Results
Of 35,570 new users of cholinesterase inhibitors, 30,174 were users of donepezil, 1,176 were users of galantamine, and 4,220 were users of rivastigmine. Table 1 shows the baseline characteristics of the study population. Mean age was 81.6 ± 7.3, 26,542 (74.6%) of the participants were women, and 29,265 (82.3%) were white. Some participants had a history of pneumonia (13.1%) and chronic obstructive pulmonary disease (28.2%). Galantamine users generally had lower rates of health service utilization. A greater proportion of rivastigmine users had Parkinson's disease (13.9%), Lewy body disease (7.0%), and use of antipsychotics (19.9%) and anti-Parkinson medication (11.8%) than of donepezil (5.1%, 2.8%, 14.7%, and 4.4%, respectively) and galantamine (7.5%, 4.5%, 17.4%, and 6.5%, respectively) users.
There were 1,249 cases of pneumonia during the mean follow-up period of 246.9 days. The cumulative incidence of pneumonia was 51.9 per 1,000 person-years. Incidence was highest in donepezil users (52.6/1,000 person-years), followed by galantamine users (42.9/1,000 person-years) and rivastigmine users (42.0/1,000 person-years). The incidence curves in Figure 1 indicate that the rate of pneumonia was significantly lower in rivastigmine users than donepezil users (P < .001). The unadjusted pneumonia risk was lower with rivastigmine (HR = 0.76, 95% CI = 0.62–0.95) and galantamine (HR = 0.76, 95% CI = 0.52–1.10). After multivariable adjustment, pneumonia risk remained significantly lower with rivastigmine (HR = 0.75, 95% CI = 0.60–0.93), whereas the risk was not significantly different between galantamine and donepezil (HR = 0.87, 95% CI = 0.62–1.23). When pneumonia risk with the oral and transdermal forms of rivastigmine was compared, 39% and 14% lower risks, respectively, were observed than with donepezil (Table 2).
(Enlarge Image)
Figure 1.
Cumulative incidence of pneumonia risk in new users of cholinesterase inhibitors. P < .001 for comparisons between the index medications.
As shown in Table 2, pneumonia risk increased with advancing age and was greater in men; participants with a history of pneumonia; and participants with heart failure, chronic obstructive pulmonary disease, and previous use of antipsychotics, antidepressants, antiepileptic agents, and loop diuretics.
In subgroup analyses and sensitivity analyses, the HRs were similar to those in the primary analysis (Figure 2). Lower risk of pneumonia was found in rivastigmine users than in donepezil users (HRs ranging from 0.38 to 0.95), and the risks were similar between galantamine and donepezil users (HRs ranging from 0.75 to 1.10). Pneumonia risk was even lower in rivastigmine than donepezil users in those who had Parkinson's disease or Lewy body disease (HR = 0.43, 95% CI = 0.22–0.94), whereas it was not significantly different in those who did not have these conditions (HR = 0.84, 95% CI = 0.67–1.05).
(Enlarge Image)
Figure 2.
Subgroup and sensitivity analyses of pneumonia risk in new users of cholinesterase inhibitors. Hazard ratios (HRs) for pneumonia outcomes comparing (A) rivastigmine and (B) galantamine with donepezil from the primary analysis and subgroup and sensitivity analyses. High-dimensional propensity scores were generated using two models: a donepezil–galantamine group and a donepezil–rivastigmine group. HDPS = high dimensional propensity score; ED = emergency department; CI = confidence interval.