James black (jab254) viva corrections

1 parent 74734fc commit 09259760ac11d740624bc7880dd22ac0557e0e26 committed
Showing with 1,644 additions and 1,830 deletions.
1. +3 −3 Chapter_cvd/chapter_cvd.tex
2. +3 −2 Chapter_factors/chapter_factors.tex
3. Chapter_hrqol/.DS_Store
4. +2 −2 Chapter_hrqol/chapter_hrqol.tex
5. +19 −15 Chapter_med/chapter_med.tex
6. +2 −2 Chapter_modelled/chapter_modelled.tex
7. +9 −7 Chapter_source/chapter_source.tex
8. +6 −6 Chapter_traj/chapter_traj.tex
9. +6 −4 Conclusions/conclusions.tex
10. +16 −16 Introduction/introduction.tex
11. ThesisFigs/.DS_Store
13. ThesisFigs/fig_davis_wt_9.png
14. ThesisFigs/fig_lancetpapersbyhb.png
15. ThesisFigs/{fig_lancetpapersbyhb.jpg → fig_lancetpapersbyhb_old.jpg}
16. ThesisFigs/fig_prommed.png
6 Chapter_cvd/chapter_cvd.tex
 @@ -233,7 +233,7 @@ \subsection{Violation of proportional hazards} \label{sec_tvecox} The Pearson product-moment correlation ($r$) between scaled Schoenfeld residuals and ranked time suggests that the proportional hazards assumption for change in medication in the primary analysis was violated ($r$=-0.31, $\chi^2$=5.81, p=0.02). When separating CVD and diabetes medication into unique parameters, non-proportionality over time remained present for change in CVD medication only ($r$=-0.42, $\chi^2$=10.81, p\textless0.01). -\cref{fig_coxviolation} shows the variation in fit over time for the primary analysis (\cref{fig_coxviolationa}), and the change in CVD medication variable (\cref{fig_coxviolationb}) from a model where CVD medication and diabetes medication were separate parameters. If the proportional hazards assumption held, I would expect to see line with a slope close to zero. I found that in both \cref{fig_coxviolationa,fig_coxviolationb} it appeared that an additional medication was associated with an higher risk in the first year of follow up, but the hazard appeared to stabilise from the second year of follow up. To address the violation of the non-proportional hazards, two strategies were employed: +\cref{fig_coxviolation} shows the variation in fit over time for the primary analysis (\cref{fig_coxviolationa}), and the change in CVD medication variable (\cref{fig_coxviolationb}) from a model where CVD medication and diabetes medication were separate parameters. If the proportional hazards assumption held, I would expect to see line with a slope close to zero. I found that in both \cref{fig_coxviolationa,fig_coxviolationb} it appeared that an additional medication was associated with a higher risk in the first year of follow up, but the hazard appeared to stabilise from the second year of follow up. To address the violation of the non-proportional hazards, two strategies were employed: \begin{description} \item[Amended model 1) Splitting the model] into periods of proportional hazards. A subjective interpretation of \cref{fig_coxviolation} suggests that dividing the model into before and after 12 months after one year follow up would lead to hazards that are proportional over time. As this leaves inadequate power for the model investigating the first 12 months, I only present the second model looking from 12 months after one year follow up (effectively 2 years after diagnosis) in \cref{sec_coxpartition}. @@ -308,7 +308,7 @@ \subsection{Context within the literature} In \gls{steno2} 160 individuals with type 2 diabetes and microalbuminuria were randomised to intensive treatment of CVD risk factors through a step-wise application of lifestyle advice and pharmacotherapy to attain CVD risk factor targets.\supercite{Gaede2003} At 7.8 year follow up, individuals in the intensive treatment arm were 55\% (HR 0.45; 95\%CI 0.23,0.91) less likely to experience a composite CVD event.\supercite{Gaede2003} \gls{steno2} recruited a sample with long standing diabetes who were randomised to treatment strategies, while in \gls{additioncam} I have explored the relationship within screen-diagnosed individuals that changed medication, which means comparisons must be approached with caution. Under this caveat, in my analysis I found that an increase of three medications was associated with a predicted 51\% lower (relative hazard to no change 0.49; 95\%CI 0.39,0.61) risk of a composite CVD event. While \gls{steno2} did not report change in medication information in a comparable way, intensification of CVD risk factor control in a population with long standing diabetes can be assumed to involve pharmacotherapy intensification, and it suggests that my results suggest the effectiveness of pharmacotherapy within a multifactorial approach to diabetes care. -While this analysis uses a novel screen-detected diabetes population who received multi-factorial diabetes care, my findings are in keeping with previous research from \gls{rct}s of interventions that led to treatment intensification for individual CVD risk factors. The \gls{ukpds} demonstrated the benefit of improving both glucose control, particularly with metformin in overweight individuals, and lowering blood pressure levels.\supercite{UKProspectiveDiabetesStudyGroup1998,Holman2008} The \gls{hps} and \gls{cards} trials demonstrated the benefit of statin use in people with diabetes.\supercite{HeartProtectionStudy2008,Colhoun2004} While there is less consensus on the benefit of aspirin outside of individuals with previous CVD\supercite{ADA2014,NICE2010}, in \cref{chapter_medburden} (page~\pageref{chapter_medburden}) I demonstrated that in \gls{additionuk} 21\% reported aspirin at diagnosis, and this doubled (42\%) at one year. While a screen-detected population may have a lower event rate than many of the populations were mono-therapies and individual risk factor control were evaluated, there is evidence that the percentage risk reductions remain constant regardless of absolute risk.\supercite{HeartProtectionStudy2008} +While this analysis uses a novel screen-detected diabetes population who received multi-factorial diabetes care, my findings are in keeping with previous research from \gls{rct}s of interventions that led to treatment intensification for individual CVD risk factors. The \gls{ukpds} demonstrated the benefit of improving both glucose control, particularly with metformin in overweight individuals, and lowering blood pressure levels.\supercite{UKProspectiveDiabetesStudyGroup1998,Holman2008} The \gls{hps} and \gls{cards} trials demonstrated the benefit of statin use in people with diabetes.\supercite{HeartProtectionStudy2008,Colhoun2004} While there is less consensus on the benefit of aspirin outside of individuals with previous CVD\supercite{ADA2014,NICE2010}, in \cref{chapter_medburden} (page~\pageref{chapter_medburden}) I demonstrated that in \gls{additionuk} 21\% reported aspirin at diagnosis, and this doubled (42\%) at one year. While a screen-detected population may have a lower event rate than many of the populations where mono-therapies and individual risk factor control were evaluated, there is evidence that the percentage risk reductions remain constant regardless of absolute risk.\supercite{HeartProtectionStudy2008} \subsection{Strengths and limitations} @@ -320,7 +320,7 @@ \subsection{Strengths and limitations} Repeating the analysis without individuals that decreased their medication after diagnosis leads to estimates of an association centred on no effect. The binary inclusion of up or down into the model, and the raw survival rates from the Kaplan Meier plot, suggests that there is a broadly linear effect of change in medication, so available evidence suggests that the primary model is appropriate in treating change in medication as continuous. However, when the primary analysis is repeated in only those who increased medication, the result becomes insignificant (not presented is that the relationship is also non-significant in only those that decreased). While I hypothesise that the centred estimate is due to low power, I cannot rule out that the associations seen are primarily driven by individuals that decrease medication in the first year being more frail\supercite{Ismail-Beigi2011}, which in turn leads to an increased incidence of CVD events. -This analysis aims to expand on our knowledge of single risk factor therapy, and multifactorial interventions like \gls{steno2}, to the effect of changes in multiple medication types for multiple risk factors centred around the prevention of micro- and macro-vascular disease. I have used self-reported medication, which for repeat medication for chronic conditions like hypertension can closely mirror pharmacy redemptions.\supercite{Allin2013} However, it should be noted that there is uncertainty on how redeemed medication translates into actual use by the individual.\supercite{Chowdhury2013} +This analysis aims to expand on our knowledge of single risk factor therapy, and multifactorial interventions like \gls{steno2}, to the effect of changes in multiple medication types for multiple risk factors centred around the prevention of micro- and macrovascular disease. I have used self-reported medication, which for repeat medication for chronic conditions like hypertension can closely mirror pharmacy redemptions.\supercite{Allin2013} However, it should be noted that there is uncertainty on how redeemed medication translates into actual use by the individual.\supercite{Chowdhury2013} \subsection{Implications for practice}
5 Chapter_factors/chapter_factors.tex
 @@ -67,7 +67,7 @@ \subsection{Sensitivity analysis} The primary analysis was repeated in only the routine care arm, and results that conflicted are reported in the results (\cref{sec_rfsensitivity}, page~\pageref{sec_rfsensitivity}). Fixed effects meta-analyses of centre level regressions was used selected as a parsimonious primary model as I did not expect the true effects to vary across centres. The primary analysis was also repeated using a multilevel model of practices within centres. These multilevel models allowed for the possibility of the four \gls{addition} centres representing a distribution of true effects. -Potential interactions between baseline modelled CVD risk and education were explored in a a multilevel model analogous to the primary analysis, except applied to the entire sample rather than stratified by quartile of modelled risk at diagnosis. Only significant interactions are reported (\cref{sec_paper2socio}, page~\pageref{sec_paper2socio}). +Potential interactions between baseline modelled CVD risk and education were explored in a multilevel model analogous to the primary analysis, except applied to the entire sample rather than stratified by quartile of modelled risk at diagnosis. Only significant interactions are reported (\cref{sec_paper2socio}, page~\pageref{sec_paper2socio}). To explore the possibility that the observed associations were dependent on how modelled CVD risk was stratified, I produced scatter plots of change in each risk factor by baseline modelled \gls{cvd} risk. Using quartiles appeared to accurately summarise the continuous relationship between risk factors and baseline risk. Results were similar within randomisation groups, and they were combined into a single cohort with adjustment for trial group. A multilevel logistic model (practices within centres) was used to explore socio-demographic information that predicted loss to follow up. Regression to the mean within quartiles was explored by plotting baseline values against change scores.\supercite{TuChange2007} @@ -462,6 +462,7 @@ \subsection{Socio-economic patterning} \caption[Change in BMI and baseline CVD risk by education]{Hex-binned scatter plot of modelled CVD at diagnosis against change in \gls{bmi} from diagnosis to five years. Bins are coloured based on the count of observations they contain, with dark indicating few point, and yellow areas of high datapoint concentration. Line of best fit ($y=\alpha+\beta_{RS}x_{RS}+\epsilon$, RS = 10-year modelled risk score) for each education status is overlaid.} \label{fig_refinteraction} \end{figure} + \subsection{Sensitivity analyses} \label{sec_rfsensitivity} \sloppy @@ -506,7 +507,7 @@ \subsection{Strengths and limitations} The change in each risk factor appeared to be normally distributed within each quartile, and sensitivity analyses treating modelled CVD risk as a continuous measure suggested that the quartiles represented the underlying patterns in an easily interpretable manner. \subsection{Implications for practice} - + Calculation of modelled CVD risk might be a useful tool for guiding treatment decisions in newly diagnosed diabetes patients. It is also recommended for use during diabetes consultations in England.\supercite{GPguide2008} Identifying who is at highest risk will help target treatment to those who need it the most and is likely to lead to a reduction in treatment inequity.\supercite{JointBritishSocieties2005} This analysis provides a reference point for patients and their \gls{gp}s when considering what are achievable goals for changes in risk factors early in the course of the disease, accounting for the diverse cardiometabolic profile present in newly diagnosed patients. This is important as primary care is striving towards diabetes care that is ever more tailored to the patient\supercite{GPguide2008,ADA2014}, and the results presented provide realistic expectations for how risk factors will change in the first five years after diagnosis.
