Michael J Campbell

• PhD
• Professor Emeritus at The University of Sheffield

This chapter

Background The primary aim of the study is to test the null hypothesis that there are no statistically significant differences in intracranial volumes between male and female fetuses. Furthermore, we have studied the symmetry of the cerebral hemispheres in the cohort of low-risk fetuses. Methods 200 normal fetuses between 18 and 37 gestational wee...

Background: Peritoneal dialysis (PD) relies on the optimal functionality of the flexible plastic PD catheter present within the peritoneal cavity to enable effective treatment. As a result of limited evidence, it is uncertain if the PD catheter's insertion method influences the rate of catheter dysfunction and, thus, the quality of dialysis therap...

Background: The primary aim of the study is to test the null hypothesis that there are no statistically significant differences in intracranial volumes between male and female fetuses. Furthermore, we have studied the symmetry of the cerebral hemispheres in the cohort of low-risk fetuses. Methods: 200 normal fetuses between 18-37 gestational weeks...

This chapter explores two other regression models which are of considerable use in medical research: Poisson regression and ordinal regression . Poisson regression is an extension of logistic regression where the risk of an event to an individual is small, but there are a large number of individuals, so the number of events in a group is appreciabl...

Random effects models are used when there is more than one source of random variation. Cluster specific models have specific terms for the random effects and use some form of maximum likelihood to estimate the coefficients of the model. Random effects models can be extended to the Binomial and Poisson models. A cluster randomised trial is one in wh...

Logistic regression is a generalisation of linear regression to examine the association of a binary dependent variable with one or more independent variables that can be binary, categorical or continuous. Usually for ordinal variables the baseline is chosen as the lowest unless that category is very small. As a substitute for multiple regression wh...

A meta‐analysis is a statistical method that combines the results of multiple studies, initially trials but increasingly also observational studies. This chapter discusses fixed and random effects models for meta‐analysis, displaying data using a forest plot and the problems of missing studies and the use of a funnel plot. A key benefit of this app...

Time series regression is mainly used when the outcome is continuous, but measured together with the predictor variables serially over time. The potential for confounding in time series regression is very high – many variables either simply increase or decrease over time, and so will be correlated over time. Potential confounding factors are season...

This chapter discusses the use of dummy or indicator variables to model categories and investigate the sensitivity of models to individual data points using concepts such as leverage and influence . Multiple regression is one of the most useful tools in a statistician's armoury. The chapter describes how to interpret a computer output for linear re...

A useful model for modelling survival times on explanatory variables is known as a proportional hazards model, which is also referred to as a Cox model . In survival analysis, the key variable is the time until some event. A critical assumption in these models is that the probability that an individual is censored is unrelated to the probability th...

This chapter explains some of the basic concepts in statistical analysis, which are covered in greater depth in Statistics at Square One . The use of statistical models greatly expands the utility of statistical analysis. Data can be divided into two main types: quantitative and qualitative. Qualitative data can also be formed by categorising conti...

An international randomized phase II trial of Globo H (GH) vaccine, adagloxad simolenin/OBI-821 in 349 patients with metastatic breast cancer showed longer progression-free survival (PFS) in vaccinated patients who developed anti-Globo H (anti-GH) IgG than those who did not and the placebo group. The impacts of anti-GH IgM and GH expression on peak...

Objectives Prespecified progression criteria can inform the decision to progress from an external randomised pilot trial to a definitive randomised controlled trial. We assessed the characteristics of progression criteria reported in external randomised pilot trial protocols and results publications, including whether progression criteria were spec...

Background Pilot and feasibility studies (PAFS) often have complex objectives aimed at assessing feasibility of conducting a larger study. These may not be clear to participants in pilot studies. Methods Here, we aimed to assess the transparency of informed consent in PAFS by investigating whether researchers communicate, through patient informati...

Aim We examined the effect of spontaneous hyperglycaemia in adults with type 1 diabetes mellitus (T1DM) and without history of cardiovascular disease on heart rate variability (HRV), cardiac repolarisation and incidence of cardiac arrhythmias. Methods Thirty-seven individuals with T1DM (age 17-50 years, 19 males, mean duration of diabetes 19.3 SD(...

Misuse of statistics in medical and sports science research is common and may lead to detrimental consequences to healthcare. Many authors, editors and peer reviewers of medical papers will not have expert knowledge of statistics or may be unconvinced about the importance of applying correct statistics in medical research. Although there are guidel...

