statistical analysis – American ME and CFS Society

Bayesian Statistics Improves Biological Interpretability of Metabolomics Data from Human Cohorts

Abstract:

Univariate analyses of metabolomics data currently follow a frequentist approach, using p-values to reject a null hypothesis. We here propose the use of Bayesian statistics to quantify evidence supporting different hypotheses and discriminate between the null hypothesis versus the lack of statistical power.

We used metabolomics data from three independent human cohorts that studied the plasma signatures of subjects with myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS). The data are publicly available, covering 84-197 subjects in each study with 562-888 identified metabolites of which 777 were common between the two studies and 93 were compounds reported in all three studies. We show how Bayesian statistics incorporates results from one study as “prior information” into the next study, thereby improving the overall assessment of the likelihood of finding specific differences between plasma metabolite levels.

Using classic statistics and Benjamini-Hochberg FDR-corrections, Study 1 detected 18 metabolic differences and Study 2 detected no differences. Using Bayesian statistics on the same data, we found a high likelihood that 97 compounds were altered in concentration in Study 2, after using the results of Study 1 as the prior distributions. These findings included lower levels of peroxisome-produced ether-lipids, higher levels of long-chain unsaturated triacylglycerides, and the presence of exposome compounds that are explained by the difference in diet and medication between healthy subjects and ME/CFS patients.

Although Study 3 reported only 92 compounds in common with the other two studies, these major differences were confirmed. We also found that prostaglandin F2alpha, a lipid mediator of physiological relevance, was reduced in ME/CFS patients across all three studies. The use of Bayesian statistics led to biological conclusions from metabolomic data that were not found through frequentist approaches. We propose that Bayesian statistics is highly useful for studies with similar research designs if similar metabolomic assays are used.

Source: Brydges C, Che X, Lipkin WI, Fiehn O. Bayesian Statistics Improves Biological Interpretability of Metabolomics Data from Human Cohorts. Metabolites. 2023 Aug 31;13(9):984. doi: 10.3390/metabo13090984. PMID: 37755264; PMCID: PMC10535181. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10535181/ (Full text)

Understanding pediatric long COVID using a tree-based scan statistic approach: an EHR-based cohort study from the RECOVER Program

Abstract:

Objectives: Post-acute sequalae of SARS-CoV-2 infection (PASC) is not well defined in pediatrics given its heterogeneity of presentation and severity in this population. The aim of this study is to use novel methods that rely on data mining approaches rather than clinical experience to detect conditions and symptoms associated with pediatric PASC.

Materials and methods: We used a propensity-matched cohort design comparing children identified using the new PASC ICD10CM diagnosis code (U09.9) (N = 1309) to children with (N = 6545) and without (N = 6545) SARS-CoV-2 infection. We used a tree-based scan statistic to identify potential condition clusters co-occurring more frequently in cases than controls.

Results: We found significant enrichment among children with PASC in cardiac, respiratory, neurologic, psychological, endocrine, gastrointestinal, and musculoskeletal systems, the most significant related to circulatory and respiratory such as dyspnea, difficulty breathing, and fatigue and malaise.

Discussion: Our study addresses methodological limitations of prior studies that rely on prespecified clusters of potential PASC-associated diagnoses driven by clinician experience. Future studies are needed to identify patterns of diagnoses and their associations to derive clinical phenotypes.

Conclusion: We identified multiple conditions and body systems associated with pediatric PASC. Because we rely on a data-driven approach, several new or under-reported conditions and symptoms were detected that warrant further investigation.

Source: Lorman V, Rao S, Jhaveri R, Case A, Mejias A, Pajor NM, Patel P, Thacker D, Bose-Brill S, Block J, Hanley PC, Prahalad P, Chen Y, Forrest CB, Bailey LC, Lee GM, Razzaghi H. Understanding pediatric long COVID using a tree-based scan statistic approach: an EHR-based cohort study from the RECOVER Program. JAMIA Open. 2023 Mar 14;6(1):ooad016. doi: 10.1093/jamiaopen/ooad016. PMID: 36926600; PMCID: PMC10013630. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10013630/ (Full text)

Modeling diurnal hormone profiles by hierarchical state space models

Abstract:

Adrenocorticotropic hormone (ACTH) diurnal patterns contain both smooth circadian rhythms and pulsatile activities. How to evaluate and compare them between different groups is a challenging statistical task. In particular, we are interested in testing (1) whether the smooth ACTH circadian rhythms in chronic fatigue syndrome and fibromyalgia patients differ from those in healthy controls and (2) whether the patterns of pulsatile activities are different. In this paper, a hierarchical state space model is proposed to extract these signals from noisy observations. The smooth circadian rhythms shared by a group of subjects are modeled by periodic smoothing splines. The subject level pulsatile activities are modeled by autoregressive processes. A functional random effect is adopted at the pair level to account for the matched pair design. Parameters are estimated by maximizing the marginal likelihood. Signals are extracted as posterior means. Computationally efficient Kalman filter algorithms are adopted for implementation. Application of the proposed model reveals that the smooth circadian rhythms are similar in the two groups but the pulsatile activities in patients are weaker than those in the healthy controls.

