severe life-threatening” or psychiatric illness
															(8) Planned migration to another country in the year after randomization
Lee & Bae (2015)	South Korea	Int J Clin Exp Med	2008–2009	Automated open lumbar discectomy (AOLD) or Microdiscectomy (MD)	42.7 ± 11.5	/	L3-4/L4-5/L5-S1 1/14/25	13 (32.5%)	MC0:24	MC0:18	Pain/Neurologic deficit	4.6 ± 4.9 Wks	20 months	(1) The presence of disc herniation as determined by magnetic resonance imagining (MRI) and pain that persisted for 4-8 weeks after conservative treatment involving rest, analgesia and physical therapy.	(1) Patients age older than 69 years, previous surgery, severe lumbar stenosis, spondylosis, spondylolisthesis, extraforaminal far lateral disc herniation, foraminal spur or bony compression, metabolic bone disease,	Type 1 Modic changes were hypointense on T1-weighted imaging (T1WI) and hyperintense on T2-weighted imaging (T2WI), type 2 Modic changes were hyperintense on T1WI and isointense or slightly hyperintense on T2WI, and type 3 Modic changes were hypointense on both T1WI and T2WI.
									MC1:1	MC1:7
									MC2:3	MC2:3				(2) Patients with progressive neurologic deficit underwent emergency operations.	(2) Patients receiving worker’s compensation, and patients who have coexisting lumbar spinal disease.
									MC3:0	MC3:0

DOI: 10.7717/peerj.17851/table-2

Postoperative MCs

Meta-analysis of the conversion rates towards MC0, MC1, and MC2 groups from all included studies was conducted (Fig. 4–7). Conversion rates towards MC0 from both MC1 and MC2 groups, as well as those towards MC3 across all groups, approached 0% (MC1→MC0, 0%, p = 0.969, 95% CI [0–1.3]; MC2→MC0, 0.8%, p = 0.113, 95% CI [0–6.7]; MC0→MC3, 0.1%, p = 0.849, 95% CI [0–0.1]; MC1→MC3, 0%, p = 0.959, 95% CI [0–0]; MC2→MC3, 0%, p = 0.613, 95% CI [0–1.6]). MC0 group had a higher proportion of conversion to MC1 than the MC2 group (MC0→MC1, 17.7%, p < 0.001, 95% CI [6.8–31.7]; MC2→MC1, 12.4%, p < 0.001, 95% CI [0.4–32.7]). Conversely, the MC1 group showed a higher tendency to convert into MC2 than the MC0 group (MC0→MC2, 13.1%, p < 0.001, 95% CI [6.7–21.0]; MC1→MC2, 19.0%, p = 0.329, 95% CI [7–33.8]). Moderate to high heterogeneity persisted across most subgroups (MC0→MC1 and MC2; I² = 89.689% and 73.367%, respectively; MC1→MC2; I² = 13.159%; MC2→MC0 and MC1: I² = 43.911% and 82.178%, respectively).

Discussion

To the best of our knowledge, this review presents the first examination of long-term outcomes (more than 1-year post-surgery) regarding MC transformations in patients with LDH, particularly those with a mix of preoperative MC types. The evidence, of moderate to high quality, indicates a higher likelihood of Modic type transitions in patients initially without preoperative MC (MC0) than in those with pre-existing combined MC types (MC+). Specifically, a significant proportion of patients transitioned to MC1 (17.7%), followed by MC2 (19.0%), while transitions to MC3 were notably infrequent (0%). Interestingly, individuals with a preoperative blend of MC1 types were more prone to such transformations than those with MC2 phenotypes, although the prevalence were similar (MC1 = 20.5%, MC2 = 19.1%). However, neither group showed evidence of reverting back to MC0. The process illustrating the conversion of Modic changes is depicted in Fig. 8.

Previous research indicates that patients with LDH exhibiting combined preoperative MC1 are associated with diminished functional status post-discectomy. Traditionally, prognostic assessments post-discectomy have focused on preoperative MCs (Karadağ et al., 2022; Li et al., 2021; Ulrich et al., 2020; Xu et al., 2019), yet the impact of surgical intervention on MC trajectory has not been thoroughly examined. The transformation of MCs following surgery may significantly contribute to persistent postoperative pain, underscoring the importance of analyzing the prevalence and patterns of MCs post-lumbar discectomy for understanding surgical outcomes and refining clinical management strategies. Furthermore, the progression of Modic changes post-discectomy is clinically significant as it can influence long-term outcomes and pain management strategies in patients with lumbar disc herniation (Albert & Manniche, 2007; Jensen et al., 2012; Kuisma et al., 2007). A comprehensive understanding of these changes can aid in tailoring postoperative care and potentially enhancing patient prognosis.

In our aggregate analysis, the preoperative incidences of MC0, MC1, and MC2 were found to be 37.0%, 20.5%, and 19.1%, respectively. Importantly, our study’s MC prevalence substantially exceeded the rates observed in the general Danish population’s natural progression (Jensen et al., 2009). This disparity may be attributed to the strong association between disc degeneration and MCs. Evidence from various studies supports a positive association between the emergence of MCs and progression of lumbar disc degeneration (Kjaer et al., 2005), with severe lumbar disc degeneration being more prevalent in patients exhibiting MCs (Özcan-Ekşi et al., 2021).

Recent scholarly efforts advocate for a nuanced classification of MCs, such as Xu et al. (2016)’s observation of mixed types combining endplate edema and sclerosis in MC patients. However, our study focused on traditional MC typology due to limited representation of mixed-type cases, potentially increasing heterogeneity and confounding factors in our findings. It is widely accepted that MCs can evolve from one type to another, representing different phases of the same pathological process. The concept of mixed types likely represents a transitional phase (Weishaupt et al., 1998), with evidence suggesting surgical interventions may accelerate this transition (Albert et al., 2008; Kanna et al., 2017), contrasting with the average follow-up duration in our included studies.

Moreover, Bostelmann et al.’s (2019) study, which met our study’s inclusion criteria, examined a population with low back pain persisting for just over 6 weeks, notably shorter than that in all included studies. This brief duration may indicate a higher prevalence of unstable MC1 in patients, potentially leading to more frequent MC transitions post-surgery. In eight studies, patients who underwent lumbar discectomy exhibited conversions between MC1 and MC2, with a higher incidence of transitioning from MC1 to MC2 preoperatively than from MC2 to MC1. Notably, no conversion to the more stable MC3 type was observed post-surgery, possibly due to the brief follow-up timeframe, highlighting the need for extended follow-up in future research.

Conclusion

Our findings suggest an increased likelihood of postoperative MC transformation, particularly from no preoperative MC (MC0) to MC1 or MC2, with a higher prevalence of MC1. Notably, patients exhibiting a combination of preoperative MC1 types experienced more frequent postoperative transformations than those with MC2 types. This study elucidates the characteristic changes in Modic change (MC) types observed among patients undergoing lumbar discectomy, thereby enhancing comprehension of surgical outcomes and facilitating improvements in postoperative clinical care quality.

Supplemental Information