Development and internal validation of a machine learning prediction model for low back pain non-recovery in patients with an acute episode consulting a physiotherapist in primary care

Background

While low back pain occurs in nearly everybody and is the leading cause of disability worldwide, we lack instruments to accurately predict persistence of acute low back pain. We aimed to develop and internally validate a machine learning model predicting non-recovery in acute low back pain and to compare this with current practice and ‘traditional’ prediction modeling.

Methods

Prognostic cohort-study in primary care physiotherapy. Patients (n = 247) with acute low back pain (≤ one month) consulting physiotherapists were included. Candidate predictors were assessed by questionnaire at baseline and (to capture early recovery) after one and two weeks. Primary outcome was non-recovery after three months, defined as at least mild pain (Numeric Rating Scale > 2/10). Machine learning models to predict non-recovery were developed and internally validated, and compared with two current practices in physiotherapy (STarT Back tool and physiotherapists’ expectation) and ‘traditional’ logistic regression analysis.

Results

Forty-seven percent of the participants did not recover at three months. The best performing machine learning model showed acceptable predictive performance (area under the curve: 0.66). Although this was no better than a’traditional’ logistic regression model, it outperformed current practice.

Conclusions

We developed two prognostic models containing partially different predictors, with acceptable performance for predicting (non-)recovery in patients with acute LBP, which was better than current practice. Our prognostic models have the potential of integration in a clinical decision support system to facilitate data-driven, personalized treatment of acute low back pain, but needs external validation first.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12891-022-05718-7.

THU0637-HPR THE ASSOCIATION BETWEEN THE RECALL PERIOD AND THE AMOUNT OF INFORMATION ABOUT REPORTED ADVERSE DRUG REACTIONS BY PATIENTS USING BIOLOGICALS

Background:

In order to monitor the safety of medicines pharmacovigilance, it is important that patients report their adverse drug reactions (ADRs). Theoretically, the quality of the reported information might be affected by the elapsed time between the onset of the ADR and the moment of reporting. Real-life evidence demonstrating a negative relationship between this recall period and the quality of reported ADRs is however lacking.

Objectives:

To assess the effect of recall period on the amount of information that patients report about their ADR (information density) in patients using a biologic for an immune-mediated inflammatory disease (IMID).

Methods:

The Dutch Biologic Monitor is a multi-center cohort ADR monitoring system collecting data on reported ADRs by patients using a biologic for an IMID. Per patient, every first unique reported ADR between 1 February 2017 and 1 September 2019 was eligible. ADR reports were selected by stratified random sampling based on length of recall period and biologic. The recall period was defined by the number of days between the onset and reporting date of the ADR. The amount of information in an ADR report (information density) was determined based on eleven domains: specification, location, frequency, time-to-onset, course, causality, cause or consequence, health care professional (HCP) visits, HCP action, patient action and ADR burden. Information density was calculated by the number of reported domains divided by the number of domains deemed relevant in the ADR report. The association between the information density of the ADR reports and different recall periods was compared using a one-way ANOVA test. One-way ANOVA and independent t-tests were used to assess the impact of gender, age, type biologic and burden of the ADR on the information density of the reported ADRs.

Results:

Out of 1109 reported ADRs by 531 IMID patients, we included 402 ADR reports of 294 patients (55%) (see table 1). Included reports were equally divided over seven different recall periods: 0-1, 1-2, 2-4, 4-8, 8-12, 12-26 and 26-52 weeks. Results have shown no association between the information density in patient-reported ADRs and the length of recall period (p=0.805) (figure 1). However, the proportion of reported information about HCP visits for the ADR increased with increasing recall period: 0-1 week (14%), 1-2 weeks (24%), 2-4 weeks (34%), 4-8 weeks (40%), 8-12 weeks (48%), 12-26 weeks (50%) and 26-52 weeks (46%).

Table 1.

