DESCRIPCIÓN DEL ALCANCE

The specific limitations of each study are explored within the discussion section of the

corresponding chapter. However there are a number of limitations that are applicable to all the documented works, which warrant further comment.

A single measure of the impact of FFB was used throughout this programme of research only, the FIS questionnaire (Helliwell et al. 2005). At the time of research, this was the only tool that had been previously validated for use in patients with RA and was therefore selected for this reason. However, the FIS questionnaire has not been validated for longitudinal use, although Turner et al. (2007) suggested that a score change of three or more, in either direction, was clinically meaningful. As such, these margins of change were adopted in this work.

Consequently, the reported estimations of disability impact associated with FFB may be biased by measurement inaccuracy. Such bias is likely to be systematic throughout the reported results, however the direction (e.g. over/under estimation) is unknown. Use of secondary measures, that identify concurrent validity between items, would improve the certainty of an accurate

assessment of disability (Silman and Hochberg 2001). Validation of FIS longitudinally would also be of benefit in improving the construct validity of these reported measures. Thus, the variation in disability reported in Chapter four, may be a consequence of measurement error. The

additional use of margins of clinically meaningful change in disability, suggested by Turner et al. (2007), potentially increased the stringency of evaluation of change, thereby minimising the effect of measurement error. Conversely, the loss of sensitivity to change, introduced through the grouping of continuous data, may have resulted in an underestimation of clinically

meaningful fluctuation in disability (Altman and Bland 1999, Bland and Altman 1996).

The differences in longitudinal disability demonstrated between patients with high/low FFB count are also of interest. Foot impairment improved for both patients with high and low FFB at

baseline, with the reported impairment converging between groups at year-three. This may represent an overall regression to the mean disability level for the studied cohort (Bland and Altman 1994b). Alternatively, patients entering the study at baseline were offered podiatric care if required and thus the reduction in impairment may represent a treatment effect (Campbell and Machin 1999). Furthermore, it is possible that those patients with an elevated burden of foot disease, to which FFB presence may be contributing, preferentially sought intervention accounting for the greater reduction in impairment shown in this group. Additional post-hoc analysis of treatment provision would benefit the interpretation of these results (Campbell and Machin 1999). It is also plausible to suggest that those patients with a pre-existing foot complication were more likely to participate within this study and as such the burden of foot disease may be over represented within this cohort. Future inclusion of recruitment analyses may provide further insight into this potential bias. However, the prevalence of reported foot impairment and activity limitation within this study is consistent with those reported by previous authors (Helliwell et al. 2005; Turner et al. 2007). However it should be noted, that the same preferential bias for recruitment may also be evident within these comparative investigations.

It is interesting, that in contrast to improved foot impairment, reported activity limitation increased. It is plausible that in a treated cohort, patients who feel better and have less foot impairment attempt to undertake more weight-bearing tasks of daily living (Campbell et al. 2012, Platto et al. 1991). Such patients may become increasingly aware of activity limitations that were otherwise masked by comorbid disease or other social influences/distractions (Bjork et al. 2011). Conversely, while foot impairment may be modifiable and can improve, factors contributing to activity limitation perhaps accumulate only (van der Leeden et al. 2007, van der Leeden et al. 2008). Therefore it may not be possible for improvements in foot impairment to be mirrored with improvements in foot-related activity limitation. In this instance, the assessment of FFB presence may be a useful indicator of potential activity limitation; FFB assessment could be used clinically to identify those patients at greatest risk of worsening activity limitation and to whom

management should be targeted.

The degree and impact of pain associated with FFB has not been evaluated in this programme of research. This is a key omission and the potential influence of pain or pain related anxiety, as a confounding or colinear factor, should be considered (Otter et al. 2011, Mustafa et al. 2012,

McWilliams et al. 2012, Campbell et al. 2012). Interestingly, Otter et al. (2011) note a positive correlation between reported pain and healthcare access. Furthermore, various authors have reported a high incidence of foot specific pain associated with the development of soft tissue lesions of the forefoot (Koski 1998, Iagnocco et al. 2001, Ashman et al. 2001, Zielaskowski et al. 2000). Thus, if longitudinal differences in reported disability are to be considered as a treatment effect, the presence of pain associated with FFB may further confound differences in disability observed between high/low FFB count groups.

