Association between oral contraceptives and serious infections: A population-based cohort study
Abstract
Aims
Oral contraceptives (OC)s are commonly used worldwide. In a recent study, we showed that the use of OCs is associated with an increased risk for neutropenia. We aimed to investigate the clinical implications of this finding by examining the infection rates of 4 serious infections before, during and after OCs.
Methods
A retrospective cohort study using the electronic medical records of a large health organization. We selected 2 retrospective cohorts of women aged 16–40 between years 2005 and 2019. The first cohort examined infection rates during 2 years before OC use and 2 consecutive years of adherent OC use. The second cohort included women who consumed OCs adherently for 2 years and then discontinued their use for 2 consecutive years. Women's infection rates were compared by χ2 test, results were stratified by OC type and age.
Results
Overall, 21 595 and 20 728 women were included in Cohorts 1 and 2 respectively. We found a statistically significant higher relative risk for infection while using OCs; the overall risk ratios (95% confidence intervals) for infection in Cohorts 1 and 2 were 1.35 (1.32–1.38) and 1.27 (1.24–1.31), respectively. The overall infection risk remained statistically significant when stratified by age.
Conclusions
This study demonstrates a high statistically and clinically significant risk for all infections followed during OC consumption, which is likely to have major clinical and economic implications. These findings may have implications to millions of women worldwide and should lead to more research on the safety of the pill.
What is already known about this subject
- Recent studies have shown that the use of oral contraceptives is associated with increased risk of neutropenia and high circulating oestrogen levels are associated with staphylococcal nasal carriage among women.
What this study adds
- This study demonstrates a high statistically and clinically significant risk for serious infections during oral contraceptive consumption, which is likely to have major clinical and economic implications.
1 INTRODUCTION
Oral contraceptives (OCs) are among the most commonly used contraceptive methods worldwide. There are 2 main types of OCs: combined OCs, which contain oestrogen and progesterone; and the progestin-only pills. Combined OCs are more commonly used while progestin-only pills appear to be an appropriate contraceptive choice mainly for women with contraindications to oestrogen-containing contraceptive.1 In the USA, 35% of adult women are currently using hormonal methods of contraception and 88% were exposed to some form of contraceptive.2 In a recent study we have shown that the use of OCs is associated with increased risk of neutropenia (risk ratio [RR] = 1.2, 95% confidence interval [CI] 1.1–1.35, P < .01).3 Oestrogen receptors4 are expressed in various lymphoid tissue cells including lymphocytes, macrophages, and dendritic cells.5 To date, numerous studies have investigated the relationship between sex hormones and the immune system, demonstrating immunomodulatory effects of oestrogens.6-9 Oestrogen has been shown to depress cell-mediated immunity, impair the activity of natural killer cells, and suppress some aspects of neutrophil function9. Studies in cell culture, animals and humans indicate that pregnancy, oestrogen supplementation and menstrual stage can affect the acquisition and severity of certain bacterial, parasitic and viral infections.10 The potential effect of oestrogen on neutrophil count is also suggested by the known fluctuating neutrophil counts throughout the menstrual cycle.11-13 This is further strengthened by in vitro illustration of an inhibitory effect of oestrogen and progesterone on neutrophil responsiveness,14 and by the known anti-inflammatory state during pregnancy, where high levels of oestrogen and progesterone are present. Moreover, the association between high circulating oestrogen levels and staphylococcal nasal carriage among women, which is linked to clinically significant infections,15, 16 has been reported. A study conducted in 2012 found that women consuming OCs were more likely to be persistent carriers,17 and another study published recently showed that women consuming OCs had an odds ratio of 2.3 for Staphylococcus aureus nasal carriage.18 A variety of mechanisms contributing to this immunomodulation have been described.19-23
The effect of progestins on the immune system has also been extensively studied. In vitro studies demonstrate that cell exposure to progesterone 4 alters the immune environment in different tissues by promoting an anti-inflammatory status.24-28 Epidemiological studies have demonstrated a higher susceptibility to HIV transmission and acquisition while using oral contraception.29-32
Since neutropenia results in an increased susceptibility to infections, and due to demonstrated interactions between sex hormones and the immune system, we chose to examine potential OC-induced changes in incidence rates of 4 common infections: acute pharyngitis, bronchopneumonia, urinary tract infection (UTI) and cellulitis.