Characteristics of included patients with adverse drug reactions

Characteristics (N=294)N (%)Gender (female)202 (69%)Age (years) (mean ± SD)53 ± 13Smoking59 (20%)BMI (kg/m2) (mean ± SD)25.7 ± 5.3Reported ARDs (mean ± SD)1.4 ± 0.8IndicationRheumatoid arthritis129 (44%)Psoriatic arthritis51 (17%)Axial spondyloarthritis43 (15%)Crohn’s disease42 (14%)Other indications29 (10%)BiologicAdalimumab97 (33%)Etanercept72 (24%)Infliximab27 (9%)Tocilizumab16 (5%)Secukinumab15 (5%)Rituximab14 (5%)Other biologics53 (18%)

Female patients reported more information about their ADR (p=0.002), whereas the patient’s age was not associated with information density (p=0.221). Etanercept (ETA) users report significantly more information than adalimumab (ADA) users (p=0.019). The number of patients using other biologics was too low for further analysis. A higher ADR burden tended (p=0.120) to result in more reported ADR information (figure 2).

Conclusion:

The length of recall period did not affect the amount of information that patients report about their ADR(s). The recall period was longer for patients reporting information about their HCP visit. Furthermore, female patients tend to report more information about their ADR than male patients and ETA-users tend to report more than ADA-users.

Disclosure of Interests:

Alexandra Laurijssen: None declared, Jette van Lint: None declared, Bart van den Bemt Grant/research support from: UCB, Pfizer and Abbvie, Consultant of: Delivered consultancy work for UCB, Novartis and Pfizer, Speakers bureau: Pfizer, AbbVie, UCB, Biogen and Sandoz., Lilian Beijer: None declared, Naomi Jessurun: None declared

(1-->3)-beta-D-glucan does not induce acute inflammation after nasal deposition

To assess if (1→3)-β-D-glucan, a microbial cell wall agent normally present in pollen, has the ability to produce pollenlike response, sensitive persons received a nasal deposition of two doses of (1→3)-β-D-glucan. The percentage of eosinophils and amount of eotaxin were measured in nasal lavage 30 minutes and 24 hours after challenge. No effect could be demonstrated. The absence of an inflammatory response after (1→3)-β-D-glucan application confirms earlier findings in inhalation studies.

Mould exposure at home relates to inflammatory markers in blood

Living in damp buildings has been associated with airway symptoms, suspected to be due to inflammatory reactions. The relationship between home exposure to mould and signs of inflammation was, therefore, studied. Nonsmoking subjects with a high (G-high, > 4.0 ng x m(-3), n = 17) or low (G-low, < 2.0 ng x m(-3), n = 18) amount of airborne beta(1 --> 3)-D-glucan, an indicator of mould biomass, in the home were recruited. Blood samples were analysed for granulocytic enzymes, T-cell subsets and the secretion of cytokines from in vitro incubated peripheral blood mononuclear cells (PBMCs). In the G-high group, the proportion of cytotoxic T-cells (CD8+S6F1+) was lower and secretion of tumour necrosis factor-alpha from PBMCs higher than in the G-low group. There were no significant differences in secretion of interferon gamma and interleukin (IL)-4 from PBMCs between the two groups. Among nonatopic subjects, the ratio between interferon gamma and IL-4 was significantly higher in the G-high group than in the G-low group and was related to the amount of beta(1 --> 3)-D-glucan in the home. No significant differences were found regarding secretion of IL-10 or IL-Ibeta from PBMCs, eosinophil cationic protein or myeloperoxidase in serum, or differential cell counts in blood. The effects found on inflammatory markers in relation to beta(1 --> 3)-D-glucan in the home suggest upregulation of some parts of the inflammatory/immunological system due to mould exposure.

Work related symptoms among sewage workers: a nationwide survey in Sweden

AIMS

To assess the risk for work related symptoms among sewage workers in Sweden using a postal questionnaire.

METHODS

All municipalities in Sweden were contacted and asked to provide addresses of sewage workers and controls. Controls were recruited among other municipal workers not exposed to sewage, such as workers in drinking water plants and gardeners. A questionnaire was sent to the subjects and after two reminders, the response rate was 74% among sewage workers and 59% among controls.

RESULTS

Significantly increased risks for airway symptoms, chronic bronchitis, and toxic pneumonitis, as well as central nervous system symptoms such as headache, unusual tiredness, and concentration difficulties were found among the sewage workers compared with controls. Furthermore, an increased risk for non-specific work related gastrointestinal symptoms was found among the sewage workers; an increased risk for joint pains, related to pains in more than four joints but not with loading, was also found.

CONCLUSIONS

The results of this questionnaire survey show an increased risk for airway, gastrointestinal, and general symptoms such as joint pains and central nervous system symptoms among sewage workers. Clinical investigations are needed to determine the cause of the reported symptoms among sewage workers, and further field studies are required to assess the causal agents.

