Lumbar Puncture increases risk of Lumbar Degenerative Disc Disease: Analysis from the Rochester Epidemiology Project

INTRODUCTION

Degenerative disc disease (DDD) is a leading cause of chronic low back pain (LBP) where the annual cost exceeds $190 billion in the US.^1,2 The exact etiology of DDD has not yet been fully elucidated; to date, studies have shown that the degeneration may be attributed to multiple factors including ageing, living conditions, biomechanical loading and several molecular and genetic factors.³ On a cellular level, reduction in active cell numbers, depletion of extracellular matrix, altered phenotype of normal disc cells, and presence of pro-inflammatory cytokines and mediators have all been shown to be associated with degeneration.⁴ Among these, inflammatory mediators such as Interleukin (IL) 1β, IL-6, and IL-8 have recently garnered much interest.⁵ The source of these mediators has been shown to be both disc cells and circulating macrophages.⁵ Moreover, such an inflammatory response can be activated by mechanical strain in human disc cell⁶, such as an annular puncture during discography.⁷

Lumbar puncture (LP) is a common diagnostic procedure which also induces local inflammation causing LBP.⁸ Bedside LP relies on manual and mental estimation of the bony anatomy to guide the needle in the subarachnoid space which may result in frequent attempts and misplacements.⁹ Hypothetically, LP-induced back injury might initiate an inflammatory cascade, thus increasing the risk of DDD progression. Herein, we conducted a retrospective cohort study using the Rochester Epidemiology Project (REP) to estimate the incidence of degenerative disc disease in patients with a history of having a LP. We also determined the procedural factors of lumbar puncture, such as needle size, operator, use of fluoroscope, CSF color and complications, associated with an increased risk of developing DDD.

METHODS

Data Source

Rochester Epidemiology Project (REP), established in 1966, is a unique medical records-linkage system that links and indexes the records of virtually all medical providers in Olmsted County.¹⁰ It allows a user to search medical records, medical diagnoses, surgical interventions, and demographic information of virtually all persons residing in Olmsted County, Minnesota.^11,12 Nearly all health care to county residents is provided by the two major participating institutions, Mayo Clinic and Olmsted Medical Center and their affiliated hospitals, and Rochester Family Medicine Clinic. The study was approved by the Institutional Review Boards of both Mayo Clinic and Olmsted Medical Center (registration number17–010821 and 061-OMC-17, respectively). Residents of Olmsted County who had granted permission for their medical records about 90.6% of the population to be reviewed were included in the study.¹³

Study population

A population-based, retrospective cohort study was designed to investigate the association between LP and DDD. This study was reviewed and approved by Mayo Clinic and Olmsted Medical Center Institutional Review Boards. All participants were identified by using the resources of REP.A total of 1309 unique patients aged >=18 with a history of lumbar puncture procedure (Current procedural terminology, CPT code 62770 & 62772) were identified using REP during the years 2004–2009 in Olmsted County, Minnesota. Further, we identified the patients with lumbar DDD who received the occurrence of International Classification of Disease, Ninth (ICD-09) and Tenth Edition (ICD-10), codes listed in the Supplementary Table S1 and reviewed their medical reports to confirm the diagnosis. Among these, 292 were excluded for having a diagnosis of Lumbar DDD or for having a spine procedure based on ICD-09 and ICD 10, codes in Supplementary Table S2 prior to the identified lumbar puncture date. An additional 43 were excluded for not having greater than 3 months of follow-up post LP. The final cohort consisted of a total of 974 patients. To conduct a matched analysis, up to 2 controls matched on age, sex, and BMI to each LP case were identified using the following criteria: 1) patients with no prior history of lumbar puncture, DDD, or spine procedure and 2) with at least 3 months of follow-up post the matching date. At least one suitable control was found for 950 of the eligible 974 cases.

Validation:

To validate our screening methodology and case ascertainment, two authors (F.M and R.D.F) randomly selected 10 patients each from the four groups: lumbar puncture with and without DDD; no lumbar puncture with and without DDD, and performed a manual chart review.

Statistical Methods

Data were descriptively summarized by case status using frequency and percent for categorical variables, and medians and interquartile ranges for continuous variables. Data distributions across case status were compared using fisher/chi-square tests (where appropriate) for categorical variables and Wilcoxon rank-sum tests for continuous variables.