BIN  Chapter_hrqol/.DS_Store
Binary file not shown
4 Chapter_hrqol/chapter_hrqol.tex
 @@ -173,7 +173,7 @@ \subsection{Cohort characteristics} %---------------------------------------------Figure \begin{figure}[!p] -\centering +\centering \includegraphics[width=1.3\textwidth,angle=90]{fig_promhists.png} \caption[Distribution of change in HRQoL by change in medication]{Distribution of change in HRQoL (\gls{eq5d}, \gls{addqol} \gls{mcs}, and \gls{pcs}) from 0-1 and 1-5 years, colour coded by change in medication.} \label{fig_promhists} @@ -186,7 +186,7 @@ \subsection{Change from baseline to one year} %---------------------------------------------Table \begin{table}[!tb]\footnotesize -\caption[Change in medication and HRQoL after diabetes diagnosis]{Associations between change in number of cardio-protective agents and \gls{hrqol} in \gls{additioncam} cohort.} +\caption[Change in medication and HRQoL after diabetes diagnosis]{Associations between change in number of cardio-protective agents and \gls{hrqol} in \gls{additioncam} cohort.} \label{tab_medhrqolchange} \begin{adjustbox}{width=\linewidth} \begin{threeparttable}
34 Chapter_med/chapter_med.tex
 @@ -12,7 +12,7 @@ \section{Introduction and aims} The pharmacological treatment burden among individuals with screen-detected or recently diagnosed diabetes is unknown. Given that population screening for diabetes has been recommended by several national organisations and the \gls{nhs} currently includes assessment of risk of diabetes in its Health Checks programme\supercite{HealthChecks2009}, more individuals will be found earlier in the disease trajectory. Further, there is growing evidence for the benefit of intensive treatment of risk factors early in the course of the disease\supercite{Holman2008,griffin2011}, which suggests that screen-detected patients may be prescribed a larger number of cardio-protective drugs earlier than they might previously have been. Although there is some evidence that improved medication adherence may improve health-related quality of life in symptomatic diabetic patients\supercite{Hajos2012,Shim2012}, individuals earlier in the disease trajectory are unlikely to have symptoms and may be less likely to adhere to complex medication regimes.\supercite{Murphy1995,Miller1997} Guidelines promote a multifactorial approach to diabetes care from diagnosis that includes pharmacotherapy for multiple CVD related conditions.\supercite{NICE2002,NICE2010} Despite the increasing number of individuals with screen-detected diabetes, many of whom have comorbidities, there is an absence of knowledge about what the pharmacotherapy burden is at diagnosis in this population, and how it changes in the first five years. It is important that this is described so that patients and practitioners are informed about the likely course and burden of treatment. -\subsection{Aims} +\subsection{Aims} I aimed to (i) describe medication burden at diagnosis, one and five years in individuals with screen-detected diabetes and (ii) examine in detail if age, sex, intensive treatment, or modelled 10-year CVD risk was associated with the change in the number of medications that individuals were prescribed in the five years after diagnosis. @@ -123,7 +123,7 @@ \subsection{Change in cardio-protective medication} \label{sec_changeinmed} %%---------------------------------------------Figure %\begin{figure}[htb] %\centering -% +% % %\end{figure} @@ -281,12 +281,12 @@ \subsubsection{Total medication burden} \centering \includegraphics[width=\textwidth]{fig_additionuk.eps} \caption[Medication use from diagnosis to 5 years]{Proportion of participants prescribed medication, by agent, in \gls{additionuk} from diagnosis to 5 years.} -\label{fig_additionuk} +\label{fig_additionuk} \end{figure} \subsubsection{Diabetes-related and cardio-protective medication} - + After diagnosis, both the variety and number of cardio-protective and diabetes medications increased (\cref{fig_medboxplots} \& \cref{fig_additionuk}). At one year, 27\% of individuals were prescribed any type of diabetes medication, which increased to 62\% at five years. Between diagnosis, one and five years, the prescription of anti-hypertensive (57\% to 60\% to 77\%), lipid lowering (24\% to 57\% to 81\%) and anti-thrombotic (21\% to 43\% to 54\%) medication increased. In this screen-detected population, many individuals reported using no glucose lowering medication at one and five years (73\% and 38\%, respectively, \cref{fig_medboxplots} \& \cref{fig_additionuk}). @@ -313,24 +313,24 @@ \subsubsection{Predictors of prescribed medication at five years} & Change in CVD medication & Change in other medication \\ %\cmidrule(r){2} \cmidrule(r){3} \cmidrule(r){4} \cmidrule(r){5} - & $\beta$ (95\%CI) - & IRR\tnote{\textdagger} (95\%CI) - & $\beta$ (95\%CI) + & $\beta$ (95\%CI) + & IRR\tnote{\textdagger} (95\%CI) + & $\beta$ (95\%CI) & $\beta$ (95\%CI) \\ -\midrule +\midrule Number of medications at diagnosis\tnote{\textdaggerdbl} & \color[HTML]{FF0000}-0.49 (-0.56,-0.42) & - - & -0.50 (-0.56,-0.44) - & -0.30 (-0.37,-0.22) \\ + & \color[HTML]{FF0000}-0.50 (-0.56,-0.44) + & \color[HTML]{FF0000}-0.30 (-0.37,-0.22) \\ Male gender & -0.25 (-0.57,0.06) - & 0.86 (0.75,0.99) + & \color[HTML]{FF0000}0.86 (0.75,0.99) & -0.11 (-0.33,0.10) - & 0.12 (-0.10,0.34) \\ + & 0.12 (-0.10,0.34) \\ Intensive treatment arm - & 0.44 (0.10,0.78) - & 1.14 (1.01,1.30) + & \color[HTML]{FF0000}0.44 (0.10,0.78) + & \color[HTML]{FF0000}1.14 (1.01,1.30) & \color[HTML]{FF0000}0.39 (0.09,0.69) & -0.08 (-0.30,0.13) \\ Age at diagnosis (years) @@ -345,10 +345,12 @@ \subsubsection{Predictors of prescribed medication at five years} & 0.00 (-0.01,0.01) \\ \bottomrule \end{tabular} -} +} \begin{tablenotes} \item[\textdagger] IRR = Incidence Rate Ratio \item[\textdaggerdbl] Number of medications of the medication type that is the dependent variable in that columns regression. +\end{tablenotes} +\end{threeparttable} \end{adjustbox} \end{table} @@ -386,6 +388,8 @@ \subsection{Context within the literature} \end{figure} \FloatBarrier + +%---------------------------------------------SECTION \section{Discussion} In a population of individuals with screen-detected type 2 diabetes, I described the prevalence of diabetes-related, cardio-protective and other medications from diagnosis to five years. The majority of cardio-protective medication changes happened immediately after diagnosis, although there was a gradual increase in glucose lowering medication after the initial increase for the full five years of follow up. At diagnosis, 45\% of individuals reported being prescribed blood pressure lowering, lipid lowering, or both types of medication in \gls{addition}. Many individuals in \gls{additionuk} reported medications not related to cardio-protection before diagnosis (42\%), and this increased along with a rise in the number of diabetes-related and cardio-protective drugs. The screen-detected diabetes population had a degree of poor cardio-metabolic health, with only 34\% of the sample free of high blood pressure, high cholesterol and \gls{cvd} at diagnosis. At five years, individuals were typically prescribed six medications, including one diabetes-related medication, four cardio-protective medications, and one other medication. This suggests that there is a significant degree of multi-morbidity and polypharmacy present in individuals with screen-detected diabetes. Following diagnosis, individuals were more likely to be prescribed diabetes-related medication if they were younger, female, had a high modelled CVD and if they were randomised to the intensive treatment arm of the trial (\cref{tab_medpredict}). Higher modelled CVD risk at baseline was associated with a greater increase in cardio-protective medication, but not a increase in other medications. As recommended in national guidelines, these results suggest that the treatment of diabetes was influenced by the underlying risk of CVD.