Effects after inhalation of (1-->3)-beta-D-glucan in healthy humans

BACKGROUND AND AIM

This study was performed to assess the effects of an exposure to a pure (1-->3)-beta-D-glucan, a cell wall component of fungi, plants and certain bacteria.

METHODS

Twenty-one healthy subjects inhaled saline or (1-->3)-beta-D-glucan suspended in saline in a random, double-blind, cross-over design. They were examined before exposure and 24 and 72h afterwards with spirometry, blood sampling and collection of induced sputum. Differential cell counts and eosinophilic cationic protein (ECP) were determined in blood and sputum, and myeloperoxidase (MPO), tumour necrosis factor-alpha (TNF-alpha), and interleukin (IL)-8 and IL-10 were determined in sputum supernatants. TNF-alpha was determined after cultivation of blood mononuclear cells.

RESULTS

In sputum, inhalation of saline caused a significant increase in ECP and TNF-alpha. (1-->3)-beta-D-Glucan inhalation caused a further increase in these cytokines, although not statistically significantly different from the increase induced by inhalation of saline alone. In blood, the number of eosinophils was significantly decreased 72 h after the challenge with (1-->3)-beta-D-glucan. This effect was not found after the inhalation of saline alone. TNF-alpha production from stimulated blood mononuclear cells was significantly decreased 72 h after the (1-->3)-beta-D-glucan inhalation as compared with the increase induced by saline inhalation.

CONCLUSIONS

The results suggest that (1-->3)-beta-D-glucan causes a different type of response as compared with inflammatory agents such as bacterial endotoxin that cause a neutrophil-dominated inflammatory response.

Airways inflammation and glucan exposure among household waste collectors

A field study was made on 17 workers collecting unsorted household waste, eight workers collecting organic/nonorganic separated waste, and 24 controls. Measurements of airborne endotoxin and (1-->3)-beta-D-glucan were made in their working environments. Examinations consisted of a questionnaire for symptoms, spirometry, airway responsiveness, and blood and sputum sampling for determination of cell counts, eosinophilic cationic protein (ECP), and myeoloperoxidase (MPO). A higher proportion of waste collectors reported diarrhea, congested nose, and unusual tiredness as compared to controls. The number of blood lymphocytes was higher among waste collectors and were dose-related to the amount of airborne (1-->3)-beta-D-glucan at the workplaces. The amount of ECP and the number of macrophages were lower in sputum among waste collectors as compared with controls. The results suggest that certain dusts from household waste may cause airway inflammation as well as general symptoms, and the effects were associated with higher (1-->3)-beta-D-glucan levels.

Monocyte responsiveness and a T-cell subtype predict the effects induced by cotton dust exposure

The aim of the study was to evaluate whether peripheral cellular parameters could predict susceptibility to decreased lung function and associated symptoms, in response to a single exposure to cotton dust. Previously nonexposed subjects (n = 42) inhaled an aerosol of cotton dust in a model cardroom during a period of 5 h. The subjects were examined before the exposure for FEV1, procoagulant activity (PCA) in blood mononuclear cells (BMNC), and serum IgE antibodies against a pool of inhalant antigens. Blood lymphocytes were typed into the helper/inducer (CD4+) and cytotoxic/suppressor (CD8+) T cells in combination with surface markers subdividing these populations. A questionnaire was used to identify atopic and nonatopic subjects. Immediately after exposure, the subjects were tested for FEV1 and PCA, and symptoms were recorded with a questionnaire. The dust exposure induced a decrease in FEV1 that was larger for the atopic group, but did not change the PCA in BMNC. The decrease in FEV1 was positively related to the preexposure PCA in both atopics and nonatopics. Symptoms from the airways after the exposure were reported to the same extent in the atopic and nonatopic group, and the subject group reporting chest tightness had a larger preexposure PCA. The atopic group had a larger proportion of blood CD8+ T lymphocytes negative for the monoclonal antibody anti-S6F1 (CD8+S6F1-), and in this group the decrease in FEV1 was significantly related to the proportion of this cell type. Also, in the atopic group, the proportion of CD8+S6F1- cells correlated positively with the preexposure PCA, and a negative correlation was found for this cell and serum levels of IgE.(ABSTRACT TRUNCATED AT 250 WORDS)

