Background Periprosthetic joint infection is a serious complication and a major reason for revision surgery after primary shoulder arthroplasty. The prophylactic antibiotics for primary shoulder arthroplasty that have predominantly been used in Sweden are cloxacillin and clindamycin. To address Cutibacterium acnes, benzylpenicillin has recently increasingly been added to cloxacillin, but it is unclear which antibiotic prophylaxis regimen is the most effective to prevent periprosthetic joint infection. Questions/purposes After controlling for baseline differences among patients such as age, gender, previous surgery, cement fixation, and arthroplasty type, was the risk of reoperation for infection higher in patients who received cloxacillin than in those who received clindamycin or the combination of benzylpenicillin and cloxacillin? Methods Data from the Swedish Shoulder Arthroplasty Register were used for this study. The inclusion criterion was registered antibiotic prophylaxis in primary arthroplasty. Between January 1, 1999, and December 31, 2019, 22,470 primary shoulder arthroplasties, including total shoulder, hemiarthroplasty, and reverse shoulder arthroplasties, were entered into the Swedish Shoulder Arthroplasty Register. Reporting of antibiotic prophylaxis to the register was introduced on January 1, 2013. Since then, the completeness of information on the type of antibiotic prophylaxis in the reports has been 85.3%. Consequently, 10,706 arthroplasties were eligible and fulfilled the inclusion criterion of reported antibiotic prophylaxis. A further 129 were excluded because of unusual prophylaxis regimens, leaving 10,577 shoulder arthroplasties for analysis. The Swedish Shoulder Arthroplasty Register gathers information from all 60 hospitals performing shoulder arthroplasty in Sweden, and through a comparison with the National Patient Register, it has been estimated that more than 90% of all primary shoulder arthroplasties and shoulder reoperations are reported to the register. The age of the study population ranged between 16 and 98 years; the mean age at the primary surgery was 70 +/- 10 years for the entire cohort, with a mean age of 67 +/- 10 years and 72 +/- 9 years for men and women, respectively. The mean observation period was 989 +/- 669 days. From 2013 to 2019, there was a clear change in prophylaxis; in particular, the use of the combination of benzylpenicillin and cloxacillin increased dramatically and the use of cloxacillin alone decreased. Clindamycin prophylaxis increased moderately. The primary study endpoint was reported reoperation for infection. In the register, this is defined as repeat procedures of any kind, including biopsy, lavage of the joint, or revision, defined as secondary surgery in which a component was exchanged, removed, or added. To compare the reoperation rate in relation to the different antibiotics used, which changed over time, we controlled for age, gender, previous surgery, cement fixation, and arthroplasty type using a Cox proportional hazards model. Results When adjusting for age, gender, previous surgery, cement fixation, and arthroplasty type, cloxacillin prophylaxis was associated with an increased relative risk of reoperation for infection compared with the combination of cloxacillin and benzylpenicillin (hazard ratio [HR] 2.40 [95% confidence interval (CI) 1.35 to 4.25]; p = 0.003) and compared with clindamycin alone (HR 1.78 [95% CI 1.11 to 2.85]; p = 0.02). No difference was found between the cloxacillin and benzylpenicillin combination and clindamycin (HR 0.74 [95% CI 0.42 to 1.32]; p = 0.31). Conclusion Our results indicate that prophylaxis against C. acnes may be warranted in shoulder arthroplasty. Because the absolute number of infections was low and infections could have been underreported to the register, our results should be interpreted with caution. There is no available information about the causative microorganisms. The study lays the groundwork for further investigations of antibiotic prophylaxis regimens in shoulder arthroplasty. Because large randomized controlled trials would be impractical to perform, prospective register-based randomized controlled studies might be a viable method.