Planned analyses
The investigators will perform the following analyses, with terms based on final categorization above:
Primary analyses:
I. Dependent variable (outcome): positive = significant growth, negative = no growth or possible significant growth. Main explanatory variable (group): MC1 vs. control II. Dependent variable (outcome): positive = significant growth, negative = no growth or possible significant growth. Main explanatory variable (group): MC2 vs. control
For each of the primary analyses (I and II), we will perform the following sensitivity analyses:
1. st sensitivty analysis: Dependent variable (outcome): positive = significant growth or possible significant growth, negative = no growth and 16S rDNA nanopore sequencing positive.
2. nd sensitivity analysis: Dependent variable (outcome): positive = significant growth, possible significant growth or 16S rDNA nanopore sequencing positive, negative = no growth
Exploratory analyses:
The investigators will perform exploratory analyses using the following as main explanatory variables (using the same dependent variable as the primary analysis):
I - MC1 and MC2 in previously operated vs. control II - Large MCs vs. control (significant growth from disc, yes/no). Large MCs are defined as MCs with volume ≥ 25 % of vertebral body volume or height \> 50 % of vertebral body height.
III - MC1 and MC2 vs. control in fusion group (significant growth from vertebral body biopsy, yes/no). Vertebral body biopsies are not performed in the disc herniation group.
For each of these three exploratory analyses, the investigators will consider to perform similar sensitivity analyses as for the primary analyses, depending on the results of those.
Statistics and power:
The main aim of this study is to investigate if the proportions of patients with significant bacterial growth from perioperative disc biopsies differ between cases (MC1 patients or MC2 patients) vs controls without MCs. The null hypothesis is that there is no difference between cases and controls. The alternative hypothesis is that there is a difference.
The sample size calculation is based on previously published data and a pre-specified relevant difference in proportions of bacterial growth among cases vs. controls
The investigators calculated the sample size using a two-sided Pearson's chi-squared test. For the primary analysis, with two primary endpoints, the investigators aim to achieve 80 % power to detect a difference in bacterial growth in 25 % of cases with MC1 or MC2 vs. 5% of controls. Due to multiple testing the investigators use Bonferroni correction (alfa 0.025). The investigators therefore plan to analyze at least 60 cases with MC1, 60 cases with MC2 and 60 controls.
The MC2 sample is likely to become larger than n = 60, since the investigators recruit MC1 and MC2 patients consecutively and MC2 is more common than MC1.
The primary endpoint will be analyzed with a logistic regression model with bacterial growth (positive/negative) as the outcome and group (MC1 or MC2 vs. control) as the main explanatory variable.
After fitting the model, the model-predicted marginal probabilities of positive bacterial findings will be estimated for both groups. The effect measure will be the difference between the two probabilities, and will be reported with a 95 % confidence interval and a P-value for the null hypothesis of a zero difference. The standard error of the difference will be estimated using the delta method.
The exploratory and sensitivity analyses will be carried out with the same model, after replacing outcome and main explanatory variable as appropriate. The subgroup analysis of previously operated will be carried out by adding previously operated as an interaction term between groups, and previously operated as covariates in the logistic model. A significant coefficient for the interaction term will indicate a subgroup effect.