Misuse of statistics in medical and sports science research is common and may lead to detrimental consequences to healthcare. Many authors, editors and peer reviewers of medical papers will not have expert knowledge of statistics or may be unconvinced about the importance of applying correct statistics in medical research. Although there are guidel...

Introduction The successful treatment of type 1 diabetes (T1D) requires those affected to employ insulin therapy to maintain their blood glucose levels as close to normal to avoid complications in the long-term. The Dose Adjustment For Normal Eating (DAFNE) intervention is a group education course designed to help adults with T1D develop and sustai...

All statistical estimates from data have uncertainty due to sampling variability. A standard error is one measure of uncertainty of a sample estimate (such as the mean of a set of observations or a regression coefficient). Standard errors are usually calculated based on assumptions underpinning the statistical model used in the estimation. However,...

The UK Brexit debate and the current Covid pandemic have been fertile grounds for people seeking poor use of statistics, and demonstrate a need to reiterate some basic principles of data presentation. Communicating basic numbers to convey the correct message is a vital skill for a public health professional but even basic numbers can be difficult t...

Objectives To determine the rate of sudden unexpected death in infancy (SUDI) for infants born after a previous SUDI in the same family, and to establish the causes of death and the frequency of child protection concerns in families with recurrent SUDI. Design Observational study using clinical case records. Setting The UK’s Care of Next Infant (...

Introduction Pilot/feasibility studies assess the feasibility of conducting a larger study. Although researchers ought to communicate the feasibility objectives to their participants, many research ethics guidelines do not comment on how informed consent applies to pilot studies. It is unclear whether researchers and research ethics boards clearly...

Background: Accurate centre-level medication adherence measurement allows identification of highly performing CF centres, drives shared learning and informs quality improvement. Self-reported adherence is unreliable but data-logging nebulisers can capture objective data. However, adherence levels in current literature are limited by the use of agr...

Background Ultrasonography has been the mainstay of antenatal screening programmes in the UK for many years. Technical factors and physical limitations may result in suboptimal images that can lead to incorrect diagnoses and inaccurate counselling and prognostic information being given to parents. Previous studies suggest that the addition of in ut...

Background : Precision medicine is the Holy Grail of interventions that are tailored to a patient’s individual characteristics. However, conventional clinical trials are designed to find differences in averages, and interpreting these differences depends on untestable assumptions. Although only an ideal, a constant effect of treatment would facilit...

Non-Gaussian dynamic models are proposed to analyse time series of counts. Three models are proposed for responses generated by a Poisson, a negative binomial and a mixture of Poisson distributions. The parameters of these distributions are allowed to vary dynamically according to state space models. Particle filters or sequential Monte Carlo metho...

Introduction: In CF, people with higher FEV1 are less aggressively treated with intravenous (IV) antibiotics, with resultant negative impact on their health outcomes. This could be entirely clinician-driven, but patient choice may also influence IV use. In this prospective observational study, we explored IV recommendations by clinicians and IV ac...

Background : Precision medicine is the Holy Grail of interventions that are tailored to a patient’s individual characteristics. However, conventional clinical trials are designed to find differences in averages, and interpreting these differences depends on untestable assumptions. Although ideal, a constant effect would facilitate individual manage...

Background Habit, a psychological process that automatically generates urges to perform a behavior in associated settings, is potentially an important determinant of medication adherence. Habit is challenging to measure because, as a psychological construct, it cannot be directly observed. We describe a method of using routinely available objective...

Rationale, aims, and objective We previously demonstrated that annual review %FEV1 underestimates lung health of adults with CF compared with %FEV1 captured during periods of clinical stability. This has implications in the comparisons against registries with encounter‐based FEV1, such as the United States. It is uncertain whether this bias affects...

Background It has been argued that true endpoints (or ‘hard’ endpoints) for clinical trials, which are meaningful to clinicians, researchers and patients alike, are limited to those that measure health status, survival and cost. Other endpoints are termed 'surrogate' endpoints and are intended to substitute and predict the true endpoint. A number o...

Background : Precision medicine is the Holy Grail of interventions that are tailored to a patient’s individual characteristics. However, conventional clinical trials are designed to find differences with the implicit assumption that the effect is the same in all patients within the eligibility criteria. If this were the case, then there would be no...

Aims To quantify the frequency, characteristics, geographical variation and costs of emergency hospital care for suspected seizures. Design Cross-sectional study using routinely collected data (Hospital Episode Statistics). Setting The National Health Service in England 2007–2013. Participants Adults who attended an emergency department (ED) or...