4 Chapter_modelled/chapter_modelled.tex
16 Chapter_source/chapter_source.tex
 @@ -12,7 +12,7 @@ \chapter{Key data sources} \subsubsection{Summary} -The \gls{addition} study was a cluster \gls{rct} comparing intensive multifactorial treatment with routine care among people with screen-detected diabetes in primary care. The primary endpoint (composite \gls{cvd} event) at five years was available for 99.9\% ($\frac{3055}{3057}$) of the screen-detected participants. After a median follow up of 5.9 years, there were significant improvements in CVD risk factors (\gls{hba1c}, cholesterol and blood pressure) in both groups. The incidence of first cardiovascular event was 7.2\% (13·5 per 1000 person-years) in the intensive treatment group and 8.5\% (15·9 per 1000 person-years) in the routine care group. A small non-significant reduction in both \gls{cvd} events (HR 0.83, 95\% CI 0.65–1.05) and all-cause mortality (HR 0.91; 0.69, 1.21) was observed.\supercite{griffin2011} +The \gls{addition} study was a cluster \gls{rct} comparing intensive multifactorial treatment with routine care among people with screen-detected diabetes in primary care. The primary endpoint (composite \gls{cvd} event) at five years was available for 99.9\% ($\frac{3055}{3057}$) of the screen-detected participants. After a median follow up of 5.9 years, there were significant improvements in CVD risk factors (\gls{hba1c}, cholesterol and blood pressure) in both groups. The incidence of first cardiovascular event was 7.2\% (13·5 per 1000 person-years) in the intensive treatment group and 8.5\% (15·9 per 1000 person-years) in the routine care group. A small non-significant reduction in both \gls{cvd} events (HR 0.83, 95\% CI 0.65, 1.05) and all-cause mortality (HR 0.91; 0.69, 1.21) was observed.\supercite{griffin2011} \subsubsection{ADDITION-\emph{Europe} aims} @@ -21,7 +21,9 @@ \subsubsection{ADDITION-\emph{Europe} aims} \subsubsection{Methods used in ADDITION-\emph{Europe}} \label{sec_additionmethods} -The \gls{addition} trial protocol\supercite{Lauritzen2000} and primary outcome paper\supercite{griffin2011} have been published (Clinical Trials.Gov registration \jbcode{NCT00237549)}. \gls{addition} was a primary-care based study of screening for type 2 diabetes followed by a pragmatic cluster \gls{rct} comparing intensive multifactorial treatment with routine care in four centres (Cambridge, UK; Denmark; Leicester, UK; the Netherlands; \cref{fig_additioncentres}). Of 1312 general practices invited to participate, 379 (29\%) agreed and 343 (26\%) were independently randomised to screening plus routine care of diabetes or screening followed by intensive multifactorial treatment. Between April 2001 and December 2006, practices undertook stepwise screening of patients. Within \gls{addition}, median prevalence of known diabetes was 3.5\% (excluding Denmark, where it was unknown). Screening and treatment protocols differed by centre (\cref{tab_addcentremethods}). Specific explanations of centre level variation from the \gls{addition} methods is given in \cref{tab_addcentremethods} and in later sections for for the Danish (\cref{sec_additiondk}), UK (Cambridge \& Leicester, \cref{sec_additionuk}) and Cambridge studies (\cref{sec_additioncam}). Due to an increase in missing data for smoking status, in all primary analyses smoking status at baseline was carried forward if missing at follow up.\label{sec_smoking} +The \gls{addition} trial protocol\supercite{Lauritzen2000} and primary outcome paper\supercite{griffin2011} have been published (Clinical Trials.Gov registration \jbcode{NCT00237549)}. \gls{addition} was a primary-care based study of screening for type 2 diabetes followed by a pragmatic cluster \gls{rct} comparing intensive multifactorial treatment with routine care in four centres (Cambridge, UK; Denmark; Leicester, UK; the Netherlands; \cref{fig_additioncentres}). Of 1312 general practices invited to participate, 379 (29\%) agreed and 343 (26\%) were independently randomised to screening plus routine care of diabetes or screening followed by intensive multifactorial treatment. While participating practices in Leicester appeared to be have an deprivation profile similar to national estimates\supercite{Webb2010}, further information on whether included practices were more motivated or differed in unmeasured factors is not available. + +Between April 2001 and December 2006, practices undertook stepwise screening of patients. Within \gls{addition}, median prevalence of known diabetes was 3.5\% (excluding Denmark, where it was unknown). Screening and treatment protocols differed by centre (\cref{tab_addcentremethods}). Specific explanations of centre level variation from the \gls{addition} methods is given in \cref{tab_addcentremethods} and in later sections for the Danish (\cref{sec_additiondk}), UK (Cambridge \& Leicester, \cref{sec_additionuk}) and Cambridge studies (\cref{sec_additioncam}). Due to an increase in missing data for smoking status, in all primary analyses smoking status at baseline was carried forward if missing at follow up.\label{sec_smoking} \subsubsection{Screening in ADDITION-\emph{Europe}} @@ -47,16 +49,16 @@ \subsubsection{Screening in ADDITION-\emph{Europe}} \midrule Cambridge & Electronic medical records of patients aged 40-69 years were used to calculate the Cambridge diabetes risk score\supercite{Griffin2000score} for each individual. Individuals with a score $\geq$0.17 (those in the top 25\% of the risk distribution) were invited to a stepwise screening programme, including capillary random blood glucose, fasting blood glucose and \gls{hba1c} tests. - & Practice-based educational meetings held with family physicians and nurses to discuss treatment targets, algorithms, and lifestyle advice. Audit and feedback via follow-up practice-based meetings up to twice per year Practice staff provided with educational materials for patients. Small financial incentives given to family physicians equivalent of three 10 min consultations with a family physician and three 15 min consultations wit a nurse, per patient, per year, for 3 years. \\ + & Practice-based educational meetings held with family physicians and nurses to discuss treatment targets, algorithms, and lifestyle advice. Audit and feedback via follow-up practice-based meetings up to twice per year Practice staff provided with educational materials for patients. Small financial incentives given to family physicians equivalent of three 10 min consultations with a family physician and three 15 min consultations with a nurse, per patient, per year, for 3 years. \\ Leicester & All patients aged 40-69 years were invited to undergo an oral glucose tolerance test. - & Patients referred to structured education programme. Follow up every 2 months in the first year offered, and every four months after. Clinic staff chased up missed appointments, and financial incentives given for participating equivalent of three 10 min consultations with a family physician and three 15 min consultations wit a nurse, per patient, per year, for 3 years. \\ + & Patients referred to structured education programme. Follow up every 2 months in the first year offered, and every four months after. Clinic staff chased up missed appointments, and financial incentives given for participating equivalent of three 10 min consultations with a family physician and three 15 min consultations with a nurse, per patient, per year, for 3 years. \\ Denmark & All patients aged 40-69 years were either sent or opportunistically asked to complete a questionnaire containing the Danish Diabetes Risk Score.\supercite{Glumer2004} Patients with a score $\geq$5 were then invited for stepwise diabetes screening. - & Practice based meetings to discuss treatment targets. Follow up and feedback up to twice a year. Educational materials provided to patients. Financial incentives given to practices for participating equivalent of three 10 min consultations with a family physician and three 15 min consultations wit a nurse, per patient, per year, for 3 years. \\ + & Practice based meetings to discuss treatment targets. Follow up and feedback up to twice a year. Educational materials provided to patients. Financial incentives given to practices for participating equivalent of three 10 min consultations with a family physician and three 15 min consultations with a nurse, per patient, per year, for 3 years. \\ Netherlands & Participants aged 60-69 years were sent the symptom risk questionnaire from the Hoorn study.\supercite{HoornQ1997} Individuals that scored $\geq$4 (in the 41 practices near the study centre) or $\geq$6 (38 practices further away from the study centre) were invited to attend their practice for a diabetes screening assessment. - & Practice based meetings to discuss treatment targets. Follow up and feedback up to twice a year. Patient sent reminders if overdue for assessment. Financial incentives given to practices for participating equivalent of three 10 min consultations with a family physician and three 15 min consultations wit a nurse, per patient, per year, for 3 years. \\ + & Practice based meetings to discuss treatment targets. Follow up and feedback up to twice a year. Patient sent reminders if overdue for assessment. Financial incentives given to practices for participating equivalent of three 10 min consultations with a family physician and three 15 min consultations with a nurse, per patient, per year, for 3 years. \\ \bottomrule \end{tabular} @@ -189,7 +191,7 @@ \section{UKPDS CVD risk engine} child { node[concept] {\textbf{Albumin} }} child { node[concept] {\textbf{Systolic BP} }} child { node[concept] {\textbf{HbA$_{1C}$} }} - child { node[concept] {\textbf{$frac{Total}{HDL}$}} } + child { node[concept] {\textbf{$\frac{Total chol}{HDL}$}} } } child[concept color=orange] { node[concept] {\textbf{Smoker}}
12 Chapter_traj/chapter_traj.tex
