The significant frequency of injuries in the first period in both drug groups may be related to the occurrence of dizziness and vertigo. Other disorders, especially cardiovascular conditions, that were significant in the lorazepam patients tended to occur in the elderly (Table 4) and may be related to the slight preponderance of older patients in this drug group. There were no hospitalizations for rare, potentially life-threatening events, such as agranulocytosis and erythema multiforme, occurring soon after starting treatment in either drug group.

These findings suggest a similar event profile for the two benzodiazepines following the start of treatment and that there is no particular concern about the safety of either drug. Results from 10,895 patients in a PEM study also showed that serious events, including uncommon ADRs, do not occur during treatment with alprazolam ‘to any great extent’.

In the previous evaluation of this methodology, the importance of the impact of a pre-existing condition () on the development of an adverse effect (peptic ulceration) of nonsteroidal anti-inflammatory drugs was identified . In this analysis, differences in the occurrence of events following the receipt of the benzodiazepines in patients with or without previous health care services for psychiatric disorders and patients who had or had not previously taken a benzodiazepine were examined. However, apart from a higher occurrence of psychiatric conditions (especially anxiety, depression and sleep disorders) in patients with previous services for psychiatric disorders and those who had previously had a benzodiazepine, there were no apparent different outcomes in terms of potential ADRs.

There are at least three important limitations to this study. First, there were only 8525 alprazolam patients compared with 20,000 lorazepam patients; therefore, statistical significance is more likely in the lorazepam group than in the alprazolam group. Second, as previously mentioned, data were supplied in aggregate form, preventing an evaluation of individual patients, which should be possible with these datafiles , including an assessment of the effect of any concurrent medications on the event profile. Third, due to limited resources and the provision of data in aggregate form, there was no examination of medical records for validation purposes or to obtain clinical information about particular events . Follow-up of physician services with three-digit ICD codes that may imply a serious adverse event (eg, code 288 may indicate a mild white blood cell disorder or agranulocytosis, and code 695 may indicate a common erythematous condition or erythema multiforme) is essential.