A number of investigative techniques have been used to evaluate the mechanical function of the foot. Clinically used static postural measures include navicular height, arch height and valgus index [86, 87]. However, these have been demonstrated to have significantly different inter-rata reliability (p = 0.001- 0.005) even when evaluated in non-pathological populations (Weiner- Ogilvie and Rome 1998; Menz et al. 2005). The foot posture index (FPI) was the only clinical static postural tool validated for use for patients with RA.at the time of this study (Redmond et al. 2006). The authors acknowledge that the FPI was only able to predict 64% of the variance in static standing position and 41% of the postural variance during the stance phase of gait in this patient group (Redmond et al. 2006). None the less, this has been shown to have good internal consistency (Cronbachs alpha=0.83) and allows simple multi-segment, multi-plane evaluation in a clinical setting and as such has clear advantages over the alternative postural tools.

However, it should be noted that when applied in this work, substantial inter-rater variation was recorded, even after two episodes of dedicated researcher training. As such, and despite moderate agreement subsequently being demonstrated, the findings of this score should be treated with caution; there is likely to be some reporter bias within the recorded results.

Furthermore, the clinical relevance of a static measure of alignment is yet to be demonstrated. It has previously been hypothesised that variations in static weight bearing alignment can be related to dynamic biomechanical stress (Hicks 1953; Root et al. 1977; Dananberg 2000, Fuller 2000; Kirby 2001). However there is much contention surrounding the theoretical underpinning of this inferred relationship (Wold et al.2008). None the less, the results of the work completed as part of this thesis do appear to suggest that, despite these limitations, the FPI score may have clinical relevance when considering the relationship between static foot alignment and FFB prevalence. Further work is required to fully determine the nature of this association.

The provision of ongoing rheumatological care, concurrent to the longitudinal study completed in Chapter four, may introduce treatment effect as a possible confounder within the reported results. Changes in the provision of pharmacological care were not reported or analysed within the scope of this investigation. However, it is plausible that a number of patients will be in receipt of disease modifying anti-rheumatic drugs, such as Methotrexate, which have been previously linked with soft tissue change and nodulosis, or will have had an escalation in drug therapy to include newer biologic therapies, the secondary effects of which remain unknown. As such,

future work which clearly evaluates the potential contributory or colinear effect that differing treatment regimens may have on the soft tissue structures of the foot is warranted.

The FFB-score was validated for use in patients with RA against MRI-determined measures of disease activity within the forefoot, adopted from the RAMRIS score (Ostergaard et al. 2003). However, these measures have been validated for application to the metacarpophalangeal joints and not the metatarsophalangeal joints (Lassere et al. 2003). None the less, there is biological plausibility to the application of this tool to these forefoot joints, although the accuracy of

localised disease activity assessment, within the reported validation study in Chapter six, should be considered.

8.4.1 Management of bias

There are a number of potential biases within the reported study findings. The following text therefore summarises such potential sources of bias and considers the methodological adjustments used to manage this where possible. The implications of such biases to the conclusions of this work are considered.

Selection bias, the introduction of error due to systematic differences in the characteristics between those selected or not selected to participate within the study, may be evident within the represented sample population (Silman and Macfarlan 1995). Those patients attending a rheumatology outpatient clinic, within a secondary care setting, were consecutively invited to participate within the baseline study reported in Chapters four and five. As such, the target samples are likely to include those patients with chronic or more active disease and may not be representative of all patients with a rheumatological diagnosis. Indeed, the demographical analysis completed appears to suggest that those patients with more established RA disease are represented within the studied population (Bland and Altman 1994b). The applicability of the study findings to those patients with early disease should therefore be questioned. Future work, that includes an inception cohort, would benefit further study in this area. Post-hoc response analyses were completed for the longitudinal study completed in Chapter four, the findings of which suggest that there may be a systematic difference in those patients willing to return for follow-up assessment and those who withdrew from the study. As such, systematic bias in the inclusion of those patients with the greatest burden of foot or inflammatory disease may result in an over-representation of these or associated characteristics within the study findings. The generalisability of the results of this work the broader population should therefore be considered.