Source: Liu Z, Guo W. Modeling diurnal hormone profiles by hierarchical state space models. Stat Med. 2015 Oct 30;34(24):3223-34. doi: 10.1002/sim.6579. Epub 2015 Jul 7. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4592415/ (Full article)

A randomised trial of adaptive pacing therapy, cognitive behaviour therapy, graded exercise, and specialist medical care for chronic fatigue syndrome (PACE): statistical analysis plan.

Abstract:

BACKGROUND: The publication of protocols by medical journals is increasingly becoming an accepted means for promoting good quality research and maximising transparency. Recently, Finfer and Bellomo have suggested the publication of statistical analysis plans (SAPs).The aim of this paper is to make public and to report in detail the planned analyses that were approved by the Trial Steering Committee in May 2010 for the principal papers of the PACE (Pacing, graded Activity, and Cognitive behaviour therapy: a randomised Evaluation) trial, a treatment trial for chronic fatigue syndrome. It illustrates planned analyses of a complex intervention trial that allows for the impact of clustering by care providers, where multiple care-providers are present for each patient in some but not all arms of the trial.

RESULTS: The trial design, objectives and data collection are reported. Considerations relating to blinding, samples, adherence to the protocol, stratification, centre and other clustering effects, missing data, multiplicity and compliance are described. Descriptive, interim and final analyses of the primary and secondary outcomes are then outlined.

CONCLUSIONS: This SAP maximises transparency, providing a record of all planned analyses, and it may be a resource for those who are developing SAPs, acting as an illustrative example for teaching and methodological research. It is not the sum of the statistical analysis sections of the principal papers, being completed well before individual papers were drafted.

TRIAL REGISTRATION: ISRCTN54285094 assigned 22 May 2003; First participant was randomised on 18 March 2005.

Source: Walwyn R, Potts L, McCrone P, Johnson AL, DeCesare JC, Baber H, Goldsmith K, Sharpe M, Chalder T, White PD. A randomised trial of adaptive pacing therapy, cognitive behaviour therapy, graded exercise, and specialist medical care for chronic fatigue syndrome (PACE): statistical analysis plan. Trials. 2013 Nov 13;14:386. doi: 10.1186/1745-6215-14-386. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4226009/ (Full article)

Chronic fatigue syndrome: a clinical and laboratory study with a well-matched control group

Comment on: Chronic fatigue syndrome: a clinical and laboratory study with a well matched control group. [J Intern Med. 1995]

Dear Sir,

It is an ongoing debate whether concurrent occurrence of particular additional symptoms should be part of the definition of chronic fatigue syndrome (CFS) [1–5] or not. Studies on the similarities and differences between patients satisfying the various definitions are indispensable to solve this dispute.

Swanink et al. [6] studied CFS patients satisfying the criteria described by Sharpe et al. [3], i.e. additional symptoms may be present but are not required. Part of the group also satisfied the more stringent CFS criteria by the Centers for Disease Control (CDC) [1], which require the additional presence of at least eight specific symptoms. When the number of complaints was included as the covariate, no significant differences on fatigue severity, depression and functional impairment were found between CFS patients who fulfilled the CDC criteria and who did not. Furthermore, the authors remarked that the sole effect of applying the CDC symptom criteria to their study group is separating patients with few symptoms from patients with many symptoms.

These results are very misleading and have often been misinterpreted. The authors’ analysis of variance (anova) yielded a lot of significant differences between CDC–CFS and non-CDC–CFS patients. That these were lost in their subsequent analysis of covariance (ancova) is because the level of the covariate and the treatment (fulfilment of the CDC criteria) are highly dependent, as fulfilment of the CDC criteria requires the presence of at least nine symptoms (fatigue included). Because the ancova assumption that the covariate is statistically independent of the treatment is not met, the ancova results are artificial and have little practical meaning [7, 8].