The glpT and glpQ genes of the glycerol regulon in Bacillus subtilis

The cloning of the Bacillus subtilis glpT and glpQ genes and their nucleotide sequences are reported. Analysis of mRNA indicates that glpT and glpQ constitute one operon which is transcribed from a sigma A type promoter. The steady state amount of glpTQ mRNA is increased in cells grown in the presence of glycerol 3-phosphate. The 5' untranslated leader sequence of glpTQ mRNA contains an inverted repeat which shows sequence similarity to repeats present in the leader sequences of glpFK and glpD transcripts. These repeats seem therefore to be essential control elements for all B. subtilis glp genes.

The glpP and glpF genes of the glycerol regulon in Bacillus subtilis

The Bacillus subtilis glpPFKD region contains genes essential for growth on glycerol or glycerol 3-phosphate (G3P). The nucleotide sequence of glpP encoding a regulatory protein and the previously unidentified glpF encoding the glycerol uptake facilitator was determined. glpF is located immediately upstream of glpK and the two genes were shown to constitute one operon which is transcribed separately from glpP. A sigma A-type promoter and the transcriptional start point for glpFK were identified. In the 5' untranslated leader sequence (UTL) of glpFK mRNA a conserved inverted repeat is found. The repeat is believed to be involved in the control of expression of glpFK by termination/antitermination of transcription, a control mechanism previously suggested for the regulation of glpD encoding G3P dehydrogenase. Expression of glpFK and glpD requires the inducer G3P and the glpP gene product. A 2.9 kb B. subtilis chromosomal DNA fragment containing the glpP open reading frame was cloned to give plasmid pLUM7. pLUM7 contains a functional glpP gene as shown by its ability to complement various glpP mutants. Immediately upstream of glpP an open reading frame is found (ORF1). Disrupting ORF1 by plasmid integration in the B. subtilis chromosome does not affect the ability to grow on glycerol as sole carbon and energy source. With the present report all B. subtilis glp genes located at 75 degrees on the chromosomal map have been identified.

Utilisation of glycerol and glycerol 3-phosphate is differently affected by the phosphotransferase system in Bacillus subtilis

Glycerol and glycerol 3-phosphate uptake in Bacillus subtilis does not involve the phosphotransferase system. In spite of this, B. subtilis mutants defective in the general components of the phosphotransferase system, EnzymeI or Hpr, are unable to grow with glycerol as sole carbon and energy source. Here we show that a Hpr mutant can grow on glycerol 3-phosphate and that glycerol 3-phosphate, but not glycerol, can induce glpD encoding glycerol-3-phosphate dehydrogenase. Induction of glpD also requires the glpP gene product which is a regulator of all known glp genes. Thus the phosphotransferase system general components do not interfere with the overall regulation of the glp regulon. Revertants of a Hpr mutant which can grown on glycerol carry mutations closely linked to the glp region at 75 degrees on the B. subtilis chromosomal map. This region contains the glpP, the glpFK and the glpD operons. The glpFK operon encodes the glycerol uptake facilitator (glpF) and glycerol kinase (glpK). The present results demonstrate that one of these genes, or their gene products, is the target for phosphotransferase system control of glycerol utilisation. Furthermore we conclude that utilisation of glycerol and glycerol 3-phosphate is differently affected by the phosphotransferase system in B. subtilis.

Glycerol catabolism in Bacillus subtilis: nucleotide sequence of the genes encoding glycerol kinase (glpK) and glycerol-3-phosphate dehydrogenase (glpD)

The glpPKD region of the Bacillus subtilis chromosome was cloned in its natural host in plasmid pHP13. The glpPKD region contains genes required for glycerol catabolism: glpK coding for glycerol kinase, glpD coding for glycerol-3-phosphate (G3P) dehydrogenase and glpP, proposed to code for a positively acting regulatory protein. The cloned 7 kb fragment carries wild-type alleles of glpK, glpD and glpP. It can also complement a strain deleted for the entire glpPKD region. The wild-type alleles were mapped to different subfragments, establishing the gene order glpP-glpK-glpD. The nucleotide sequence of glpK and glpD was determined. Immediately upstream of glpK, an additional open reading frame was found, possibly being part of the same operon. Putative transcription terminators were found in the region between glpK and glpD and downstream of glpD. In a coupled in vitro transcription/translation system, two proteins were found, corresponding in size to those predicted from the deduced amino acid sequences of glycerol kinase and G3P dehydrogenase (54 kDa and 63 kDa, respectively).