1.1 Settings
Maccabi Healthcare Services (MHS) is the second largest health maintenance organization in Israel, insuring over 2 million members. According to the Israeli National Health Insurance act, MHS is forbidden to reject any citizen who wishes to be insured. Ergo, MHS insure every section of the Israeli population and its data are representative of the Israeli population.32 The annual attrition rate of MHS membership is 1%. Since the early 1990s MHS has established an electronic patient record, where all medical data are collected, including diagnoses, medication prescriptions and purchases.33
2 STUDY POPULATION AND RATIONALE
We selected 2 retrospective cohorts between 2005 and 2019.
The first cohort consisted of all women aged 16–40 years, who have had their first dispensed OCs after at least 2 years during which they had no purchases, and then purchased OCs continuously for at least 2 consecutive years.
To account for changes over time in treatment patterns, reporting issues and awareness during our observational pre–post design study period, we built a second cohort in which 2 years of full OCs use were compared with 2 consecutive years of no OC use. OC records of consumption were extracted from MHS purchase records. Exposure to OCs was estimated using proportion of days covered during follow-up period. Patients with a proportion of days covered level of 80% or above were considered as adherent.34 We chose a 2-year window, sufficiently long to be able to discern changes in infection rate. A longer period would have significantly reduced the number of subjects available for this study.
We excluded women who were prescribed any antiepileptic agent, antipsychotics or immunosuppressants (Figure 1) due to their known associated risks for neutropenia. In addition, we excluded women who were pregnant anytime during the 2 years prior to OC initiation Cohort 1) or after discontinuation (Cohort 2), due to pregnancy's known effect on neutrophil counts. Women who changed OC type (combined vs. progesterone only) throughout this time period were also excluded (Figure 1). Women were allowed to participate twice if they had 2 periods that fulfilled the above criteria: Cohort 1 n = 192, Cohort 2 n = 155. A total of 21 595 and 20 728 periods were eventually included in Cohorts 1 and 2. The rational for this design is that the number of women who began and then discontinued oral contraception after 2 years of use, together with our exclusion criteria, was very small and unlikely to affect the results.
For the 4 types of infections studied by us the following ICD-9 diagnoses were extracted from the patient's medical records. Bronchopneumonia: 481, 481.9, 482.0,482.1, 482.2, 482.3, 482.4, 482.82, 483, 483.9, 484.1, 484.3, 484.6. 485.9, 486, 486.9, 487.0, 507.0, 507.1, 516.8, 518.3. Acute & Strep pharyngitis: 0.34.0, 0.34.1, 034.9, 038.0, 463, 462. Cellulitis; 682.0, 682.3, 682.6, 682.7, 682.9, 681, 376.01, 380.1, 478.1, 478.71, 528.3, 528.5. UTI: 599.0. Aphthous stomatitis; 528. Infections were included if there was no identical diagnosis during the prior 2 weeks.
We chose to combine acute pharyngitis with streptococcal pharyngitis rather than authenticated streptococcal pharyngitis because physicians commonly document acute pharyngitis before having results of throat culture and do not document another diagnosis.
We present the results for aphthous stomatitis separately as unlike the other infections, this is inflammatory lesions associated with a T-cell response.
The poverty index (socioeconomic status [SES] level) is based on several parameters, including household income, educational qualifications, crowding, material conditions, car ownership, census of population and housing. SES levels range between 1 (lowest) and 10 (highest).
There was no patient or public involvement in this retrospective cohort study, which was totally based on electronic medical records.