 @@ -2,7 +2,7 @@ \section{Introduction} -In \cref{chapter_medburden} I presented information on the large increase in glucose lowering medication after screen-detected diagnosis, and the continued gradual intensification over time. Good glycaemic control in combination with management of other CVD risk factors is promoted after diagnosis of diabetes to reduce the risk of micro- and macrovascular disease\supercite{NICE2010,griffin2011}, yet little is known about who glycaemic control is patterned over time in an early detected population. +In \cref{chapter_medburden} I presented information on the large increase in glucose lowering medication after screen-detected diagnosis, and the continued gradual intensification over time. Good glycaemic control in combination with management of other CVD risk factors is promoted after diagnosis of diabetes to reduce the risk of micro- and macrovascular disease\supercite{NICE2010,griffin2011}, yet little is known about how glycaemic control is patterned over time in an early detected population. \subsection{Glycaemic control after diagnosis} @@ -60,7 +60,7 @@ \subsection{Data collection} Only the intensive treatment arm was included in this analysis. \gls{gp}s in the intensive treatment arm of \gls{additiondk} were encouraged to regularly test \gls{hba1c}, providing a a large number of measurements to explore trajectories within. The intensive treatment group family physicians were encouraged through guidelines, educational meetings, and audits with feedback to introduce a stepwise target-led drug treatment regime to reduce hyperglycaemia, hypertension and hyperlipidaemia\supercite{Echouffo-Tcheugui2009,Webb2010} based on the \gls{steno2} study.\supercite{Gaede2003} Targets included \gls{hba1c} \textless\SI{53}{\ifcc} (\textless 7.0\%), blood pressure $\leq$$\frac{135}{85}$ mmHg, cholesterol \textless \SI{5}{\lipid} without ischaemic heart disease or \textless \SI{4.5}{\lipid} with ischaemic heart disease, prescription of aspirin to those treated with anti-hypertensive medication and prescription of a statin to all patients with a cholesterol level $\geq$ \SI{3.5}{\lipid} within four weeks of the diagnosis of diabetes. -Redeemed prescriptions were collected via linkage to the Danish National Prescription Registry, which has complete coverage of all redeemed prescriptions in Denmark since 1994 (see \cref{sec_dkmed} on page~\pageref{sec_dkmed} for details).\supercite{Juul2009} Information on \gls{hba1c} data collection is in \cref{sec_dkhb} on page~\pageref{sec_dkhb}. +Redeemed prescriptions were collected via linkage to the Danish National Prescription Registry, which has complete coverage of all redeemed prescriptions in Denmark since 1994 (see \cref{sec_dkmed} on page~\pageref{sec_dkmed} for details).\supercite{Juul2009} Information on \gls{hba1c} data collection is in \cref{sec_dkhb} on page~\pageref{sec_dkhb}. Baseline characteristics of the population are presented for the entire cohort and for each of the identified trajectory groups. Patients saw their family physicians every three months for the first year, and then every six months. I also report the percentage of individuals that redeemed a prescription for any glucose lowering, lipid lowering, anti-hypertensive or anti-thrombotic medication in the previous 90 days at each measurement time point. @@ -212,7 +212,7 @@ \subsubsection{Follow up truncation} \caption[Source of HbA\textsubscript{1C} values in ADDITION-\emph{Denmark}]{Source of \gls{hba1c} values from \gls{additiondk} used in this analysis. If available, lab reported \gls{hba1c} was used. CRF = Case Report Form completed by doctors, lab results collected from database.\label{fig_trajsource}} \vspace*{1cm} \includegraphics[width=\textwidth]{fig_histogramlasthb.png} -\caption[When individuals provided their last HbA\textsubscript{1C} measurement]{Histograph showing when individuals provided their last \gls{hba1c} measurement, truncated at the 2,000 day cut off.\label{fig_histogramlasthb}} +\caption[When individuals provided their last HbA\textsubscript{1C} measurement]{Histograph showing when individuals provided their last \gls{hba1c} measurement, truncated at the 2,000 day cut off.\label{fig_histogramlasthb}} \end{figure} While lab reported \gls{hba1c} was the primary source (as explained in \cref{sec_dkhb} on page~\pageref{sec_dkhb}), for some measurements only \gls{hba1c} variables collected by the practice were available (\cref{fig_trajsource}). I was unable to identify why some measurements were entered by \gls{gp}s, but not submitted by the laboratory. As the values should theoretically be identical, when laboratory measurements were missing values were collected from the study case report form completed in the practice. \cref{fig_trajsource} shows the number of measurements that came from either source, or were missing, at each time point. @@ -367,15 +367,15 @@ \section{Discussion} %---------------------------------------------SECTION \section{Discussion} - + In this prospective cohort of individuals with screen-detected diabetes, I identified four clinically distinct trajectories of glycaemic control. The majority of individuals (87.5\%) had slightly elevated \gls{hba1c} at diagnosis and were able to maintain good glycaemic control over the following five years (the low-low trajectory, \sparklowlow). A small proportion had high levels of blood glucose at diagnosis that improved over five years (high-med \sparkhighmed, 2.1\%). The final two sub-sets of the sample had similar high levels of glycaemia at diagnosis. One group was able to attain and maintain glycaemic targets (med-low \sparkmedlow, 8.2\%), while the other initially attained good control, but then deteriorated over five years (med-high \sparkmedhigh, 2.3\%). The majority of individuals (96\%) experienced a glucose trajectory that was predominately below an \gls{hba1c} threshold of 8\% from three months after diagnosis. While power was limited due to the small size of divergent trajectories, traditional risk factors and medication choices at diagnosis did not explain why some individuals diverge into less ideal glycaemic trajectories. -\subsection{Context within the literature} +\subsection{Context within the literature} The majority of studies exploring changes in glycaemia in diabetes assume a single latent process, variations from which can be adjusted for using collected covariates. This analysis does not rely on a single underlying trajectory, and instead allows the observed trajectories to be clustered into similar groups. This novel approach prevents direct comparisons with the literature to date. Soon after diabetes diagnosis, an individual’s glucose levels typically reduce as a result of changes in lifestyle behaviour and/or treatment with hypoglycaemic medication.\supercite{Holman2008,Patel2008} In the \gls{ukpds}, after an initial decrease, \gls{hba1c} values increased gradually over the next 15 years.\supercite{Prospective1998} This contrasts with the maintenance of good glycaemic control experienced by the majority of individuals in \gls{additiondk}, and is likely related to both the earlier diagnosis and temporal shifts in what constitutes best practice in diabetes care. The \gls{ukpds} recruited between 1977 and 1991, while the \gls{additiondk} recruited from 2001 to 2004. In addition to \gls{additiondk} participants being screen-detected, rather than recently diagnosed individuals, the treatment protocol applied (\cref{tab_addcentremethods}, page~\pageref{tab_addcentremethods}) suggests that participants in this analysis experienced a level of care close to contemporary guidelines.\supercite{Danish1999} Long term post trial follow up of the \gls{ukpds} showed that the gradual deterioration in glyceamic control had begun to plateau in the first five years after the trial intervention ended for both treatment groups (1997 to 2002).\supercite{Holman2008} - + \subsection{Medication and glycaemic control} Elevated values of CVD risk factors have been associated with both a higher probability of being prescribed glucose-lowering medicine as well as greater decreases in \gls{hba1c} after diagnosis in \gls{addition}.\supercite{Black2014BJGP} In all four identified trajectories \gls{hba1c} values decreased immediately after diagnosis, with larger decreases present in those with a higher baseline \gls{hba1c}. Of particular concern were the small subset (n=21) of individuals that had gradually increasing \gls{hba1c} levels after an initial decrease after diagnosis. Differentiating the characteristics of individuals that had a poor long term \gls{hba1c} from those that had similar baseline \gls{hba1c} measurements at diagnosis but stable trajectories was difficult due to the low number of individuals. Individuals with a suboptimal trajectory appeared to be more likely to receive sulphonylureas, but otherwise their prescription redemption history was very similar to individuals with a preferred trajectory. Khunti et al, in a sample of 81,573 individuals with diabetes, demonstrated that increases in the prescription of glucose lowering drugs can lag more than seven years behind changes in \gls{hba1c} values.\supercite{Khunti2013} Information on potential external reasons for not intensifying treatment is unavailable for this analysis. The trajectory that experienced the worse glycaemic trajectory after diagnosis (med-high, \sparkmedhigh) appears to experience a delay in treatment intensification. The presence of clinical inertia provides one potential explanation for this lag, although other plausible explanations include potential variation in pharmacogenetics, adherence and body composition to list a few.\supercite{Cantrell2010}
10 Conclusions/conclusions.tex
 @@ -107,7 +107,7 @@ \subsection{Individual variation} Guidelines promote individualised care, yet setting treatment goals as a collaboration between \gls{gp} and patient requires the succinct translation of the relationship between the potential harm and benefit of medications and individual patient characteristics, which can contribute to a large variation in target attainment. I demonstrated that individuals with cardiometabolic risk factors close to guideline recommended values are usually able to maintain low values, while those with poor cardiometabolic health have a larger achievable change, and a larger variation in the subsequent change in their CVD risk factors over the five years after diagnosis. I identified four clusters of \gls{hba1c} trajectories in \gls{additiondk}, and 87\% of individuals were allocated to a trajectory group that had an \gls{hba1c} centred on \SI{46}{\ifcc} (SD 9; 6.8\%, SD 1.5) at diagnosis, and remained low for the following five years (\sparklowlow). Two of the remaining trajectories were predominately patterned by their initial \gls{hba1c} levels being higher (\sparkmedlow and \sparkhighmed). The remaining cluster followed a divergent trajectory (\sparkmedhigh), that was not able to be well characterised due to its rarity. This is a direct contradiction to the \gls{ukpds}, where the divergent trajectory was the norm.\supercite{UKProspectiveDiabetesStudyGroup1995} -The term shared decision making' has entered current guidelines\supercite{ADA2014,NICE2010}, and embodies the aim of general practice to empower patients to be able to make informed decisions. Conflicting with this aspiration of how care should be delivered - a recent survey commissioned by the British Medical Association found that 54\% of \gls{gp}s felt their current workload was \emph{unmanageable or unsustainable}''\supercite{BMAgps2014} and the average consultation in 2006/07 was 11.7 minutes for \gls{gp}s and 15.5 minutes for practice nurses and nurse practitioners.\supercite{RoyalCollegeofGeneralPractitioners2013} While peripatetic clinics incorporating lay educators are able to defer some of the care burden from the practice\supercite{Carey2014}, communicating the complex relationship between the benefits and potential harms of intensifying medication at diagnosis and subsequent visits as the individual ages and the disease progresses is a difficult task. The simplified descriptions of how medication (\cref{chapter_medburden}), glycaemic control (\cref{chapter_hrqol}) and CVD risk factors (\cref{chapter_factors}) change, as well as what the potential impact of intensification is (\cref{chapter_modelled,chapter_cox,chapter_hrqol}) I present, will aid in describing the expected prognosis and relative benefits of different treatment strategies. +The term shared decision making' has entered current guidelines\supercite{ADA2014,NICE2010}, and embodies the aim of general practice to empower patients to be able to make informed decisions. Conflicting with this aspiration of how care should be delivered - a recent survey commissioned by the British Medical Association found that 54\% of \gls{gp}s felt their current workload was \emph{unmanageable or unsustainable}''\supercite{BMAgps2014} and the average consultation in 2006/07 was 11.7 minutes for \gls{gp}s and 15.5 minutes for practice nurses and nurse practitioners.