For cases, the start time for follow-up was the lumbar puncture. For controls, the start date was time of matching. Follow-up was calculated as the number of days from lumbar puncture/match to either a) loss of follow-up (patient no longer resided in Olmsted County MN), b) DDD diagnosis, or c) death. Associations between lumbar puncture and DDD diagnosis were analyzed using univariate and multivariable Cox proportional hazards regression models (multivariable models were additionally adjusted for the effects of age, BMI, gender, smoking status, HIV, malignancy, spinal cord injury, lower back injury, stroke, DM, and hypertension (HTN)). Death and loss to follow-up were modeled as censoring events. For illustrative purposes, cumulative incidence curves by case status were generated using competing risk methods with death modeled as a competing risk¹⁴. The incidence rate of DDD per 1,000 person-years of follow-up was also estimated by case status.

In patients with LP, association between LP procedure characteristics and DDD were analyzed using univariate and multivariable Cox regression. We reviewed the procedure description for LP in each case to abstract the following variables: needle size (18, 20 and 22 Gauge), operator (Physician, Nurse and trainee), use of fluoroscope (yes vs no), cerebrospinal fluid color (clear, stained and cloudy) and complication (yes vs no). A total of 730 out of 950 cases had an available clinical note in the medical chart for their LP procedure. Of those patients, 46 were missing needle size and 33 were missing CSF color. To run the regression models, multiple imputation was used with 5 independent datasets. Data were assumed to be missing at random.

All analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC). All tests were 2-sided, and p-value < 0.05 was determined to be significant.

RESULTS

Patient Cohort and Demographic Characteristics

A total of 950 LP cases were matched on age, sex, and BMI to 1876 controls without LP exposure. The matches were well balanced in regards to age, sex and BMI (Table 1). Compared to controls, we found that LP cases had a higher proportion of current smokers (31.0% vs. 22.2%, p-value=<0.001), patients with stroke (16.7% vs. 6.3%, p-value=<0.001), DM (28.1% vs. 19.8%, p-value=<0.001), and HTN (42.1% vs. 34.2%, p-value=<0.001) (Table 1). This indicated that LP group had higher co-morbidity than non-LP group which was expected as LP group received LP procedure due to their ailing clinical condition. A total of 503 patients were found to have developed DDD. The median follow-up of the study sample overall was 7.6 years, with medians of 7.3 years among LP cases and 7.7 years among controls (Table 1).

Table 1.

Baseline Demographics and Clinical Characteristics By Case Status

Characteristic	Control N=1876	Case N=950	p-value
Matched Variables
Age *	48.9 (34.1, 70.7)	49.2 (34.1, 70.4)	0.98†
Gender			0.87‡
Female	1021 (54.4%)	520 (54.7%)
Male	855 (45.6%)	430 (45.3%)
BMI *	26.7 (24.0, 30.5)	26.9 (23.5, 31.5)	0.90†
Demographics
Smoking Status			<.001‡
Missing	157	16
Never	975 (56.7%)	415 (44.4%)
Former	363 (21.1%)	229 (24.5%)
Current	381 (22.2%)	290 (31.0%)
Comorbidities
Stroke	118 (6.3%)	159 (16.7%)	<.001‡
DM	371 (19.8%)	267 (28.1%)	<.001‡
HTN	642 (34.2%)	400 (42.1%)	<.001‡
HIV	4 (0.2%)	22 (2.3%)	<.001§
Malignancy	399 (21.3%)	239 (25.2%)	0.02‡
Spinal Cord Injury	38 (2.0%)	30 (3.2%)	0.06‡
Lower Back Injury	47 (2.5%)	30 (3.2%)	0.31‡

Numbers indicate N (%) unless otherwise noted.

^*Median (Q1, Q3)

^†Wilcoxon

^‡Chi-square

^§Fisher exact

Association between LP and DDD

Patients with LP exposure were at higher risk of DDD diagnosis in univariate and multivariable comparisons (Table 2). Compared to their matched controls, LP had significantly increased risk of DDD progression (HR=1.33, 95% CI=1.10–1.60, p-value=0.003). Increased age, BMI and former/current smokers were also found to be associated with increased risk of DDD. At one year, the cumulative incidence of DDD was 3.6% (95% CI=2.5%−4.9%) among cases compared to 1.8% (95% CI=1.3%−2.5%) among controls (Figure 1). At 5 years, it was found to be 13.8% (95% CI=11.5%−16.1%) for cases vs. 9.4% (95% CI=8.1%−10.8%) for controls.