Background : Precision medicine is the Holy Grail of interventions that are tailored to a patient’s individual characteristics. However, conventional clinical trials are designed to find differences with the implicit assumption that the effect is the same in all patients within the eligibility criteria. If this were the case, then there would be no...

Among adults with cystic fibrosis (CF), medication adherence is low and reasons for low adherence are poorly understood. Our previous exploratory study showed that stronger ‘habit’ (ie, automatically experiencing an urge to use a nebuliser) was associated with higher nebuliser adherence. We performed a secondary analysis of pilot trial data (n=61)...

Rationale, aims and objective Cross‐country comparisons of cystic fibrosis (CF) outcomes can potentially identify variation in care but are dependent on data quality. An important assumption is that the UK annual review FEV1 is only collected during periods of clinical stability. If this assumption does not hold, results of FEV1 comparisons may be...

This chapter describes the general nature of feasibility studies and, amongst these, what distinguishes pilot studies. Two specific types of pilot study are described. One type is an Internal‐Pilot, which is designed mainly to enable a reassessment of the sample size within an ongoing (Main) study and in which the data of the pilot is to be regarde...

The majority of clinical trials involve a simple comparison between two interventions or treatments. When there are more than two treatments, the situation is much more complicated. One problem arising at the time of analysis is that such situations may lead to multiple significance tests, resulting in misleading p‐values. Various solutions have be...

This chapter considers sample‐size calculations for comparisons between two groups where the outcome of concern is a count which is expressed as a rate per unit of time. The anticipated effect size between groups is expressed either as a difference between the two rates or by the risk ratio. The chapter presents the sample size expressions for some...

The Sample Size Software implements many of the sample size computation methods discussed in this book. The primary function of Sample Size Software is as a sample size calculator, which can be accessed under Sample Size Calculator in the main menu, to compute sample sizes corresponding to various choices of input values. The calculator takes the f...

The strength of the linear association between two continuous or ranked variables assessed in the same individuals is estimated by the Pearson or Spearmen correlation coefficients respectively. This chapter describes the formula for sample sizes to detect a correlation of a pre‐specified magnitude. The approach to identifying a lack of association,...

This chapter describes aspects of how the groups to compare are identified, situations where more than two groups are concerned, some of which may involve identifying systematic trends across the groups, and also factorial designs which can simultaneously investigate different questions within one study. It examines the role of covariates and the f...

An important part of the process of examining a patient is to check clinical measures taken from the patient against a 'normal' or 'reference' range of values. Evidence of the measure lying outside these values may be taken as indicative of the need for further investigation. This chapter describes sample sizes for establishing such reference inter...

If the null hypothesis is rejected in a comparative two‐group clinical trial, then superiority is claimed for one of the groups. In contrast, the concern in this chapter is with situations in which the aim is to claim either therapeutic non‐inferiority for the Test as compared to the Standard or, in appropriate circumstances, therapeutic equivalenc...

This chapter discusses Therapeutic Exploratory trials which include the Case‐Morgan design, which concerns survival time endpoints, and the Bryant‐Day design, which involves the simultaneous consideration of the dual endpoints of response and toxicity. To determine the Expected Total Study Length (ETSL) design, the same procedures are followed as f...

A common clinical trial design is one which measures the outcome variable of concern repeatedly over time. In any event, the sample size calculation requires a specific model known as the random intercepts model for the analysis to be identified. The model estimates the common intercept using only post‐randomisation observations whereas in equation...

Among many of the sample size approaches, and within clinical trials, different designs are available. This chapter gives an overview of the sample size approaches for two types of clinical trials—those involving drugs for a specific genomic target and those determining an appropriate dose of a new drug to be given in the first trials involving pat...

One type of cluster trial is that in which the clusters are maintained over successive periods of time during which, for example, all will commence on the standard intervention for the condition concerned and, as periods go by, more of the clusters will receive the test intervention. Such trials have what is termed a stepped wedge design (SWD). The...

This chapter describes methods for calculating sample sizes for studies which yield paired data for the situations where the outcomes are binary, ordered categorical or continuous. For binary and ordered categorical data, the components required for sample size determination comprise a measure related to the odds ratio and information on the number...

This chapter explores the sample‐size calculations for comparisons between groups where the outcome of concern is binary to when the outcome is an ordered categorical variable. The corresponding summary for a single group will be the proportion of subjects falling into each of the categories. When planning a comparative two‐group study, the anticip...