\supercite{RoyalCollegeofGeneralPractitioners2013} While peripatetic clinics incorporating lay educators are able to defer some of the care burden from the practice\supercite{Carey2014}, communicating the complex relationship between the benefits and potential harms of intensifying medication at diagnosis and subsequent visits as the individual ages and the disease progresses is a difficult task. The simplified descriptions of how medication (\cref{chapter_medburden}), glycaemic control (\cref{chapter_traj}) and CVD risk factors (\cref{chapter_factors}) change, as well as what the potential impact of intensification is (\cref{chapter_modelled,chapter_cox,chapter_hrqol}) I present, will aid in describing the expected prognosis and relative benefits of different treatment strategies. \section{Limitations} \label{sec_limitations} @@ -115,11 +115,11 @@ \section{Limitations} \label{sec_limitations} Modelling work using \gls{addition} suggests that the combination of screening and early intervention is beneficial over routine diagnosis and delayed treatment initiation.\supercite{Herman2015} While this finding supports my conclusions, a more granular breakdown of the benefits of early detection vs. intensive treatment after detection suggests that the majority of this finding is driven by the early diagnosis, and not the intensity to which CVD risk factors are lowered after diagnosis.\supercite{Herman2015} This thesis supports a small protective effect of intensive treatment from early diagnosis, but the possibility exists that simply the knowledge that an individual has the diagnosis is sufficient to enable routine care to successfully manage CVD risk. This potential is strong in \gls{addition}, as the difference between routine care and the intensive treatment protocol decreased while the study recruitment was ongoing.\supercite{GPguidelinesDK2002,ADA2000,NICE2002bplipid,NICE2002} -The effect of treatment intensification ' is a reoccurring exposure in this thesis that attempts to capture the total pharmacotherapy burden of cardioprotective medication that spans multiple medication classes. In \cref{chapter_hrqol,chapter_cox,chapter_medburden} I used the number of classes as the primary measure of treatment intensification '. This method does not account for the different adverse effects that are specific to each type of medication. The role of individual medications has been addressed by seminal \gls{rct}s\supercite{UKProspectiveDiabetesStudyGroup1998,Prospective1998,Group1998,Kelly2009,Colhoun2004,HOPE2000}, but a diagnosis of diabetes leads to an uptake in multiple medications (\cref{chapter_medburden}). Alternatively, a count of pills rather than agents would reflect what the individual with diabetes experiences, but then there is less connection between the medication and underlying burden of active ingredients as doses can vary and a medication split across two pills or multiple doses a day. Where possible, I also conducted multiple sensitivity analyses, which included looking at associations by medication class. There was often a degree of concordance when uptake in one medication class (e.g. $\beta$ blockers) was correlated with uptake of another (e.g. statins) which led to increased variance and attenuation in mutually adjusted models making it difficult to detect independent effects. This though is not necessary a limitation, as my analysis remains focused on the total treatment intensification' experienced after diabetes diagnosis. +The effect of treatment intensification' is a reoccurring exposure in this thesis that attempts to capture the total pharmacotherapy burden of cardioprotective medication that spans multiple medication classes. In \cref{chapter_hrqol,chapter_cox,chapter_medburden} I used the number of classes as the primary measure of treatment intensification'. This method does not account for the different adverse effects that are specific to each type of medication. The role of individual medications has been addressed by seminal \gls{rct}s\supercite{UKProspectiveDiabetesStudyGroup1998,Prospective1998,Group1998,Kelly2009,Colhoun2004,HOPE2000}, but a diagnosis of diabetes leads to an uptake in multiple medications (\cref{chapter_medburden}). Alternatively, a count of pills rather than agents would reflect what the individual with diabetes experiences, but then there is less connection between the medication and underlying burden of active ingredients as doses can vary and a medication split across two pills or multiple doses a day. Where possible, I also conducted multiple sensitivity analyses, which included looking at associations by medication class. There was often a degree of concordance when uptake in one medication class (e.g. $\beta$ blockers) was correlated with uptake of another (e.g. statins) which led to increased variance and attenuation in mutually adjusted models making it difficult to detect independent effects. This though is not necessary a limitation, as my analysis remains focused on the total treatment intensification' experienced after diabetes diagnosis. A more technical limitation is that a feedback mechanism exists in pharmacotherapy for CVD risk factors, particularly for glucose lowering medication.\supercite{Patorno2014} This is because pharmacotherapy will be reviewed at each consultation. Within \gls{addition} medication change was primarily intensification of medication as diabetes progresses, but the possibility exists for medication to be decreased due to external factors like frailty, or improvements in lifestyle allowing medication to be discontinued.\supercite{Ismail-Beigi2011} I have attempted to address this by taking changes from diagnosis to one year to represent the period in which treatment strategies are tested and refined. The one year time frame, while conveniently reflecting a patients one year review, was dictated by the available data in \gls{addition}. More detailed information on continuous medication changes, and whether stability of medication regimes or time spent on different regimes influenced outcomes, alongside information on diet, physical activity, and general frailty would provide a much greater level of detail on the role of medication as part of a multifactorial therapy. -\gls{addition} participants were predominantly of white ethnic origin.\supercite{griffin2011} While the Leicester centre was expected to recruit 30\% of it's sample from the British South Asian\supercite{Webb2010}, it remains difficult to generalise the results of \gls{addition} to minorities. Beyond ethnicity, \gls{addition} participants were drawn from a large population-based sample, and biological data and self-reported characteristics were collected using standardised protocols and questionnaires. +\gls{addition} participants were predominantly of white ethnic origin.\supercite{griffin2011} While the Leicester centre was expected to recruit 30\% of it's sample from the British South Asian\supercite{Webb2010} community, it remains difficult to generalise the results of \gls{addition} to minorities. Beyond ethnicity, \gls{addition} participants were drawn from a large population-based sample, and biological data and self-reported characteristics were collected using standardised protocols and questionnaires. %%---------------------------------------------Figure %\begin{figure}[htb] @@ -145,13 +145,15 @@ \section{Areas for future research} \label{sec_future} Questions also remain over the interaction between the patient and \gls{gp} that led to treatment decisions. While I did not explore \gls{gp} level variation in prescribing patterns, in \gls{additiondk} variation in lipid lowering rates at the \gls{gp} level has been linked to differing risk of all-cause mortality. Further, decisions to not set aggressive CVD risk factor goals, regardless of potential increases in risk of CVD events, may in fact of been informed and valid decisions made as part of a shared decision with the \gls{gp}. The ongoing \emph{Introdia} (\url{introdia.com}) is an example of a study attempting to understand the complex relationship interaction that leads to treatment decisions, and will provide more information on how much of the clinical inertia and apparent lack of intensity in prescribing is in fact a deficiency in care. -Stand along screening and management of diabetes does not reflect the shared risk factors diabetes has with CVD and other conditions like kidney disease. Future studies of the effectiveness of combined programs like the NHS Health Checks are likely to create a more accurate picture of the cost of both screening and intensive treatment than standalone tests of each of these related conditions. +Stand alone screening and management of diabetes does not reflect the shared risk factors diabetes has with CVD and other conditions like kidney disease. Future studies of the effectiveness of combined programs like the NHS Health Checks are likely to create a more accurate picture of the cost of both screening and intensive treatment than standalone tests of each of these related conditions. %---------------------------------------------SECTION \section{Implications} \label{sec_implications} The benefits of intensive therapy will not only be limited to macrovascular disease, which is what I have focused on. The relative contribution of glycaemic control compared to other CVD risk factors in preventing microvascular disease is stronger\supercite{Laakso2012}, which is reflected in the literature supporting intensive glucose control to prevent microvascular disease. As such, the case for intensive treatment is likely to be stronger when the entire burden of complications is addressed. +This thesis has expanded our knowledge of what an early diagnosed population looks like, and how the disease is likely to progress. I have shown that there is likely to be a reduction in \gls{cvd} events over a 15-year period after early diagnosis if an individual is treated intensively compared to routine care, but questions remain over whether this remains a significant benefit when adjusting for the greater resource costs required to provide both the medication, and the monitoring for elevated risk factors, and how the effect of further improvements in routine care will decrease the difference between intensive treatment and routine care. As such this thesis does does not provide strong evidence in favour of screening for diabetes, but does suggest that further research is needed to discern whether disease progression is delayed through earlier detection. + The implications for this thesis are derived from a pragmatic promotion of intensive care filtered through \gls{gp}s, and associations derived from individuals that did increase their medication count. This is an important distinction, as the \gls{gp} helps ensure that treatment decisions are appropriate to the individual (e.g. if they are frail\supercite{Ismail-Beigi2011}), potentially ensuring a \gls{hrqol} burden is averted. Type 2 diabetes is being diagnosed earlier in the disease trajectory, while most of our information on how to treat it is based on populations much further along the disease trajectory. My research suggests that intensification of treatment in an early diagnosed population is protective for CVD, and does not lead to an excess \gls{hrqol} burden.
