To the Editor: I note with interest the findings of Diamond et al.1 However, I would like to make some comments about the validity of the results presented.

The authors state that an intention-to-treat analysis was used. However, this was unusual as the intention was to treat everybody, with the control group being made up of patients who were offered the intervention but refused it because of the lack of published data on the safety of the procedure.

There was a marked decrease in the pain scores at 24 hours in the intervention group. The methods state that all patients were offered similar analgesia titrated to individual need. However, there was no mention of how many of the intervention group (if any), compared with the control group, received parenteral analgesia. The intervention group may have tended to receive more parenteral analgesia than the control group, but this was not mentioned. Indeed, Predey et al2 specifically mention this possibility in their review.

Diamond et al state that lower pain scores persisted in the vertebroplasty-treated group at 6 weeks. However, from the results given in Box 3, it seems that there was no clinically significant difference between the intervention and control groups at 6 weeks. I would have liked the report to have included pain scores at 1 week.

The decision to use means ± 1 SD instead of 95% CIs in the results is interesting. One SD will only include 66% of a normal population whereas a 95% CI would refer to the mean ± 2 SDs. More importantly, the size of the SD introduces the possibility of an enormous spread in the pain scores for both groups, which was not commented on in the Results or the Discussion.

Finally, it was unclear whether the patients were treated in the private or public hospital system.

Regardless of my comments above, I applaud any attempt to treat back pain in elderly people (especially that caused by osteoporotic crush fractures) in a time-expedient fashion. I look forward to the day when this therapy is first-line treatment for this disease and not something to consider 2 weeks down the track. The cost savings in reduced length of stay and the decreased morbidity associated with this treatment, as well as the reduced geriatric workload with the more rapid improvement in function, must outweigh the initial up-front costs.

In reply: We thank Mallows for his constructive comments, and would like to reply as follows.

1. Intention-to-treat analysis is normally associated with randomised controlled trials (RCTs). Our study was not an RCT, as patients who refused vertebroplasty formed the comparison group.1 We used intention-to-treat analysis to indicate that all patients who were assessed as eligible for our study at the outset were enrolled and were all included in the results presented, irrespective of subsequent events (ie, whether they were completely compliant, had died or were lost to follow-up). We thought it was important to make clear that the study was prospective, that all patients entered into the study were represented in the outcomes, and that patients had not been selected on the basis of later events or results.

2. A description of the analgesic requirements of a subgroup of the patients in our study has been published.2 More than 90% of the vertebroplasty-treated cohort were either able to cease or reduce their analgesia dose by at least 50% within 24 hours after the procedure. Before the procedure, an intravenous injection of pethidine (50–100 mg) was the only parenteral analgesia given routinely and would not have affected the pain scores at 24 hours.

3. Pain scores at 6 weeks were statistically lower in the vertebroplasty-treated group compared with the control group. The clinical significance was not apparent, as measured by the changes in the Barthel indices. This indicator is only a crude assessment of patients’ wellbeing, and a more detailed questionnaire of activities of daily living may have shown a difference. Pain scores recorded at 24, 48, 72 and 96 hours after the procedure would have been more sensitive, so that the additional pain, without vertebroplasty, could have been calculated from the area under the pain curve; the more data points, the more accurate the curve.

4. We elected to report the results as means ± 1 SD instead of 95% CIs. We used SDs to be consistent with our previous publications and other reports in this field. We agree that we could have used 95% CIs.

5. Patients were referred to Sydney Imaging Group from the inpatients and outpatients departments of St George public and private hospitals. All patients were treated without any out-of-pocket patient expenses. Private hospital inpatients were billed according to standard Medicare fees and outpatients were bulk-billed, which was revenue neutral for the radiology practice.