Table 2.

Associations Between Baseline Demographics and Clinical Characteristics With DDD Using Univariate and Multivariable Cox Regression

			Univariate		Multivariable†
Characteristic	Total N=2844	N DDD (%)	Hazard Ratio (95% CI)	p-value	Hazard Ratio (95% CI)	p-value
Lumbar Puncture				<.001		0.003
No	1876	308 (16.4%)	1.00 (ref)		1.00 (ref)
Yes	950	195 (20.5%)	1.37 (1.15, 1.64) ***		1.33 (1.10, 1.60) **
Age (per 10 years)				<.001		<.001
Ordinal Effect	2826	503 (17.8%)	1.38 (1.31, 1.44) ***		1.38 (1.30, 1.47) ***
BMI				0.11		0.002
Ordinal Effect	2826	503 (17.8%)	1.01 (1.00, 1.03)		1.03 (1.01, 1.04) **
Gender				0.41		0.10
Female	1541	271 (17.6%)	1.00 (ref)		1.00 (ref)
Male	1285	232 (18.1%)	1.08 (0.90, 1.28)		0.86 (0.71, 1.03)
Smoking Status				<.001		0.006
Never	1390	205 (14.7%)	1.00 (ref)		1.00 (ref)
Former	592	154 (26.0%)	1.89 (1.53, 2.33) ***		1.38 (1.11, 1.71) **
Current	671	131 (19.5%)	1.43 (1.15, 1.78) **		1.40 (1.12, 1.75) **
Unknown	173	13 (7.5%)	0.65 (0.37, 1.15)		1.02 (0.58, 1.80)
HIV				0.10		0.27
No	2800	495 (17.7%)	1.00 (ref)		1.00 (ref)
Yes	26	8 (30.8%)	1.80 (0.90, 3.62)		1.49 (0.73, 3.02)
Malignancy				<.001		0.18
No	2188	342 (15.6%)	1.00 (ref)		1.00 (ref)
Yes	638	161 (25.2%)	1.87 (1.55, 2.26) ***		1.15 (0.94, 1.41)
Spinal Cord Injury				0.002		0.06
No	2758	484 (17.5%)	1.00 (ref)		1.00 (ref)
Yes	68	19 (27.9%)	2.09 (1.32, 3.31) **		1.56 (0.98, 2.48)
Lower Back Injury				0.98		0.96
No	2749	489 (17.8%)	1.00 (ref)		1.00 (ref)
Yes	77	14 (18.2%)	0.99 (0.58, 1.69)		1.01 (0.59, 1.74)
Stroke				<.001		0.88
No	2549	438 (17.2%)	1.00 (ref)		1.00 (ref)
Yes	277	65 (23.5%)	1.75 (1.35, 2.28) ***		1.02 (0.78, 1.34)
DM				<.001		0.27
No	2188	343 (15.7%)	1.00 (ref)		1.00 (ref)
Yes	638	160 (25.1%)	1.87 (1.54, 2.25) ***		1.12 (0.91, 1.38)
HTN				<.001		0.32
No	1784	248 (13.9%)	1.00 (ref)		1.00 (ref)
Yes	1042	255 (24.5%)	2.07 (1.74, 2.47) ***		0.89 (0.72, 1.12)

^*P <.05

^**P <.01

^***P <.001

^†Multivariable Estimates and P-values are adjusted for all covariates listed in the table

Figure 1.

Cumulative probability of developing DDD from time of LP exposure between LP cohort and control cohort.

Patients with LP were at increased risk of DDD in all tested subgroups (Table 3). Overall, the incidence rate of DDD per 1,000 person years of follow-up was 31.64 among cases compared to 23.15 among controls.

Table 3.