This chapter describes methods for calculating sample sizes for studies in which the outcome of primary concern is assessed by the period between a specific time‐point and the subsequent occurrence of a particular event. The reliability of a 'time‐to‐event' study depends on the total number of events that are eventually observed. In certain types o...

This chapter considers sample‐size calculations for comparisons between means of groups where the outcome of concern is continuous. It describes the situations when the data can be assumed to have a Normal distribution form, and when they do not. The chapter also describes the problem of comparing an estimated mean from one group with an assumed kn...

This chapter describes sample size calculations for observer agreement studies, with respect to the degree of either self‐agreement of a single observer assessing the same material twice or agreement between two observers both independently assessing the same specimens to make a decision with respect to a definitive diagnosis. In these situations,...

This chapter considers sample‐size calculations for comparisons between two groups where the outcome of concern is binary. The anticipated effect size between groups is expressed either as a difference between two proportions or by the odds ratio. The situation in which one of the proportions can be assumed known is described. Attention is drawn to...

Cluster randomised trial designs are growing in popularity in many clinical areas, and parallel statistical developments are numerous. When the randomised allocation applies to the clusters, the basic principles for sample size calculation still apply, although modifications are required. Cluster trials consume considerable logistical and other res...

This chapter describes single‐ and two‐stage designs for Therapeutic Exploratory (TE) clinical trials involving a single arm and where the outcome of interest is binary in nature. It discusses the single‐stage Fleming‐A'Hern design as well as two‐stage designs such as the Simon‐Optimal and the Simon‐Minimax designs. The designs by Tan‐Machin and Ma...

This chapter describes how sample sizes may be derived by pre‐specifying the width or relative width of the confidence interval (CI) the investigator wishes to obtain at the end of the study. It presents the formulae for both binary and continuous outcome measures for a single group, for the comparison between two independent groups and for paired...

Background: The prevalence of childhood obesity has been increasing but the causes are not fully understood. Recent public health interventions and guidance aiming to reduce childhood obesity have focused on the whole family, as opposed to just the child but there remains a lack of empirical evidence examining this relationship. Methods: Using d...

Background: Precision medicine is the Holy Grail of interventions that are tailored to a patient’s individual characteristics. However, the conventional design of randomized trials assumes that each individual benefits by the same amount. Methods: We reviewed parallel trials with quantitative outcomes published in 2004, 2007, 2010 and 2013. We coll...

Objectives: To assess the long-term cost-effectiveness of insulin pumps and Dose Adjustment for Normal Eating (pumps+DAFNE) compared with multiple daily insulin injections and DAFNE (MDI+DAFNE) for adults with type 1 diabetes mellitus (T1DM) in the UK. Methods: We undertook a cost-utility analysis using the Sheffield Type 1 Diabetes Policy Model...

( Lancet . 2017;389:538–546) Ultrasonography is the predominant mode of antenatal screening for fetal brain abnormalities. It has been suggested that in-utero magnetic resonance imaging (iuMRI) in conjunction with ultrasound examination could further improve diagnosis of these abnormalities. The magnetic resonance imaging (MRI) to enhance the diagn...

Introduction Suspected seizures are a common reason for admission to hospital. Rates of unscheduled admission are of interest as it may reflect the quality of local services for epilepsy. Methods We searched the Hospital Episodes Statistics (HES) in-patient database for unscheduled/emergency NHS admissions in adults in England between 1 April 2007...

Background: The Epidemiologic Study of Cystic Fibrosis using 1995-1996 and 2003-2005 data found that CF centres with lowest FEV1 tended to use fewer intravenous antibiotics. We repeated the analyses using 2013-2014 UK CF registry data to determine if this was still the case. Methods: Analysing data for 2013 and 2014 separately, 28 adult CF centr...

Background and objective: Intravenous (i.v.) antibiotics are needed for rescue when preventative therapy fails to achieve stability among adults with cystic fibrosis (CF). Understanding the distribution of i.v. days can provide insight into the care that adults with CF need. We aim to determine the baseline characteristics that are associated with...

Objectives: Adherence to nebulised treatment is typically low among people with cystic fibrosis (CF). This study sought to identify factors differentiating high or low nebuliser adherence patterns (i.e. ≥80% or <50% of all nebulised treatments over one year) among adults with CF. Design: A mixed methods cross-sectional exploratory comparison of low...