3 STATISTICAL METHODS
We recorded women's infection rate in all time periods described above. Infection rates were compared between the study periods using χ2 test. The percent of women with these infections were calculated. Results were stratified by OC type (combined vs. progesterone only). In addition, results were stratified by age into 3 subgroups at the time of OC initiation (cohort 1) and discontinuation (cohort 2); 16–23, 24–31, 32–40 years.
To further strengthen the analysis- 4 random year time periods amongst women included in the cohorts above were tested and infection rates in the first vs. last 2 years were compared.
We calculated the number needed to harm as 1/absolute risk increase ×100.
3.1 Nomenclature of targets and ligands
Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, and are permanently archived in the Concise Guide to PHARMACOLOGY 2019/20 (Alexander et al., 2019 a,b).
4 RESULTS
The mean age of the women included in Cohort 1 was 24.78 (standard deviation 6.1), and 11.5% were smokers (Table 1). The mean age of women included in Cohort 2 was 29.7 (standard deviation 5.1), 11.6% were smokers. Their SES is detailed in Table 1.
| Characteristic | Cohort 1 | Cohort 2 |
|---|---|---|
| n = 21 595 periods | n = 20 728 periods | |
| Age (y), mean ± SD | 24.78 ± 6.15 | 29.73 ± 5.1 |
| 16–23 (n = 10 377) | 16–23 (n = 2.409) | |
| 24–31 (n = 7982) | 24–31 (n = 11.830) | |
| 32–40 (n = 3236) | 32–40 (n = 6482) | |
| SES, mean ± SD | 6.1 ± 1.6 | 6.9 ± 1.6 |
| Smoking status | Smoker (n = 2484, 11.5%) | Smoker (n = 2420, 11.6%) |
| Nonsmoker (n = 16 150, 74.7%) | Nonsmoker (n = 16 831. 81.1%) | |
| Stopped (n = 237,1%) | Stopped (n = 344, 1.6%) | |
| Unknown (n = 2724. 12.6%) | Unknown (n = 1228. 5.9%) |
- SD, standard deviation; SES, socioeconomic status
We found a statistically significant higher relative risk of infection while using oral contraceptives (Table 2a); the overall RR (CI) for infection in Cohorts 1 and 2 were 1.35 (1.32–1.38) and 1.27 (1.24–1.31) respectively. In Cohort 1, 23.12% had at least 1 infection during the 2-year period prior to OC use while 29.08% had at least 1 infection during OC use. In Cohort 2, during OC use, 27.09% had at least 1 infection as compared to 22.33% after discontinuation (Table 2a). These percentages are presented because a woman can suffer from recurrent infections. Thus, the number needed to harm is 16.8 and 21 in Cohorts 1 and 2 respectively. The risk remained statistically significant in all infections examined except cellulitis in cohort 2. The RR for aphthous stomatitis in Cohorts 1 and 2 were 1.71 (1.44–2.03) and 2.05 (1.69–2.47) respectively (Table 2b).
| Cohort 1 (n = 21 595, 192 women participated with 2 different time periods) | Cohort 2 (n = 20 728, 155 women participated with 2 different time periods) | |||||
|---|---|---|---|---|---|---|
| Infection type | Infection cases before OC use (n) | Infection cases during OC use (n) | RR (CI) | Infection cases during OC use (n) | Infection cases after OC cessation (n) | RR (CI) |
| The number and % of subjects infected | The number and % of subjects infected | RR | The number and % of subjects infected | The number and % of subjects infected | RR | |
| Overall infection rate | 7016 | 9491 | 1.35(1.32–1.38) | 8199 | 6426 | 1.27 (1.24–1.31) |
| 4867 | 6100 | 5455 | 4533 | |||
| 22.5%a | 28.2%a | 26.3%a | 21.8%a | |||
| Acute pharyngitis | 3410 | 3670 | 1.07 (1.03–1.12) | 3384 | 2672 | 1.26 (1.29–1.32) |
| 2678 | 2916 | 2653 | 2210 | |||
| 12.4% | 13.5% | 12.8% | 10.7% | |||
| Pneumonia | 396 | 466 | 1.17 (1.03–1.34) | 559 | 370 | 1.51 (1.32–1.70) |
| 338 | 399 | 482 | 313 | |||
| 1.6% | 1.8% | 2.33% | 1.5% | |||
| Cellulitis | 177 | 220 | 1.24 (1.02–1.51) | 224 | 188 | 1.19 (0.98–1.44) |
| 169 | 208 | 216 | 175 | |||
| 0.7% | 0.9% | 1% | 0.8% | |||
| UTI | 3033 | 5135 | 1.69 (1.62–1.76) | 4032 | 3196 | 1.26 (1.20–1.31) |
| 2181 | 3324 | 2722 | 2297 | |||
| 10.1% | 15.4% | 13.1% | 11.1% | |||
- a A woman was counted only once if she had 1 or more infection types or episodes during the study period.
