Acute symptom-reporting after routine exposure to OP pesticides in sheep dip Using data reduction to provide a clinical profile of symptoms: a clinical profile of symptoms: Prof. Craig A. Jackson Division of Psychology Birmingham City University
Aims Using data reduction to provide a useful clinical profile of symptoms Subjectivity of objective methods – even in quantitative studies Factor AnalysisFactor Analysis Kaiser CriteriaKaiser Criteria Scree AnalysisScree Analysis K-means ClusteringK-means Clustering
Prevalence of non-specific symptoms (Western Australia n = 3016) SymptomPrevalence % Stuffy nose46.2Headaches33.0 Tiredness29.8Cough25.9 Itchy eyes24.7Sore throat22.4 Skin rash12.0Wheezing10.1 Respiratory10.0Nausea9.0 Diarrhoea5.7Vomiting4.0 Heyworth & McCaul, 2001 Symptom reporting in society
Common Present Day Symptoms Multiple Chemical Sensitivity (MCS) 90’s Chronic Fatigue Syndrome (formerly ME)90’s Sick Building Syndrome80’s Gulf War Syndrome90’s Musculoskeletal problems90’s Electrical Sensitivity90’s Historical Symptoms Railway Spine1860’s Combat Syndrome1850’s onward Wool sorter’s syndrome1860’s Unexplained symptom syndromes
Sheep Dipping
Organophosphate Pesticides (OPs) & Sheep Dipping UK Sheep dipped twice yearly, and was compulsory 1984–1988 OPs were the dip of choice & recommended by HSE Routine sheep dipping is wet and messy work NOT usually an acute exposure Chronic and low level exposure Non-specific symptoms alleviate 48 hours post-dip “Dippers’ Flu” AnxietyDepressionFatigueAches & Pains HeadacheFever Neurobehavioural problems (memory, concentration) The UK Sheep Dipping Saga
Dippers’ Flu Dipping sheep with Organophosphate Pesticides Traditionally tied to collection of non-specific symptoms Manifests shortly after dipping Spontaneously remits usually after 48 hours Is there really any truth in this collection of symptoms? Or is it just “background” symptomology General weakness Muscle weakness Fever Aches and pains Headaches Loss of appetite “Does Dippers’ Flu really exist?”
Past work Previous study – Stephens et al Investigated symptom reports in 82 farmers recently exposed to OPs when dipping sheep Compared with controls (quarry workers) Overall symptom reporting, and reporting of symptom groups was not elevated in exposed relative to controls This was not consistent for all symptoms Possible that this (intuitive) grouping of symptoms may mask some genuine symptom patterns. A statistical approach was needed...
Aims & Objectives a) Establish a plausible basis for grouping of symptoms b) Identify recognisable core-symptoms consistently present in exposed workers c) Determine if exposed and controls differ in these core-symptoms e) Determine if any excess in core-symptoms is dose- related (e.g. number of sheep, or years dipping)
Strategy for Re-analysis a) Cluster analysis of symptom data using original symptom groups b) Rank individual symptoms by frequency c) Chi square analyses of individual symptoms d) Factor analysis of 73 original symptoms e)Investigation of dose-effect relationships f)Cluster analysis of the symptom data
v
Individual symptoms at 24 hours after dipping 12 symptoms reported more by exposed than controls 15 symptoms reported more by controls than exposed
Factor Analysis of all symptoms Kaiser Criteria – use factors with Eigenvalue >1
Factor Analysis of all symptoms Twenty-one unwieldy factors
Scree Analysis
Factor Analysis 6 factors remained as the best representation of symptom data The product of 38 individual symptoms Accounted for 47% of the variance in symptom scores
Factor Analysis Little physiological commonality between symptoms in each factor
Kaiser Criteria versus Scree Analysis Twentyone factors representing 81% of variance or Six factors representing 46% of variance Dilemma!!!!! Kaiser criteria of gives too many factors Scree analysis often gives too few factors Both good under optimal conditions.... This study not optimal – 73 variables and 100 cases!!!
Dose-effect relationship? Weak association between 6 factors & flock size (R 2 =0.4) Weak association between 6 factors & flock size (R 2 =0.4) Flock size as surrogate exposure estimate is too simplistic Flock size as surrogate exposure estimate is too simplistic Statistical problem - 38 symptoms and only 82 cases Statistical problem - 38 symptoms and only 82 cases
K means Clustering A method for reducing data based on differences not similarities (as in Factor Analysis) Uses Euclidean Distances between factors All 73 symptomsAll 73 symptoms 21 symptoms reported most frequently21 symptoms reported most frequently 12 symptoms reported sig more by dippers than controls12 symptoms reported sig more by dippers than controls Subjected to “K-Means” cluster analysis Produced 5 distinct symptom clusters With seemingly useful physiological explanations
12 symptoms reported more – K means Clustering
12 symptoms reported more by dippers than controls Subjected to “K-Means” cluster analysis Produced 3 distinct symptom clusters With seemingly useful physiological explanations
Top 21 symptoms – K means Clustering
21 symptoms reported most frequently Subjected to “K-Means” cluster analysis Produced 5 distinct symptom clusters With seemingly useful physiological explanations “Gross” “Shiver” “Flu-like” “Muscular” “Global”?
Testing the 5 cluster model Scores on the 5 clusters compared between exposed and controls Significant differences were consistently to the detriment of the exposed grossshiverflu-likemuscularglobal
Conclusions High frequency of symptoms in both occupational groups (approx ) High frequency of symptoms in both occupational groups (approx ) No evidence of more dippers’ flu symptoms in exposed than the controls No evidence of more dippers’ flu symptoms in exposed than the controls No plausible pattern was evident in symptoms reported by the exposed No plausible pattern was evident in symptoms reported by the exposed Cluster analysis of the original 9 symptom groups showed globalized and non-specific symptoms were being reported more than localised specific symptoms - suggesting general malaise than specific target organ systems Cluster analysis of the original 9 symptom groups showed globalized and non-specific symptoms were being reported more than localised specific symptoms - suggesting general malaise than specific target organ systems Factor analysis provided little clarification of the data - it reduced 21 unwieldy factors down to 6 factors, though with little physiological plausibility in the grouping together of some symptoms Factor analysis provided little clarification of the data - it reduced 21 unwieldy factors down to 6 factors, though with little physiological plausibility in the grouping together of some symptoms K means cluster analysis identified 5 distinct symptom clusters of better plausibility, 3 of which were significantly worse in exposed K means cluster analysis identified 5 distinct symptom clusters of better plausibility, 3 of which were significantly worse in exposed
Summary Tentative support for the view that certain symptoms can be identified occurring more frequently in those exposed to OPs. Such symptoms are consistent with a flu like illness. Further verification is needed from studies specifically targeted at a definition of symptom groups following acute OP exposure.