32 Introduction/introduction.tex
 @@ -10,14 +10,14 @@ \chapter{Introduction} \section{Diabetes} -The clustering of excessive thirst and sweet, honey-like, urine had been known throughout antiquity.\supercite{Polonksy2012} Yet it was Thomas Willis (1621-75), in his \emph{Pharmaceutice rationalis}, who defined the modern condition of diabetes mellitus. In 1777, Matthew Dobson (1732-84) proved conclusively that people with diabetes had elevated levels of sugar in their urine and blood. The first appearance of diabetes as a condition in the \emph{New England Journal of Medicine and Surgery} was not until 1812, and at the time the clinical definition was restricted to the then fatal condition we now call type 1 diabetes.\supercite{Polonksy2012} In 1889, it was discovered that removing the pancreas of dogs led to type 1 diabetes and death, and in 1910 it was first hypothesised that diabetes was related to a single chemical, newly named insulin, that was secreted from within the clusters of cells in the pancreas called the islets of Langerhorns.\supercite{Polonksy2012} A decade later, in 1922, bovine insulin was successfully administered to young Leonard Thompson and type 1 diabetes no longer meant a painful death before adulthood.\supercite{Bliss1993} Banting and Macleod received the Nobel Prize for developing bovine insulin, and between their award in 1923 and 1992 a flurry of research resulted in ten scientists receiving Nobel Prizes for diabetes related research. +The clustering of excessive thirst and sweet, honey-like, urine had been known throughout antiquity.\supercite{Polonksy2012} Yet it was Thomas Willis (1621-75), in his \emph{Pharmaceutice rationalis}, who defined the modern condition of diabetes mellitus. In 1777, Matthew Dobson (1732-84) proved conclusively that people with diabetes had elevated levels of sugar in their urine and blood. The first appearance of diabetes as a condition in the \emph{New England Journal of Medicine and Surgery} was not until 1812, and at the time the clinical definition was restricted to the then fatal condition we now call type 1 diabetes.\supercite{Polonksy2012} In 1889, it was discovered that removing the pancreas of dogs led to type 1 diabetes and death, and in 1910 it was first hypothesised that diabetes was related to a single chemical, newly named insulin, that was secreted from within the clusters of cells in the pancreas called the islets of Langerhan's.\supercite{Polonksy2012} A decade later, in 1922, bovine insulin was successfully administered to young Leonard Thompson and type 1 diabetes no longer meant a painful death before adulthood.\supercite{Bliss1993} Banting and Macleod received the Nobel Prize for developing bovine insulin, and between their award in 1923 and 1992 a flurry of research resulted in ten scientists receiving Nobel Prizes for diabetes related research. \subsection{Subsets of diabetes} \label{sec_typeofdiabetes} Diabetes is not a single condition, but a heterogenous clustering of clinical conditions with overlapping mechanisms and complications. The \gls{ada} divides diabetes into the following four clinical categories\supercite{ADA2014}: \begin{description} \item[Type 1 diabetes] usually has an early onset, and is characterised by an autoimmune response which leads to destruction of the pancreatic $\beta$ cells present in the islets of Langerhorns. The lack of insulin leads to an absolute deficiency, and without an external source of insulin, eventually death. As the age of onset is not a definitive means to identify type 1 diabetes, differentiation can be improved by testing for antibodies specific to islet cells and insulin that would indicate an immune response has initiated.\supercite{Winter2011} Additionally, levels of C-peptide, a molecule produced in tandem with insulin, can indicate whether insulin secretion has been interrupted.\supercite{Jones2013} - \item[Gestational diabetes] which is transient and related to pregnancy. + \item[Gestational diabetes] which is mostly transient and related to pregnancy. \item[Other types] of diabetes \emph{with specific causes} not contained within the other clusters. Examples are diabetes related to other conditions and/or their treatment (e.g. pancreatic diseases) or specific and established genetic defects in insulin production or function. \item[Type 2 diabetes,] which is the focus of this thesis. It is the most common form of diabetes and relates to varying degrees of defect in insulin secretion and resistance. Throughout the remaining thesis, diabetes and type 2 diabetes will be used interchangeably. \end{description} @@ -30,7 +30,7 @@ \subsection{Definition and diagnosis of type 2 diabetes} Type 2 diabetes is characterised by defects in insulin signalling and/or secretion that lead to metabolic imbalances.\supercite{WHO1999} Insulin, a hormone produced in the $\beta$ cells of the pancreas, promotes the uptake and storage of glucose, and inhibits the release of glucagon, a peptide that promotes the release of glucose into the blood stream. In healthy individuals, insulin and glucagon form part of a feedback mechanism that regulates fat and carbohydrate metabolism, keeping blood glucose levels high enough to provide the necessary cellular fuel, but low enough to not be toxic.\supercite{Reece2011} In individuals with diabetes the body is unable to respond to glycaemic challenges appropriately, although how much of this is caused by insulin secretion or insulin function differs amongst individuals.\supercite{WHO1999} -The 2006 \gls{who} diagnostic criteria for diabetes was a \gls{fpg} of $\geq$7.0 mmomL\textsuperscript{-1} or a 2 hour 75g glucose load \gls{ogtt} of $\geq$11.1 mmomL\textsuperscript{-1}.\supercite{WHO2006} In 2011 the WHO also recommended the use of \gls{hba1c}, at a threshold of \SI{48}{\ifcc} ($\geq$6.5\%).\supercite{WorldHealthOrganization2011} The 2014 \gls{ada} guidelines include the \gls{who} definition, and expand it to include a random plasma glucose of $\geq$11.1 mmomL\textsuperscript{-1} in the presence of the classic symptoms of hyperglyceamia.\supercite{ADA2014} +The 2006 \gls{who} diagnostic criteria for diabetes was a \gls{fpg} of $\geq$7.0 mmomL\textsuperscript{-1} or a 2 hour 75g glucose load \gls{ogtt} of $\geq$11.1 mmomL\textsuperscript{-1}.\supercite{WHO2006} In 2011 the WHO also recommended the use of \gls{hba1c}, at a threshold of \SI{48}{\ifcc} ($\geq$6.5\%).\supercite{WorldHealthOrganization2011} The 2014 \gls{ada} further expanded the guidelines to include a random plasma glucose of $\geq$11.1 mmomL\textsuperscript{-1} in the presence of the classic symptoms of diabetes.\supercite{ADA2014} As a diagnosis of diabetes may lead to changes in the individuals lifestyle and mental state\supercite{Ebarol2007,Peel2004,Parry2004}, that in the American context could extend as far dramatic increases in insurance premiums, epidemiological definitions of diabetes can allow for a higher rate of false positives than clinical definitions.\supercite{WHO1999} While a clinical diagnosis requires confirmation of testing (in the absence of symptoms that suggest hyperglycaemia)\supercite{ADA2014}, epidemiological studies routinely use a single abnormal blood glucose level to define a case of diabetes. @@ -53,7 +53,7 @@ \subsection{Definition and diagnosis of type 2 diabetes} \caption{Coronary Heart Disease and glycaemia.} \label{fig_sarwar2010} \end{subfigure} -\caption[Where glycaemia becomes diabetes]{The left figure (a) is taken from \emph{McCance et al(1994)BMJ,308(6940)1323-8} and shows an early study that suggested there was a threshold effect of glycaemia and microvascular disease. The right figure (b), taken from \emph{Sarwar et al(2011)NEJM,364(9)829-41}, published 17 years later, shows a more contemporary observation of a linear relationship between glycaemia and macrovascular disease. The reference group is 5-5.5 mmolL\textsuperscript{-1}. Figures are reproduced with permission.} +\caption[Where glycaemia becomes diabetes]{The left figure (a) is taken from \emph{McCance et al(1994)BMJ,308(6940)1323-8} and shows an early study in Pima Indians that suggested there was a threshold effect of glycaemia and microvascular disease. The right figure (b), taken from \emph{Sarwar et al(2011)NEJM,364(9)829-41}, published 17 years later, shows a more contemporary observation of a linear relationship between glycaemia and macrovascular disease. The reference group is 5-5.5 mmolL\textsuperscript{-1}. Figures are reproduced with permission.} \label{fig_mccancesarwar} \end{figure} @@ -76,7 +76,7 @@ \subsection{Risk factors and pathogenesis} Risk factors for complications of diabetes are the primary concern for this thesis, so I will only briefly touch on the the risk factors for type 2 diabetes. While genetic variants increasing the risk of type 2 diabetes have been identified\supercite{Sladek2007}, known variants only account for around 10\% of the heritability, and the presence of known genetic variants in an individual only increases risk by a similar percentage.\supercite{Polonksy2012} While current knowledge about genetic determinants is limited\supercite{Lyssenko2008}, much stronger associations are seen with modifiable risk factors like body weight.\supercite{Polonksy2012,Vistisen2014} -Modifiable risk factors for diabetes include diet, physical activity and smoking behaviour, and exposure to these risk factors can in in turn contribute to unfavourable increases in other risk factors like cardiometabolic health and obesity.\supercite{Alberti2007} Overweight/obesity, central adiposity, elevated triglycerides, low HDL and high LDL cholesterol and elevated blood pressure are all associated with increased risk of type 2 diabetes, and the presence of these risk factors tend to cluster together in individuals.\supercite{Haffner1998,Alberti2007} Non-modifiable factors include increasing age, ethnicity and a family history of diabetes.\supercite{Haffner1998} +Modifiable risk factors for diabetes include diet, physical activity and smoking behaviour, and exposure to these risk factors can in in turn contribute to unfavourable increases in other risk factors like cardiometabolic health and obesity.\supercite{Alberti2007} Overweight/obesity, central adiposity, elevated triglycerides, low HDL cholesterol and elevated blood pressure are all associated with increased risk of type 2 diabetes, and the presence of these risk factors tend to cluster together in individuals.\supercite{Haffner1998,Alberti2007} Non-modifiable factors include increasing age, ethnicity and a family history of diabetes.