Subgroup analysis of incidence and hazard ratio of DDD in LP patients compared with their matched controls

	Total	DDD	Incidence Rate (per 1000 person-years)		Incidence Rate Ratio		Hazard Ratio (95% CI)
Subgroup	N=2844	N (%)	Cases	Controls	(95% CI)	Unadjusted	p-value	Adjusted†	p-value†
Overall	2826	503 (17.80%)	31.64 (27.50, 36.41)	23.15 (20.71, 25.89)	1.37 (1.14, 1.64)	1.37 (1.15, 1.64)	<.001	1.33 (1.10, 1.60)	0.003
Sex
Female	1541	271 (17.59%)	29.80 (24.61, 36.08)	22.77 (19.56, 26.52)	1.31 (1.02, 1.67)	1.31 (1.02, 1.67)	0.03	1.34 (1.03, 1.73)	0.03
Male	1285	232 (18.05%)	34.10 (27.74, 41.93)	23.62 (20.03, 27.84)	1.44 (1.11, 1.88)	1.46 (1.12, 1.90)	0.005	1.32 (1.01, 1.74)	0.045
Age at LP
<40 years	999	77 (7.71%)	15.03 (10.84, 20.84)	8.50 (6.21, 11.63)	1.77 (1.12, 2.78)	1.72 (1.10, 2.69)	0.02	1.53 (0.96, 2.44)	0.08
>40 years	1827	426 (23.32%)	42.20 (36.13, 49.30)	30.94 (27.40, 34.94)	1.36 (1.12, 1.66)	1.37 (1.12, 1.66)	0.002	1.29 (1.05, 1.58)	0.02

^†Estimate and P-value are adjusted for the effects of age, gender, BMI, smoking status, HIV, malignancy, spinal cord injury, lower back injury, stroke, DM, and HTN.

Association between LP procedure characteristics and DDD

A total of 730 patients who underwent LP were further analyzed to evaluate the role of LP procedural characteristics. Interestingly, fluoroscope use had significantly lower rates of DDD (ref=no use, HR 0.55, 95% CI 0.30, 1.00, p-value=0.049) while increased age and BMI had higher rates of DDD (per 10 year increase in age HR 1.36, 95% CI 1.24, 1.48, p-value<0.001 and per 1 unit HR 1.04, 95% CI 1.01, 1.06, p-value=0.007) in multivariable cox regression analyses (Table 4).

Table 4.

Association between Lumbar Puncture Characteristics with DDD Using Univariate and Multivariable Cox Regression

			Univariate†		Multivariable†
Characteristic	Total N=730	DDD (%)	Hazard Ratio (95% CI)	p-value	Hazard Ratio (95% CI)	p-value
Needle Size, n=684				0.97
18	26	5 (19.2%)	1.00 (ref)
20	570	129 (22.6%)	1.08 (0.44, 2.64)
22	88	20 (22.7%)	1.04 (0.39, 2.74)
Operator				0.02		0.96
Physician	291	48 (16.5%)	1.00 (ref)		1.00 (ref)
Nurse	398	104 (26.1%)	1.60 (1.13, 2.25) **		0.94 (0.62, 1.44)
Trainee	41	7 (17.1%)	1.10 (0.50, 2.44)		0.91 (0.40, 2.08)
Use of Flouroscope				0.02		0.049
No	600	141 (23.5%)	1.00 (ref)		1.00 (ref)
Yes	130	18 (13.8%)	0.56 (0.34, 0.91) *		0.55 (0.30, 1.00) *
CSF Color, n=697				0.62
Clear	600	131 (21.8%)	1.00 (ref)
Stained	8	1 (12.5%)	0.60 (0.19, 1.90)
Cloudy	89	21 (23.6%)	1.10 (0.69, 1.74)
Complication				0.27
No	673	150 (22.3%)	1.00 (ref)
Yes	57	9 (15.8%)	0.68 (0.35, 1.34)
Age (per 10 years)				<.001		<.001
Ordinal Effect	730	159 (21.8%)	1.34 (1.24, 1.46) ***		1.36 (1.24, 1.48) ***
BMI				0.28		0.007
Ordinal Effect	730	159 (21.8%)	1.01 (0.99, 1.04)		1.04 (1.01, 1.06) **
Smoking Status				0.46
Never	314	68 (21.7%)	1.00 (ref)
Former	192	47 (24.5%)	1.16 (0.80, 1.69)
Current	224	44 (19.6%)	0.90 (0.61, 1.31)
Spinal Cord Injury				0.60
No	707	154 (21.8%)	1.00 (ref)
Yes	23	5 (21.7%)	1.27 (0.52, 3.11)

^*P <.05

^**P <.01

^***P <.001

The Univariate Models contained just the row covariate. The Multivariable models contained all variables <0.05 in univariate except BMI, known confounder.

^†Due to there being patients with missing needle size and CSF color, we used multiple imputation on 5 independent datasets to handle the missing data. The Univariate results for CSF Color and Needle size are from analyzing the imputed datasets, while the total N and DDD (%) are from the non-imputed data.