- CI, confidence interval; OC, oral contraceptive; RR, risk ratio; UTI, urinary tract infection
| COHORT 1 (n = 21 595) | COHORT 2 (n = 20 728) | |||||
|---|---|---|---|---|---|---|
| Infection type | Infection episodes before OC use (n) | Infection episodes during OC use (n) | RR (CI) | Infection episodes during OC use (n) | Infection episodes after OC cessation (n) | RR (CI) |
| % of women infected | % of women infected | RR | % of women infected | % of women infected | RR | |
| Aphthous stomatitis | 208 | 357 | 1.71 (1.44–2.03) | 322 | 157 | 2.05 (1.69–2.47) |
| 0.008% | 0.014% | 1.64 (1.38–1.96) | 0.013% | 0.007% | 1.96 (1.61–2.39) | |
The overall infection risk remained statistically significant when stratified by age (Table 3). A higher relative risk 1.55 (1.50–1.61, P < .01) and 1.36 (1.27–1.46, P < .01) in Cohorts 1 and 2 respectively) were found in both cohorts for overall infection among the young age group (ages 16–23). However, the risk was not homogeneously significant for all types of infections: when looking at specific infections in Cohort 1, the risk for acute pharyngitis on the old age group (32–40), pneumonia in the middle age group (24–31) and cellulitis in middle and old age groups were not statistically significant (Table 3). For aphthous stomatitis in the old age group the risk was also not statistically significant (Table 3).
| Cohort 1 | Cohort 2 | |||||
|---|---|---|---|---|---|---|
| Age (y) | 16–23 | 24–31 | 32–40 | 16–23 | 24–31 | 32–40 |
| Infection type | RR (CI, P) | |||||
| All infections | 1.55 (1.50–1.61, P < .01) | 1.21 (1.16–1.26), P < .01) | 1.08 (1.02–1.15, P < .01) | 1.36 (1.27–1.46, P < .01) | 1.22 (1.17–1.26, P < .01) | 1.33 (1.27–1.40, P < .01) |
| Acute pharyngitis | 1.11 (1.05–1.17, P < .01) | 1.10 (1.02–1.19, p < .01) | 0.89 (0.79–0.99. P = .65) | 1.46 (1.27–1.68, P < .01) | 1.16 (1.08–1.23, P < .01) | 1.38 (1.27–1.49, P < .01) |
| Pneumonia | 1.39 (1.10–1.74, P < .01) | 0.95 (0.76–1.18, P = .08) | 1.27 (0.98–1.64, P = .06) | 1.27 (0.82–1.94, P = .27) | 1.33 (1.10–1.60, P < .01) | 1.81 (1.48–2.21, P < .01) |
| Cellulitis | 1.44 (1.07–1.94, P < .01) | 1.13 (0.81–1.58, P = .44) | 1.02 (0.65–1.60, P = .90) | 1.35 (0.75–2.40, P = .31) | 1.24 (0.95–1.60, P = .1) | 1.06 (0.75–1.49,p = .7) |
| UTI | 2.25 (2.11–2.39, P < .01) | 1.33 (1.24–1.41, P < .01) | 1.28 (1.14–1.43, P < .01) | 1.32 (1.19–1.46, P < .01) | 1.25 (1.19–1.32, P < .01) | 1.23 (1.13–1.34, P < .01) |
- CI, confidence interval; OC, oral contraceptive; RR, risk ratio; UTI, urinary tract infection