What happened* is illustrated in Fig. 1. anova checks whether CDC–CFS and non-CDC–CFS patients have equal test score means and and . ancova, however, checks the equality of adjusted test score means and and . These are obtained by transporting and from the treatment covariate means and and along parallel regression lines to the grand covariate mean . Thus ancova predicts if test score means of CDC–CFS and non-CDC–CFS patients would have been equal if both groups had exactly the same mean number of complaints. It provides an answer to a question that has no relevance – the mean number of complaints is inherently different for these two groups. In particular, Table 1 of the article learns that the grand covariate mean as reported with the standardized questionnaire equals = 674/88 = 7.66: the adjusted mean corresponds to a group of CDC–CFS patients that does not even exist in reality!

Figure 1.

Open in figure viewer

Download Powerpoint slide

Difference between analysis of variance (anova) and analysis of covariance (ancova).anova compares the treatment means and , whereas ancova compares the adjusted treatment means and corresponding to the same level of the covariate for both groups. Note that the grand covariate mean corresponds to a group of CDC–CFS patients that does not exist in reality.

Although their ancova was inappropriate, the authors’anova did result in valuable information.anova of CDC–CFS versus non-CDC–CFS yielded significant differences (at least P < 0.05) in concentration, activity, sleep and rest, ambulation, alertness behaviour, and recreation and pastimes, which according to the authors means that CDC–CFS patients are significantly more impaired in daily functioning. As the subjective fatigue subscale of the checklist individual strength (CIS-fatigue) easily reaches the extreme end of its scale in CFS samples (see e.g. [9, 10]), it is obvious that no significant differences in fatigue severity as measured by CIS-fatigue could be found. Generally speaking, assessing fatigue severity using a scale without this flaw may well result in different outcomes (see e.g. [10]).

Because the inadequate ancova made it appear that there are no clinical differences between CDC–CFS and non-CDC–CFS patients, this study has often been cited to permit leaving out additional symptom criteria when considering CDC–CFS. This has had major consequences for scientific research as well as for clinical practice. In scientific literature, non-CDC–CFS patients are labelled as having ‘a diagnosis of CFS according to the CDC criteria’ [10] or fulfilling ‘the CDC criteria for CFS’ [11], although other sources by the same authors explicitly state that they do not [12, 13]. In a large randomized study on cognitive behaviour therapy for CFS [14], one of the two reasons that patients without the required number of additional symptoms were included is that ‘patients who fulfilled the CDC-criteria did not differ concerning the severity of the complaints from patients who did not satisfy the CDC criteria’ [13]. The CFS definition used for clinical practice in large parts of the Netherlands [15] is based on CDC criteria, but patients without the required additional symptoms are also diagnosed CFS because ‘clinically this distinction has no meaning, as it has turned out from Dutch research’ [16]. This means that if the mistakes above would have been noted at an earlier stage, literally thousands of chronically fatigued patients might have had a different diagnosis in the Netherlands.

Apparently [13] the incorrect results of the article have also been presented during a recent meeting held for revising the latest CDC–CFS definition (presentation Bleijenberg, CDC consensus meeting, Atlanta 2000). To prevent more scientific research on CDC–CFS that disregards additional symptoms and more CFS definitions that are based on statistical errors rather than on data, it is important that the mistakes in the article are corrected as soon as possible.

You can read the rest of this comment here: http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2796.2004.01378.x/full

Source: Stouten B. Chronic fatigue syndrome: a clinical and laboratory study with a well-matched control group. J Intern Med. 2004 Sep;256(3):265-7; author reply 268-9. http://onlinelibrary.wiley.com/doi/10.1111/j.1365-2796.2004.01378.x/full (Full article)

Cost-effectiveness of cognitive behaviour therapy for patients with chronic fatigue syndrome

Comment on: Cost-effectiveness of cognitive behaviour therapy for patients with chronic fatigue syndrome. [QJM. 2004]

Sir,

I read Severens et al.’s article on the cost-effectiveness of cognitive behaviour therapy for patients with unexplained chronic fatigue1 with interest, although as several subjects met the CDC criteria for ‘idiopathic chronic fatigue’ rather than ‘chronic fatigue syndrome’,2,,3 I prefer to use the term ‘unexplained chronic fatigue’ as defined by Fukuda et al.3 to describe the patient sample under consideration.

To be able to regard the presented cost estimates as a valid reflection of the medical costs of patients with unexplained chronic fatigue, it is imperative to demonstrate that there are no differences between participants who are included in the analysis and participants who are excluded from the analysis.

According to the authors: ‘An extensive comparison between participants in the cost-effectiveness analyse (n = 171) and the remaining clinical study participants (n = 99) did not reveal any statistically significant differences regarding age, duration of CFS complaints, and scores for Sickness Impact Profile, Karnofsky score, physical activity, a self-efficacy scale, a causal attribution list, and functional impairment.’ (pp. 158–9).