DISCUSSION

Lumbar puncture is a common procedure used to obtain cerebrospinal fluid for diagnostic or therapeutic purposes and spinal anesthesia.¹⁵ Post–dural puncture headache is the most common complication, occurring in up to 40 % of patients after lumbar puncture. Other complications of LP may include cranial neuropathies, prolonged backache, nerve root injury, and meningitis.¹⁶ Our results indicate that in addition to these complications, DDD may be associated with a longer term complication, i.e. development of DDD. Using a population-based retrospective cohort, we calculated a cumulative incidence of DDD of 13.8% in LP patients compared to 9.4% in non-LP patients at 5 years.

Carragee et al. conducted a prospective, match-cohort study of disc degeneration progression over 10 years among patients undergoing a lumbar discography, and found that disc puncture with a small gauge needle and limited injection pressures may be associated with accelerated disc degenerative processes, disc herniation, loss of disc height and signal and the development of reactive endplate changes compared to match-controls.⁷ Thus, it may be postulated that factors leading to an increased release of inflammatory mediators may accelerate degeneration of the disc. For instance, studies indicate that obesity is associated with significantly increased levels of IL-6 and pro-inflammatory cascades throughout the body which could contribute to an increase in inflammatory mediators that in turn promote disk degeneration. Additionally, for in vivo animal studies annular injury is commonly adopted to induce structural disc degeneration in rat lumbar discs.^18,19 All these findings suggested that inflammatory cascade induced by LP might be responsible for increased incidence of DDD among our LP cohort, which were consistent with our hypothesis (Figure 2). Previous literatures reported that DM and HTN, among other factors, might enhance DDD progression.^17,20,21 In contrast, our multivariable analysis revealed that malignancy, spinal injury, lower back injury, stroke in addition to DM and hypertension were not associated with DDD progression even though our LP cohort had higher proportion of comorbidity than control group. Overall, our results indicated that LP was strongly and independently associated with DDD progression.

Figure 2.

Summary of the hypothesis

a) Standard LP procedure where needle inserted in epidural space b) Misplaced needle causing local injury c) Inflammation induces disc degeneration.

Our results also indicate that use of fluoroscope in LP was associated with lower probability of developing DDD. This result supports our hypothesis as fluoroscopic guidance is safer and more accurate than bedside LP. A systemic review and meta-analysis by Siddharth and colleagues concluded that the failure rate of first attempt of LP was 14% - 12%.²² Fluoroscopic guidance is often requested after multiple failed attempts without imaging; or when the physician is unable to identify the bony landmarks routinely used to place the LP needle, such as in obese patients.²³ In the current health care system, it is important to note that LP is a procedure with low reimbursement and fluoroscopic image guidance can add several levels of complexity to the procedure of LP that adds considerable cost over the bedside procedure.²³ Ultrasound is relatively cheap, readily available at the point of care, compact, and involves no radiation. Ultrasound imaging can also reduce the risk of traumatic procedures and the number of needle insertions and redirections.²⁴

Our study has some limitations. The term degenerative disc disease represents disparate definitions and with greatly varying though out the publications²⁵. Although we used ICD-09 and ICD-10 for the diagnosis of DDD there could be a specific criteria for the pathology. While we adjusted for all available confounding variables associated with DDD, including age, gender, BMI, smoking status, stroke, DM, HTN, HIV, malignancy, spinal cord injury, and lower back injury, there may still be some unknown confounders that may not have been accounted for. In addition, we were unable to account for the indications and setting of the LP procedure, i.e. inpatient or outpatient, which would have further clarified our results regarding the use of fluoroscopy. Nevertheless, to the best of our knowledge this is the first study to assess the association between LP and DDD in a population based setting using an established medical records-linkage system. Additionally, our sample had long-term follow-up (median of 7.6 years) and we limited our sample to patients who resided in the locally, while censoring those who moved out during follow-up. This allows us to analyze only those who routinely visit an established academic tertiary care center for all of their care needs. Thus, we can be relatively confident in sampling all cases and in our analysis.

CONCLUSION

These results from a population based medical records-linkage system indicate that LP may be associated with an increased risk of development of DDD. Future prospective studies are warranted to prove the association between LP and DDD. Furthermore, we found that use of fluoroscopy during an LP may decrease the risk of developing a DDD.