When stratifying by OC type in Cohort 1 (Table 4), progesterone-only pills showed a slightly higher, statistically significant overall relative risk for infection; 1.15 (1.007–1.32, P = .03). Also, a statistically significant high relative risk was detected for UTI (RR 1.31 [95% CI 1.07–1.59], P < .01) and cellulitis 2.66 (95% CI 1.04–6.81, P = .04). The relative risk for UTI while using progesterone pills was lower than the 1 detected while consuming combined OCs (1.23 vs. 1.71, Table 2a). Conversely, the relative risk for cellulitis while using progesterone only pills was higher than the relative risk during use of combined OCs (2.66 [95% CI 1.04–6.81], P = .04 vs. 1.19 [95% CI 0.97–1.46], P = .08; Table 2a). In Cohort 2, progesterone-only pills showed a statistically significant high relative risk for overall infections (1.33 [95% CI 1.16–1.53], P < .01) acute pharyngitis, UTI and bronchopneumonia (Table 5). The relative risk was higher in bronchopneumonia.
| Infection type | Combined OCs | Progesterone only OCs | ||||
|---|---|---|---|---|---|---|
| Infection episodes before OC use- (n) | Infection episodes during OC use- (n) | RR (CI, P) | Infection episodes before OC use-(n) | Infection episodes during OC use- (n) | RR (CI, P) | |
| Overall infection rate | 6644 | 9061 | 1.36(1.32–1.39, P < .01) | 371 | 430 | 1.15(1.007–1.32, P = .03) |
| Acute pharyngitis | 3251 | 3518 | 1.08 (1.03,1.13, P < .01) | 158 | 151 | 0.94 (0.76–1.18, P = .6) |
| Pneumonia | 365 | 434 | 1.18 (1.03–1.36, P < .01) | 31 | 32 | 1.03 (0.63–1.69, P = .89) |
| Cellulitis | 171 | 220 | 1.19 (0.97–1.46, P < .01) | 6 | 16 | 2.66 (1.04–6.81, P = .04) |
| UTI | 2857 | 4904 | 1.71 (1.64–1.79, P < .01) | 176 | 231 | 1.31 (1.07–1.59, P < .01) |
- CI, confidence interval; RR, risk ratio; UTI, urinary tract infection
| Infection type | Combined OCs | Progesterone only OCs | ||||
|---|---|---|---|---|---|---|
| Infections during OC use- (n) | Infections after OC discontinuation (n) | RR (CI, P) | Infections during OC use- (n) | Infections after OC discontinuation (n) | RR (CI, P) | |
| Overall infection rate | 7740 | 6082 | 1.27 (1.24–1.31, P < .01) | 459 | 344 | 1.33 (1.16–1.53, P < .01) |
| Acute pharyngitis | 3185 | 2526 | 1.26 (1.20–1.32, P < .01) | 199 | 146 | 1.36 (1.10–1.68, P < .01) |
| Pneumonia | 520 | 352 | 1.47 (1.29–1.68, P < .01) | 39 | 18 | 2.16 (1.23–2.16, P < .01) |
| Cellulitis | 214 | 176 | 1.21 (0.99–1.48, P = .05) | 10 | 12 | 0.83 (0.83–1.92, P = .67) |
| UTI | 3821 | 3028 | 1.26 (1.20–1.31, P < .01) | 211 | 168 | 1.25 (1.02–1.53, P = .01 |
- CI, confidence interval; RR, risk ratio; UTI, urinary tract infection
The overall infection incidence was 162.4 and 155 per 1000 women years when not consuming OCs and 219.7 and 197.8 during OC consumption in Cohorts 1 and 2 respectively. Specific infection rates are shown in Table 6.