Increased blood monocyte procoagulant activity in cotton mill workers

Blood monocyte procoagulant activity has previously been related to delayed type hypersensitivity. In this study, cotton workers exposed to cotton dust containing endotoxin and subjects not exposed to organic dusts, were examined. Blood mononuclear cells from the two groups were incubated with and without endotoxin and the recalcification time was measured. Mononuclear cells from cotton workers had a decreased baseline procoagulant activity but an increased response to endotoxin, suggesting cellular sensitization to the endotoxin present in cotton dust.

Modulation of pulmonary inflammation after endotoxin inhalation with a platelet-activating factor antagonist (48740 RP)

An acute pulmonary response was induced in guinea pigs and hamsters by inhalation of bacterial endotoxin in the form of a purified lipopolysaccharide (LPS). Pretreatment with the platelet-activating factor (PAF) antagonist, 48740 RP, inhibited damage to endothelial cells, decreased vascular permeability and the number of neutrophils in the airways 24 h after exposure to LPS. The increase in the number of platelets in the airways caused by endotoxin was not affected. The results suggest that PAF modulates early inflammation after endotoxin inhalation.

The involvement of platelet activating factor in endotoxin-induced pulmonary platelet recruitment in the guinea-pig

1 Exposure of conscious guinea-pigs to an aerosol of endotoxin (25-100 micrograms ml-1) resulted in a dose-related, progressive accumulation of platelets in the thoracic region. Accumulation of 111indium oxine labelled erythrocytes was not observed following exposure to an aerosol of endotoxin (50 micrograms ml-1). 2 Pretreatment of guinea-pigs with the selective platelet activating factor (Paf)-antagonists. CV-3988 or brotizolam resulted in a dose-related inhibition of endotoxin-induced pulmonary platelet recruitment. Pretreatment of guinea-pigs with the selective Paf-antagonist BN 52021 resulted in significant inhibition of endotoxin-induced pulmonary platelet recruitment, although the effects of BN 52021 were not dose-related. 3 Pretreatment of guinea-pigs with indomethacin at doses known to inhibit cyclo-oxygenase did not inhibit endotoxin-induced pulmonary platelet recruitment, whereas higher doses of indomethacin produced a reduction in platelet recruitment in the lung. 4 Pretreatment of guinea-pigs with the anticoagulant heparin and the prostacyclin analogue ZK 36374 inhibited endotoxin-induced platelet recruitment. 5 These observations suggest that endotoxin-induced pulmonary platelet recruitment in the guinea-pig is secondary to the release of platelet activating factor, but not to cyclo-oxygenase products of arachidonic acid and may also involve activation of the coagulation cascade.

Inhalation of endotoxin stimulates alveolar macrophage production of platelet-activating factor

The production of PAF was studied in alveolar macrophages (AM) and neutrophils recovered by bronchial lavage from guinea pigs exposed to aerosolized bacterial endotoxin (lipopolysaccharide, LPS). The amount of cell-associated PAF was estimated by measuring serotonin release from rabbit platelets. An increased and dose-related production was found in AM for as long as 2 h after a 40-min exposure. No production was detectable after 4 h. Prolonging the exposure did not prolong the response. When a second exposure was given, no PAF could be detected until the time interval between the 2 exposures was 72 h. The amount of neutrophils in lung lavage fluid was elevated about 100 times at 4 h after the exposure, but only a minor PAF production was found in these cells. In view of the role of LPS-contaminated dusts for the development of human lung disease, particularly airway constriction, the role of PAF needs to be further investigated.

Platelet serotonin release by human polymorphonuclear leucocytes stimulated by cotton dust bacteria

The release of serotonin from platelets was examined using polymorphonuclear leucocytes (PMN) from normal volunteers. The stimulating agents emplyed were zymosan and heat-killed bacteria from Enterobacter agglomerans, which is commonly found in cotton dust. Optimal conditions for release were established. Both zymosan and E. agglomerans yielded a release of serotonin of an equal magnitude. The data are discussed in relation to the pathogenesis of respiratory disease associated with occupational exposure to cotton dust.