Study limitations

A limitation of this study is the lack of an age matched control group. The initial design of the study was to include three groups including an age matched group of control subjects for direct comparison to the confirmed osteoporotic group. During the design phase, we considered arranging DXA scans for an age matched control group, however because of a limited budget and requirement of a large sample size, this was not feasible. Another problem with this approach is that a certain amount of disease would be identified, resulting in the research team being responsible for arranging osteoporosis treatment. We also explored the possibility of recruiting subjects that had already been screened by DXA scan and confirmed as not being osteoporotic. This involved arranging meetings to discuss the research with medical staff at a number of health centres, and osteoporosis screening facilities. Few suitable candidates were found. Following exhausting all avenues, and after discussion, it was decided to omit the age matched control group from the study. To recruit a suitable control group, we applied strict exclusion criteria to reduce the likelihood of subjects in this group being osteoporotic. However, it should be acknowledged that the BMD of this group was also unknown. While the risk factors for osteoporosis in this group were low, without BMD measurement, we cannot be certain that this group did not contain osteopenic or osteoporotic subjects. Ideally subjects in the control group should have BMD measurements, with T-scores within the normal range (> -1), excluding anyone with a T-score below -1.

We followed up 64 out of the 86 confirmed osteoporotic subjects recruited into the study. This represents a drop out rate of 25.6%. The subjects recruited to this study were keen to participate. An issue with recruiting subjects from this demographic is that they are of an age more prone to medical issues, with the result that it may be difficult to predict a drop out rate. Reasons for subject drop out included intolerance of the prescribed osteoporosis treatment, change of treatment during the observation period, non-compliance with the treatment regimen, age related cognition deterioration, and passing away. A history of cancer is a contraindication for Teriparatide, and one participant developed breast cancer shortly after commencing Teriparatide, and another who was planned for Teriparatide developed a lung mass, requiring further investigation and suspension of osteoporosis

treatment.78 _{Our power calculation showed that we needed 24 subjects in each of} the treatment subgroups, and fortunately we were able to follow up 25 subjects in the Teriparatide group and 26 in the IV Zoledronate group.

The primary researcher performed all the ultrasound measurements in this study, which introduces operator bias. Ideally the operator should be blinded to the aims of the study. This would have required the recruitment of another researcher who would be blinded to the aims of the study to carry out the scans. This would not have been realistic considering that the data collection for this study involved between five to ten hours a week, and took two years to carry out. Radiographers in St James’s Hospital, who were not aware of the aims of this study, carried out measurement of BMD. This is the only form of blinding we were able to implement. Subjects who participated in this study were not recruited by the primary researcher, they were recruited by two gatekeepers: one gatekeeper was a member of staff based in the Dublin Dental Hospital, and the other was one of the specialist bone health nurses based in St James’s Hospital. The use of the gatekeepers in this study reduces the risk of selection bias.

The arm of this study looking at response to treatment suffers from attrition bias. We had a drop out rate of 25.6% in the osteoporotic group. These were elderly patients, and therefore have a higher risk of medical issues related to ageing. This would have had an effect on our results investigating response to osteoporosis treatment.

The difficulty in taking readings from a number of sites would have effectively resulted in lower than planned numbers required for a number of statistical tests. The mandible was the only site with a 100% success rate in taking a reading and therefore was the only site that reached the numbers required when comparing the two groups. The frontal bone did achieve the number of successful readings for the control group, but fell just short of the numbers required for the osteoporotic group. The numbers of successful readings on the zygoma fell well short of the power

calculation with the result that no meaningful conclusions regarding SOS readings taken from this site can be made.

It was planned that subjects in the osteoporosis group would have DXA scans performed at or prior to recruitment, and the one-year mark, when the second set of ultrasound readings would be taken. During the study period, there have been a number of infrastructure changes in St James’s Hospital. This lead to staffing issues in the radiography department and increased waiting times to have DXA scans. As a result we were unable to arrange DXA scans for a large number of our osteoporotic cohort.

At the beginning of the study, DXA scans were performed with a GE Lunar Prodigy. This was replaced with a Hologic Horizon A after a number of months into recruitment for the study. It has been shown that there are variations between

different brands of scanners and it is difficult to compare results.154 The software

from the Hologic does incorporate an algorithm, which adjusts the score from a previous scan, compensating for the variation.

The arm of the study looking at response to osteoporosis treatment did suffer from a limited time frame to carry out this research. Subjects were followed up for one year, however those on teriparatide were on a two-year treatment course. On the Bone Health Clinic in St James’s Hospital, there is a clearly outlined protocol for teriparatide treatment. Patients deemed suitable for teriparatide, are seen by the specialist bone health nurses for education in the treatment regimen at initiation of treatment. They are then reviewed after three months to assess their tolerance to the treatment, and to have bloods to ensure there is no abnormality with calcium metabolism and to check biomarker response. They are then seen one year after initiation of treatment again to have blood tests. After two years, patients are then seen on the Bone Health Clinic to review their response to treatment with a DXA scan, and to discuss anti-resorptive treatment to ensure the anabolic effect of the teriparatide course is not lost. A treatment protocol such as this is highly conducive to research such as our study. It would be desirable to follow patients for the full duration of the teriparatide course, and to restrict the study to patients who are treatment naïve, so that the effect of previous anti-resorptive treatment does not

interfere with the remodelling effect of teriparatide. To conduct such a study, following patients on the full course of teriparatide, would need a longer timeframe to complete.