| Cohort 1 | Cohort 2 | |||
|---|---|---|---|---|
| Infection type | No OC use | OC use | OC use | NO OC use |
| Overall infection rate | 162.4 | 219.7 | 197.8 | 155 |
| Acute pharyngitis | 79 | 85 | 81.7 | 64.4 |
| Pneumonia | 9.2 | 10.8 | 13.5 | 8.9 |
| Cellulitis | 4.1 | 5.1 | 5.4 | 4.5 |
| UTI | 70.2 | 118.9 | 97.3 | 77.1 |
- OC, oral contraceptive; UTI, urinary tract infection
When comparing infection rates between random 2-year consecutive periods, without relation to their OC consumption, among women who were included in Cohorts 1 and 2, there was no difference in overall infection rate (Table 7). When looking at specific infections, no differences were found in all infections except UTI in which a statistically significant, although clinically negligible, difference was detected (RR = 0.94, CI -0.90-0.98, P = .01).
| Infection | RR (CI, P) |
|---|---|
| Overall infection rate | 0.99 (0.97–1.01, P = .87) |
| Acute pharyngitis | 0.99 (0.96–1.02, P = .83) |
| Pneumonia | 0.94 (0.93–1.17, P = .42) |
| Cellulitis | 1.04(0.87–1.25, P = .6) |
| UTI | 0.94 (0.90–0.98, P = .01) |
| Aphthous stomatitis | 1.1 (0.92–1.32, P = .24) |
- CI, confidence interval; RR, risk ratio; UTI, urinary tract infection
5 DISCUSSION
The results of the present study demonstrate a statistically and clinically significant risk for the infections followed; Acute pharyngitis, bronchopneumonia, UTI, cellulitis and for aphthous stomatitis. This is consistent with our recent preliminary study which has shown that women who used OCs were at greater risk for developing neutropenia. It is also consistent with studies showing a higher rate of nasal S. aureus carriage 17, 18 and with in vivo studies demonstrating that oestrogen reduces neutrophil responsiveness.14
The number needed to harm (16.8 and 21 in Cohort 1 and 2) reflects the commonness of infections caused by OCs, especially when taking into account the high consumption rate. The true number is probably higher as the present number does not consider women who suffered from recurrent infections while using OCs.
Neutrophils are the most abundant cellular component of the immune system and are the first line of defence against invading microorganisms. Defects in neutrophil number and function commonly present with recurrent infections. Neutropenia is usually defined as absolute neutrophil count <1500/μL, causing moderately higher infection rates when neutrophil levels are <1500/μL and significantly higher infection rate when <500/μL.25 Since neutropenia is not a common finding and the relative risk for neutropenia among women consuming OCs in our recent study was 1.2 (95% CI 1.1-1.35, P < .01), whilst the infections examined here are relatively common, it suggests that OC use affects not only the circulating numbers of neutrophils but probably also their functions.
Younger women appear to have a higher relative risk for infections when using OCs (Table 3), although, when looking at specific infections, no obvious trend is seen. This difference might be due to a lower baseline infection rate among young users or to different hormonal states in different age groups, altering the OC's hormonal effects.
The infection-specific relative risks in Cohorts 1 and 2 are relatively different in most infections, although all are statistically significant (Table 2a). This might be explained by the fact that women in Cohort 2 are 5 years older than women in Cohort 1, putting them at a different baseline risk for different infections, with OC use further expanding the gap. The age difference might also affect behavioural aspects such as rates of sexual intercourse. Furthermore, infection rates in the second cohort were recorded during the last 2 years of OC use while those in Cohort 1 were recorded during the first 2 years of use. It is possible that the duration of use affects the association between OC and specific infection risk, similar to OCs' time-related effect on venous thromboembolism.34-36 Lastly, every infection studied here involves a different tissue, which may be affected differently in aspects other than neutrophil count and functionality, but rather by changes in microbiota or other local immune properties, which may change over time.