Prevalence bias, the restriction of studied cases to those with only the disease state of interest, was minimised by the recruitment strategy chosen, and is therefore unlikely to be evident within the studied population; Included participants were not recruited based upon the presence or absence of FFB, the primary outcome of interest (Silman and Macfarlan 1997).

Survival bias, the exclusion of cases no longer able to participate within the studied population arising from factors related to the primary outcome of interest, may be evident within the represented sample population (Silman and Macfarlan 1997). It is plausible to suggest that those patients with a greater burden of foot disease and related mobility limitation may be restricted from attending follow-up study visits. Quantification of such bias was attempted via the completion of the response analyses reported in Chapter four (Bland and Altman 1994b). Furthermore, a methodological adjustment to recruitment was introduced subsequent to initial poor response rates at the year three follow-up visit. Eligible participants were offered an option to complete a postal questionnaire only, negating any need for physical attendance at a hospital examination appointment. However, changes to response rates were negligible following this amendment and as such, the likely role of disability as an inhibitor to response can be questioned.

Participant recall bias, differences in a patient’s ability to recall information of relevance, may be evident within the represented sample population (Silman and Macfarlan 1994b; Silman and Hochberg 2001). The duration and severity of episodes of poor foot heath may influence the degree of foot-related disability reported by study participants. Similarly, the impact of poor foot health may be more readily recalled if recent life events have been modified as a consequence; the period of time between FIS questionnaire completion and an episode of poor foot health may impact the patient’s ability to recall pertinent information. To overcome this, participants were encouraged to consider their responses, and answer based upon their experiences of foot health over the previous week only. Additionally, details of surgical history, previous episodes of poor foot health or related disease activity were cross references and verified by review of the patient’s medical notes by the study researcher. However, it remains plausible that FIS scores may be over or underestimated dependent upon a patient’s ability to recall related information.

Participant reporting bias, error in response that is dependent upon the willingness of a

participant to provide a true response or disclose information of relevance, may be evident within the represented sample population when considering reported foot impairment or activity

limitation (Silman and Macfarlan 1994b). However, the primary outcome of interest, FFB, was determined based upon US analysis and is therefore unlikely to be influenced by such reporting bias. To negate potential reporting bias, the researchers were cautious not to demonstrate any preference towards the identification or absence of disability or foot complications.

Researcher observer bias, the systematic error in the researcher’s measurement, reporting or documentation of the phenomena of interest, may be evident within the study findings (Silman and Macfarlan 1994b). The identification or FFB using US or MRI may be altered as a

consequence of interest or improving familiarity with the image sets. To minimise such bias, all researchers completed assessment skills training prior to the undertaking of data collection. Intra and inter-reader agreement for all methods of FFB identification were completed and the

findings of the study researcher (LH) referenced against the findings of an ‘expert’ reader (LK/CB). Thus, the relative accuracy of all researchers for the identification and characterisation of FFB has been quantifiably demonstrated to be good, subsequent to agreement analyses. However, it is of note that a learning effect was reported by the main researcher (LH) who was novice to MRI reading at the start of this study. As such, the data reported by LH was not used until consistently high image agreement with an expert reader was achieved, thereby minimising potential reporting inaccuracy. In study two, (Chapter five) there remains however some risk of non-directional misclassification bias when comparing observed phenomena between groups; equal observer bias may occur in both study population. However, the researcher is not blinded to the group from which the participant originates and thus observational bias may potentially be systematic and directional in nature.

It was not feasible to blind the researcher to the patients clinical history in studies one, three and four (Chapters four, six and seven) or patient group in study two (Chapter five). However the comparative expert readers were blinded to the patient’s disease status at the time of image acquisition. Thus any potential systematic reporting bias by the main study researcher would be evident within the completed agreement analysis. Thus, there is minimal indication there may be any systematic reporting bias by the main researcher (LH) in the reading of US or MRI images and subsequent FFB prevalence reporting.