\supercite{Haffner1998} Emerging risk factors include traffic and industrial environmental pollution\supercite{Rajagopalan2012}, although teasing out causation is difficult as exposure is related to social inequalities, and can be assumed to have a long cumulative effect. Subclinical inflammation is increasingly being suggested as process of diabetes pathogenesis.\supercite{Tamayo2014,Crook2004,Donath2011} The mechanisms for inflammation are not well explained\supercite{Crook2004}, but the excess adipose tissue present in obesity often leads to chronic inflammation.\supercite{Yudkin2007} This hypothesis over a role of inflammation is supported by evidence that the molecules associated with subclinical inflammation are further elevated in South Asians with obesity\supercite{Crook1998}, who are known to be at a higher risk of type 2 diabetes. @@ -97,7 +97,7 @@ \subsection{Increasing importance of type 2 diabetes} \subsection{Secular trends and predictions} -Type 2 diabetes has emerged as a major threat to global health.\supercite{Bonow2004,Zimmet2010Nature} The global prevalence of all diabetes, among adults aged 20-79 years, was estimated to be 8.3\% in 2011, and is expected to increase to 9.9\% by 2030.\supercite{Whiting2011} From 1990 to 2010, diabetes moved from the 21\textsuperscript{st} to the 14\textsuperscript{th} ranked cause of lost \gls{dalys}, and from 15\textsuperscript{th} to 9\textsuperscript{th} in terms of cause of death, according to a 2010 \gls{gbd} study.\supercite{Murray2013,Lozano2012} \cref{fig_diabetesatlas}, a map reproduced from the \gls{idf} Diabetes Atlas\supercite{IDF2013}, shows the millions of people estimated to be living with diabetes in each world region. These estimates include individuals that are currently undiagnosed. The black lines, which represent the undiagnosed proportion, highlight that in some heavily populated regions of the world the majority of individuals with diabetes are not clinically diagnosed. There is a great variety of sources that contribute to these global burden estimates. In Scotland, accurate information on the prevalence and trends of diabetes prevalence is available from the \gls{scidc}, a nation-wide network of diabetes care providers. In Tunisia, estimates of the prevalence of diabetes relies on identifying and testing a sample of the population, and then inferring from this sample what the national prevalence is.\supercite{BenRomdhane2014} +Type 2 diabetes has emerged as a major threat to global health.\supercite{Bonow2004,Zimmet2010Nature} The global prevalence of all diabetes, among adults aged 20-79 years, was estimated to be 8.3\% in 2011, and is expected to increase to 9.9\% by 2030.\supercite{Whiting2011} From 1990 to 2010, diabetes moved from the 21\textsuperscript{st} to the 14\textsuperscript{th} ranked cause of lost \gls{dalys}, and from 15\textsuperscript{th} to 9\textsuperscript{th} in terms of cause of death, according to a 2010 \gls{gbd} study.\supercite{Murray2013,Lozano2012} \cref{fig_diabetesatlas}, a map reproduced from the \gls{idf} Diabetes Atlas\supercite{IDF2013}, shows the millions of people estimated to be living with diabetes in each world region. These estimates include individuals that are currently undiagnosed. The black lines, which represent the undiagnosed proportion, highlight that in some heavily populated regions of the world the majority of individuals with diabetes are not clinically diagnosed. There is a great variety of sources that contribute to these global burden estimates. In Scotland, accurate information on the prevalence and trends of diabetes prevalence is available from the \gls{scidc}\supercite{NHSScotland2002}, a nation-wide network of diabetes care providers. In Tunisia, estimates of the prevalence of diabetes relies on identifying and testing a sample of the population, and then inferring from this sample what the national prevalence is.\supercite{BenRomdhane2014} In 2013 the European prevalence of type 2 diabetes was estimated as 8.5\%.\supercite{Tamayo2014} While variance in diabetes prevalence is seen across social gradients and ethnicities within countries, variance across European nations is also high, ranging from 2.4\% in Moldova to 14.9\% in Turkey.\supercite{Tamayo2014} An example of the complexities in diabetes prevalence is seen in Germany, where the less economically developed north-east is estimated to have both higher rates of diagnosed diabetes and undiagnosed diabetes than the more developed southern border regions.\supercite{Tamayo2014a} @@ -117,11 +117,11 @@ \subsection{Economic burden of diabetes} \subsection{Quality of life burden of diabetes} -Within the Hoorn study, a comparison can be made between 116 individuals with screen-detected diabetes, and 49 with clinically diagnosed diabetes at \jbtilde2 weeks after diagnosis.\supercite{Adriaanse2004} Screen detected individuals were more likely be overweight (BMI$\geq$25; 89\% vs 73\%; p=0.01) but less likely to be hypertensive (75\% vs. 59\%; p=0.04), prescribed oral glucose lowering medication (24\% vs. 78\%; p\textless0.01) or anti-depressives (0 vs 6\%; p=0.03).\supercite{Adriaanse2004} Compared to the the clinically diagnosed arm, the screen detected sample had preferable \gls{hrqol} when measured by the \gls{mcs} (mean 54, SD 9; vs mean 49, SD 12; p=0.01) and the general well-being item of the W-BQ12 (mean 28, SD 7; mean 25, SD 7).\supercite{Adriaanse2004} These estimates represent an unadjusted cross-sectional profile of screened vs. routinely diagnosed populations, which does not account for different characteristics. The measure was also only available \jbtilde2 weeks after diagnosis, so no information is given on the impact of being diagnosed via the two methods on \gls{hrqol}. +Within the Hoorn study, a comparison can be made between 116 individuals with screen-detected diabetes, and 49 with clinically diagnosed diabetes at \jbtilde2 weeks after diagnosis.\supercite{Adriaanse2004} Screen detected individuals were more likely be overweight (BMI$\geq$25; 89\% vs 73\%; p=0.01) hypertensive (75\% vs. 59\%; p=0.04), prescribed oral glucose lowering medication (24\% vs. 78\%; p\textless0.01) or anti-depressives (0 vs 6\%; p=0.03).\supercite{Adriaanse2004} Compared to the the clinically diagnosed arm, the screen detected sample had preferable \gls{hrqol} when measured by the \gls{mcs} (mean 54, SD 9; vs mean 49, SD 12; p=0.01) and the general well-being item of the W-BQ12 (mean 28, SD 7; mean 25, SD 7).\supercite{Adriaanse2004} These estimates represent an unadjusted cross-sectional profile of screened vs. routinely diagnosed populations, which does not account for different characteristics. The measure was also only available \jbtilde2 weeks after diagnosis, so no information is given on the impact of being diagnosed via the two methods on \gls{hrqol}. % UKPDS was apparently wrong? Check this reference. Bradley C. Importance of differentiating health status from quality of life. Lancet 2001; 357: 7–8. -The \gls{accord} trial included aggressive glycaemic control targets of \SI{42}{\ifcc} (\textless6\%) in the intensive arm, and 53-\SI{63}{\ifcc} (7-7.9\%) in the routine care arm, and individuals recruited had long standing diabetes and evidence of CVD. Change in \gls{hrqol} was measured by the \gls{sf36}, \gls{dsc}, \gls{dtsq} and \gls{phq9}. After randomisation, individuals receiving intensive treatment reported a larger decrease in \gls{pcs} and perceived hypoglycaemia (\gls{dtsq} item), but greater treatment satisfaction (\gls{dtsq} scale) and less hyperglycaemia (\gls{dtsq} item). The \gls{accord} researchers believed that the statistically significant change in the \gls{pcs} was \emph{trivial}', and that there was no clinically significant impact of intensive treatment on \gls{hrqol}. +The \gls{accord} trial\supercite{Gerstein2008} included aggressive glycaemic control targets of \SI{42}{\ifcc} (\textless6\%) in the intensive arm, and 53-\SI{63}{\ifcc} (7-7.9\%) in the routine care arm, and individuals recruited had long standing diabetes and evidence of CVD. Change in \gls{hrqol} was measured by the \gls{sf36}, \gls{dsc}, \gls{dtsq} and \gls{phq9}. After randomisation, individuals receiving intensive treatment reported a larger decrease in \gls{pcs} and perceived hypoglycaemia (\gls{dtsq} item), but greater treatment satisfaction (\gls{dtsq} scale) and less hyperglycaemia (\gls{dtsq} item). The \gls{accord} researchers believed that the statistically significant change in the \gls{pcs} was \emph{trivial}', and that there was no clinically significant impact of intensive treatment on \gls{hrqol}. Bohlin \emph{et al} took a qualitative approach to assessing the impact of diabetes by reviewing the way in which individuals with diabetes, who were diagnosed \textgreater1 year earlier and hadn't initiated insulin, discussed the topic of treatment burden.\supercite{Bohlen2012} They found that in 46 consultations, 83 topics relating to treatment burden were discussed. The burden of administrating treatments was discussed 28 (34\%) times, the potential effects and consequences of treatments 24 (29\%) times, patient concern over being able to attain medication 19 (23\%) times, and trouble complying with the monitoring required for safe use 12 (14\%) times. While the issues over purchasing and getting access to medications may only be prominent due to the study's American locale, it appears that the primary concern for patients when talking with their \gls{gp}s is centred around how to incorporate the medication regime into their lives. Bohlin \emph{et al} noted that two coders independently reviewed all 46 consultations, and there was an 85\% agreement rate. The coders first calibrated their coding technique on similar consultations till the reached \textgreater90\% agreement. While this suggests they attained good inter-rater reliability, there is no evidence that their method is a valid representation of a patient's concern over the burden treatment choices will have on their life. @@ -168,7 +168,7 @@ \subsubsection{Non-randomised cohorts} %---------------------------------------------Figure \begin{figure}[!tb] \centering -\includegraphics[width=\textwidth]{fig_lancetpapersbyhb.jpg} +\includegraphics[width=\textwidth]{fig_lancetpapersbyhb.png} \caption[Currie \emph{et al}'s analysis of glycaeamic control and survival]{Hazard ratios for progression to first large-vessel disease event by HbA1c decile, with Cox proportional hazards model. Vertical error bars show 95\% CIs, horizontal bars show \gls{hba1c} range. Red circle=reference decile. *Truncated at lower quartile. \dag Truncated at upper quartile. Model specification, for people with no previous cardiovascular disease only: age, sex, Charlson index (age unadjusted), total cholesterol, smoking status history, and cohort membership. This figure was published in \emph{Currie et al(2010)Lancet,375(9713):6-12}, and is reproduced with permission.