In pharmacoepidemiology, one has to ensure that comparison before vs. after exposure does not introduce immortal bias. We considered potential confounders that may bias the results. We have identified women who were prescribed any anti epileptic and antipsychotic drugs.
To further establish the relevance of our data, we compared the incidence rates of the infections recorded in our data to incidence rates reported for the same age groups in other studies. We selected 4 types of infections, which are sufficiently common to be able to get interpretable signals.
According to a population-based study conducted in Israel in 2018, the incidence of UTI among adults aged 19–44 was 57 per 100 person years, comparable to the rate found in our data. In general, women are more susceptible to UTIs than men and this study's rate, which is lower, refers to both men and women.9 The incidence rate for bronchopneumonia varies widely among different studies10-13 ranging between 2 and 23 cases per 1000 people years. The incidence found in our data lies in this range. The rate of acute pharyngitis found in our data is similar to the rate shown in a recent population-based study in a different Health Maintenance Organization in Israel.37 Among people aged 25–44 years in Australia in 2013,38 156 per 100 000 years were diagnosed with lower limb cellulitis, which accounted for only some of the cellulitis cases. We could not find studies illustrating the incidence of all cellulitis infections among young patients, although it seems that the incidence in our data is in agreement with that found above. Aphthous stomatitis is a common pathology with a prevalence of around 25%, although it is commonly not documented since it is mostly treated with over-the-counter medications; therefore, we cannot appreciate the relevance of this diagnosis, although the proportion between the time periods should not be affected,
Our results suggest the relative risk for UTI during OC use is 1.69 and 1.26 in cohorts 1 and 2 respectively (Table 2a). Since sexual activity is known to be a major risk factor for UTI, and is positively associated with OC use, one can assume that sexual activity is accountable for at least some of the rise in incidence of UTI. However, the results raise questions regarding the possible contribution of OC use to the high relative risk shown in many studies, especially due to the obvious high relative risk for all other infections investigated. Pyelonephritis, which is usually preceded by UTI, is a common cause of hospitalization among young females, causing significant short- and long-term adverse outcomes. Complications include sepsis, renal abscess, acute renal failure, renal impairment and, rarely, death38. Due to the high relative risk for UTI during OC use among this already common infection, even low incidence of pyelonephritis among this group is significant. In addition, in Table 2a, the results show differences between the relative risk for UTI episode vs. the relative risk for at least 1 infection (RR 1.69 vs. 1.52 in Cohort 1, RR 1.26 vs. 1.18 in Cohort 2). This reflects the numbers of women with recurrent infections contributing to the overall relative risk. These results suggest that the relative risk for UTI strongly reflects the risk for recurrent infections and that women with a history of UTI are at risk for recurrent UTIs.
In the present study, the relative risk for bronchopneumonia while using OCs was 1.17 (95% CI 1.03–1.34, P < .01) and 1.51 (95% CI 1.32–1.70, P < .01) in Cohorts 1 and 2, respectively (Table 2a). Bronchopneumonia is among the most common infections leading to hospitalization. However, it is not common among young patients with patients younger than 40 years accounting for ~17% of bronchopneumonia cases, with mostly mild severity and with mortality of <1%39. However, complications do occur and hospitalization rates among young patients are relatively high despite mild illness.39, 40 The apparent rise in the incidence of bronchopneumonia among women consuming OCs is translated to elevated antibiotic use, hospitalizations and, rarely, severe complications.
Oral aphthous ulcers are a common complaint with an estimated point prevalence of 4%40. In our recent study,3 we described the case that led us to investigate this topic. This patient suffered from recurrent aphthous ulcers throughout her 7-year period of OC use, which resolved after discontinuing OCs. The relative risk for aphthous stomatitis found in this study is 1.71 (95% CI 1.44–2.03, P > .01) and 2.05 (95% CI 1.69–2.47, P < .01) in Cohorts 1 and 2, respectively (Table 2b). Several studies have documented psychological impact in patients suffering from recurrent aphthous stomatitis with lower quality of life 41-43 and higher rates of anxiety and depression.41 Due to their effect on quality of life, this alone can be a reasonable cause to discontinue OC use.