Although details are lacking in the article, baseline data of the included and excluded participants are available from a publication of the Health Care Insurance Board of the Netherlands (College voor zorgverzekeringen).4 Comparing baseline variables of the two groups using two-tailed independent sample t-tests yields the results that are presented in Table 1. The table shows that physical activity (measured by a motion-sensing device called the actometer), self-efficacy, and psychological well-being (measured by the symptom checklist 90) are significantly different at the 0.05 level. The p values for physical activity (p = 0.0081) and self-efficacy (p = 0.0046) are particularly small.

You can read the rest of this comment here: http://qjmed.oxfordjournals.org/content/97/6/379.long

Source: Stouten B. Cost-effectiveness of cognitive behaviour therapy for patients with chronic fatigue syndrome. QJM. 2004 Jun;97(6):379-80. http://qjmed.oxfordjournals.org/content/97/6/379.long (Full article)

Graded exercise in chronic fatigue syndrome. Including patients who rated themselves as a little better would have altered results

Comment on:

Randomised controlled trial of graded exercise in patients with the chronic fatigue syndrome. [BMJ. 1997]

Managing chronic fatigue syndrome in children. [BMJ. 1997]

Editor—“Editor’s choice” in the issue of 7 June states, “we agree that myalgic encephalomyelitis (or chronic fatigue syndrome) is a serious condition” and “all conditions have a mental and physical component.” This is the stance of the patient organisations supporting patients with this condition. Unfortunately, some doctors have trivialised this illness; ridiculed patients and their supporters; and subjected a few of them, including children, to oppressive, perhaps even abusive, forms of treatment. Hopefully, this is now a thing of the past. We need, as Harvey Marcovitch says, to explore what might be done to help them.

You can read the rest of this comment here: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2127632/pdf/9361550.pdf

Source: Franklin AJ. Graded exercise in chronic fatigue syndrome. Including patients who rated themselves as a little better would have altered results. BMJ. 1997 Oct 11;315(7113):947; author reply 948. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2127632/

Can the chronic fatigue syndrome be defined by distinct clinical features?

Abstract:

To determine whether patients diagnosed as having chronic fatigue syndrome (CFS) constitute a clinically homogeneous class, multivariate statistical analyses were used to derive symptom patterns and potential patient subclasses in 565 patients. The notion that patients currently diagnosed as having CFS constitute a single homogeneous class was rejected.

An alternative set of clinical subgroups was derived. The validity of these subgroups was assessed by sociodemographic, psychiatric, immunological and illness behaviour variables. A two-class statistical solution was considered most coherent, with patients from the smaller class (27% of the sample) having clinical characteristics suggestive of somatoform disorders. The larger class (73% of sample) presented a more limited combination of fatigue and neuropsychological symptoms, and only moderate disability but remained heterogeneous clinically. The two patient groups differed with regard to duration of illness, spontaneous recovery, severity of current psychological morbidity, utilization of medical services and CD8 T cell subset counts. The distribution of symptoms among patients was not unimodal, supporting the notion that differences between the proposed subclasses were not due simply to differences in symptom severity.

This study demonstrated clinical heterogeneity among patients currently diagnosed as CFS, suggesting aetiological heterogeneity. In the absence of discriminative clinical features, current consensus criteria do not necessarily reduce the heterogeneity of patients recruited to CFS research studies.

Source: Hickie I, Lloyd A, Hadzi-Pavlovic D, Parker G, Bird K, Wakefield D. Can the chronic fatigue syndrome be defined by distinct clinical features? Psychol Med. 1995 Sep;25(5):925-35. http://www.ncbi.nlm.nih.gov/pubmed/8588011

Chronic fatigue syndrome. Role of psychological factors overemphasised

Comment in: Chronic fatigue syndrome and myalgic encephalomyelitis. [BMJ. 1994]

Comment on: Longitudinal study of outcome of chronic fatigue syndrome. [BMJ. 1994]

Editor,-In concluding that psychological factors are more important than immunological ones in determining the long term outcome of myalgic encephalomyelitis or the chronic fatigue syndrome Andrew Wilson and colleagues seem overconfident of the validity of their findings. Although the use of self rated measures of outcome is necessary, the validity of the investigators’ treatment of such data is questionable. For example, the five point self rated global illness outcome was dichotomised such that an original response of “not improved at all” was recorded to “worsened”-a decision the investigators fail to justify. It is also dubious whether patients’ recall of their own premorbid psychological state is accurate, given that the average onset was 9 years before recall and the finding that memory of an event is affected by subsequent events.

You can read the rest of this comment here: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2540179/pdf/bmj00440-0053a.pdf

Source: Blatch C, Blatt T. Chronic fatigue syndrome. Role of psychological factors overemphasised. BMJ. 1994 May 14;308(6939):1297. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2540179/