} \label{fig_lancetpapersbyhb} \end{figure} @@ -234,7 +234,7 @@ \subsubsection{Cochrane review on glycaemic control} \subsection{Steno-2 and multifactorial treatment} -\gls{steno2} was a small (n=160) \gls{rct} comparing routine care with behaviour modification and pharmacological therapy to lower both \gls{hba1c} (\SI{48}{\ifcc}; \textless6.5\%) and CVD risk factors (\textless$\frac{130}{80}$ mmHg blood pressure, \textless \SI{4.5}{\lipid} total cholesterol and \textless \SI{1.7}{\lipid} triglycerides). Lifestyle changes included lowing fat intake, increasing exercise and smoking cessation promotion. Aspirin, vitamin supplements and an \gls{ace} inhibitor were also advised in the intervention group. At 7.8 years the trial found that the intensive treatment group had a 53\% lower risk of \gls{cvd} (HR 0.47; 95\%CI 0.24,0.73).\supercite{Gaede2003} The benefits of multifactorial treatment were long lasting and 5.5 years after randomisation ended, \gls{cvd} risk (HR 0.41; 95\%CI 0.25, 0.67) and all-cause mortality remained lower (HR 0.54; 95\%CI 0.32,0.89).\supercite{Gaede2008} +\gls{steno2} was a small (n=160) \gls{rct} comparing routine care with behaviour modification and pharmacological therapy to lower both \gls{hba1c} (\SI{48}{\ifcc}; \textless6.5\%) and CVD risk factors (\textless$\frac{130}{80}$ mmHg blood pressure, \textless \SI{4.5}{\lipid} total cholesterol and \textless \SI{1.7}{\lipid} triglycerides). Lifestyle changes included lowering fat intake, increasing exercise and smoking cessation promotion. Aspirin, vitamin supplements and an angiotensin II receptor antagonist were also advised in the intervention group. At 7.8 years the trial found that the intensive treatment group had a 53\% lower risk of \gls{cvd} (HR 0.47; 95\%CI 0.24,0.73).\supercite{Gaede2003} The benefits of multifactorial treatment were long lasting and 5.5 years after randomisation ended, \gls{cvd} risk (HR 0.41; 95\%CI 0.25, 0.67) and all-cause mortality remained lower (HR 0.54; 95\%CI 0.32,0.89).\supercite{Gaede2008} \subsection{Changes in CVD risk factors} @@ -265,7 +265,7 @@ \subsubsection{Change in cholesterol after diabetes diagnosis} \label{fig_cholesterol} \end{figure} -\cref{fig_stenochol} shows a population with long standing diabetes from the \gls{steno2} trial randomised to a multifactorial intervention. The routine care arm had a total cholesterol target of \textless\SI{13.9}{\lipid} (\textless250 mgdl\textsuperscript{-1}) for the first six years, and then \textless\SI{10.6}{\lipid} (\textless190 mgdl\textsuperscript{-1}) for the remaining two years. The intensive care arm had a total cholesterol target of \textless10.6 (\textless190 mgdl\textsuperscript{-1}) for the first six years, and then \textless\SI{9.7}{\lipid} (\textless175 mgdl\textsuperscript{-1}) for the remaining two years. The changes seen in \cref{fig_stenochol} suggest that intensive goal setting is successful in attaining and then maintaing lower lipid levels when applied within a multifactorial intervention that also includes lifestyle promotion. In the absence of screen-detected populations to draw reference from, \cref{fig_cholesterol} suggests that in a screen detected population we can expect to see a decrease in lipids after diagnosis, that is successful maintained over the first decade after diagnosis. +\cref{fig_stenochol} shows a population with long standing diabetes from the \gls{steno2} trial randomised to a multifactorial intervention. The routine care arm had a total cholesterol target of \textless\SI{6.5}{\lipid} (\textless250 mgdl\textsuperscript{-1}) for the first six years, and then \textless\SI{4.9}{\lipid} (\textless190 mgdl\textsuperscript{-1}) for the remaining two years. The intensive care arm had a total cholesterol target of \textless\SI{4.9}{\lipid} (\textless190 mgdl\textsuperscript{-1}) for the first six years, and then \textless\SI{4.5}{\lipid} (\textless175 mgdl\textsuperscript{-1}) for the remaining two years. The changes seen in \cref{fig_stenochol} suggest that intensive goal setting is successful in attaining and then maintaing lower lipid levels when applied within a multifactorial intervention that also includes lifestyle promotion. In the absence of screen-detected populations to draw reference from, \cref{fig_cholesterol} suggests that in a screen detected population we can expect to see a decrease in lipids after diagnosis, that is successfully maintained over the first decade after diagnosis. \subsubsection{Change in blood pressure after diabetes diagnosis} @@ -283,11 +283,11 @@ \subsubsection{Change in blood pressure after diabetes diagnosis} \end{subfigure}% \begin{subfigure}[b]{.55\textwidth} \centering - \includegraphics[width=.89\linewidth]{fig_ukpdsweight.jpg} + \includegraphics[width=.89\linewidth]{fig_davis_wt_9.png} \caption{Change in weight in the UKPDS} - \label{fig_ukpdsweight} + \label{fig_davu_wt_9} \end{subfigure} -\caption[Change in BP and weight in the UKPDS]{\cref{fig_ukpdsbp} is the change in systolic BP in the first 9 years after diagnosis in the UKPDS. Reproduced from \emph{UKPDS Group (1998) BMJ 317(7160):703-13}\supercite{UKProspectiveDiabetesStudyGroup1998} with permission. \cref{fig_ukpdsweight} is the change in weight in the first 9 years after diagnosis in the UKPDS. $\bigcirc=$ white. $\blacktriangledown=$ Black. $\blacklozenge=$ South Asian. Reproduced from \emph{Davis et al (2001) Diabetes Care 24(7):1167-74}\supercite{Davis2001} with permission.} +\caption[Change in BP and weight in the UKPDS]{\cref{fig_ukpdsbp} is the change in systolic BP in the first 9 years after diagnosis in the UKPDS. Reproduced from \emph{UKPDS Group (1998) BMJ 317(7160):703-13}\supercite{UKProspectiveDiabetesStudyGroup1998} with permission. \cref{fig_davu_wt_9} is the change in weight in the first 9 years after diagnosis in the UKPDS. $\bigcirc=$ white. $\blacktriangledown=$ Black. $\blacklozenge=$ South Asian. Reproduced from \emph{Davis et al (2001) Diabetes Care 24(7):1167-74}\supercite{Davis2001} with permission.} \label{fig_bpandweight} \end{figure} @@ -297,7 +297,7 @@ \subsubsection{Change in weight after diabetes diagnosis} \subsubsection{Glycaemic control after diagnosis} -\cref{fig_ukpdsglu} shows the change in blood glucose in the first nine years after diagnosis, in a cohort analysis of the \gls{ukpds}. This figure represents one of the few times the \gls{ukpds} presented trajectories of blood glucose during the diet only run-in period (although individuals with an FPG\textgreater\SI{15}{\lipid} were randomised immediately), and highlights the large improvement in blood glucose, which is then followed by a gradual loss of glycaemic control. \cref{fig_ukpdsglu} represents only the first 15 of the 23 \gls{ukpds} centres\supercite{Davis2001}, so it remains unclear how these trajectories from a sample diagnosed 20-38 years ago (in 2015) reflects modern therapies. +\cref{fig_ukpdsglu} shows the change fasting in blood glucose in the first nine years after diagnosis, in a cohort analysis of the \gls{ukpds}. This figure represents one of the few times the \gls{ukpds} presented trajectories of blood glucose during the diet only run-in period (although individuals with an FPG\textgreater\SI{15}{\lipid} were randomised immediately), and highlights the large improvement in blood glucose, which is then followed by a gradual loss of glycaemic control. \cref{fig_ukpdsglu} represents only the first 15 of the 23 \gls{ukpds} centres\supercite{Davis2001}, so it remains unclear how these trajectories from a sample diagnosed 20-38 years ago (in 2015) reflects modern therapies. \begin{figure} \centering @@ -347,7 +347,7 @@ \subsection{Screening for diabetes} \begin{longdescription} \item[Important condition with a known natural history:] Diabetes is an important condition that increases the risk of \gls{cvd} events, early mortality and represents an economic and quality of life burden. The stages of the disease are well documented, and there is a measurable progression from healthy to poor glyceamic control. - \item[Has a known latent period:] In 1992, Harris \emph{et al} estimated individuals reach the threshold for diabetes 4-7 years before a clinical diagnosis.\supercite{Harris1992} This estimate was derived from extrapolating back from the known prevalence of retinopathy at diagnosis, to estimate when retinopathy is likely to of begun. In the Ely study, were individuals were recruited between 1990-1992, the lead time appeared to be around 3.3 years.\supercite{Rahman2012} + \item[Has a known latent period:] In 1992, Harris \emph{et al} estimated individuals reach the threshold for diabetes 4-7 years before a clinical diagnosis.\supercite{Harris1992} This estimate was derived from extrapolating back from the known prevalence of retinopathy at diagnosis, to estimate when retinopathy is likely to of begun. In the Ely study, where individuals were recruited between 1990-1992, the lead time appeared to be around 3.3 years.\supercite{Rahman2012} \item[Cost-effective primary preventions should be in place first:] Efforts to lower population levels of obesity and improve diet and exercise like Change4Life\supercite{Hardy2010} are in place, but of limited effectiveness and not at the intensity of effective strategies identified in studies. \item[There should be a simple, safe and precise screening test:] Diagnostic labels in diabetes are derived from the relationship between glyceamic control and future risk of microvascular complications. Whether individuals meet this criteria is measurable in general practice and separated sufficiently from those with `healthy' blood glucose control. Individuals at risk can by identified by risk scores and capillary testing, and diabetes can be diagnosed in the general practice by taking venous samples. The introduction of diagnosis based on \gls{hba1c} means a non-fasting sample, which minimises disruption to the individual being screened. \item[There should be an effective treatment:] While the \gls{ukpds} demonstrated a benefit of tight glyceamic control, diagnosis of diabetes will also lead to tighter management of total cardio-vascular health, the long-term effects of which are promising.\supercite{Black2014DiaMed}
BIN  ThesisFigs/.DS_Store
Binary file not shown
1,578 additions, 1,773 deletions not shown
BIN  ThesisFigs/fig_davis_wt_9.png