Acute pharyngitis is among the most common complaints in community medicine, accounting for 1–2% of ambulatory care visits in the USA.43 Of these, 5–20% are caused by streptococcal pharyngitis.43 The relative risk for acute pharyngitis found in this study is 1.07 (95% CI 1.03–1.12, P < .01) and 1.26 (95% CI 1.29–1.32, P < .01) in Cohorts 1 and 2, respectively (Table 2a). Complications include rheumatic fever, retropharyngeal or peritonsillar abscesses, post-streptococcal glomerulonephritis and arthritis, all of which can have major long-term consequences. The increased risk for acute pharyngitis under OC will result in an elevated risk for these complications.
Cellulitis is the most common soft tissue infection and is usually caused by bacterial flora colonizing the skin, such as S. aureus or streptococci. Nasal S. aureus colonization, which has been shown to occur twice as much in women consuming OCs,17, 18 is a known risk factor for S. aureus infections and specifically cellulitis.15, 44-46 A high relative risk for cellulitis was detected in Cohort 1 (RR = 1.24; 95% CI 1.02–1.51, P < .01) and in Cohort 2 (RR = 1.19; 95% CI 0.98–1.44, P = .07). However, it did not reach statistical significance. This is consistent with this finding, with complications such as sepsis, gangrene and amputation.
The potential effect of progesterone OC in our study was marginally significant; Progesterone-only pills showed a slightly higher, statistically significant overall relative risk for infection. Indeed, there is new evidence that progesterone-based compounds alter cellular signalling and activity to affect the outcome of infections at diverse mucosal sites.24
Several potential challenges in this study need to be acknowledged and addressed.
When comparing adverse effects before vs. after an index event, one needs to address the possibility of immortal bias. Here this would tend to underestimate the magnitude of the association found by us, which would mean that the true value of the risk is even higher. By using 2 different cohorts, addressing the before and after issue, we minimize the potential effect of immortal bias. We extensively considered other potential confounders; we excluded women who were prescribed any antiepileptic agents, antipsychotics or immunosuppressants due to their known risk for neutropenia. In addition, we excluded women who were pregnant any time during the 2 years prior to OC initiation (Cohort 1) or after discontinuation (Cohort 2) due to pregnancy's known effects on neutrophil counts. Women who changed the type of OCs (combined vs. progesterone only) throughout this time period were also excluded. Overall, we excluded a very small number of women changing their OC: from Cohort 1, 92 were excluded out of a total number of 21, 595 (0.08%); and from Cohort 2, 155 out of a total of 20 728 (0.07%). These exclusions accounted for <1% of the cohorts and hence could not affect the generalizability of our results.
We also considered potential changes in lifestyle in women who start using the pill, mostly being sexually active. However, a woman can be and is probably sexually active before and after using the OC, and as we did not have direct information about sexual activity, it could not be captured in models such as multivariable analysis.
Our stratification suggests potential confounding by age. The potential age effect on infection is the very reason we repeated the analysis on each subgroup of ages. This analysis reveals that we did not have a biased result, but rather that the effect has to be adjusted for age. It can be argued that one cannot investigate 4 years of follow up in patients without accounting for ageing which is a strong confounder for infections. However, if 2 years plus 2 years would cause per se a significant age effect on infection rates, then there should be a similar effect when we compared different random samples over 4 years; this was not the case.
The paired comparison of the same women during treatment and then after treatment, or before treatment and then during treatment, obviates the need to address potential confounders among patients, because the woman serves as her own control.
The strength of the association between OC and infections deserves a discussion. We are showing a RR (95%) of 1.35 (1.32–1.38) in Cohort 1 and 1.27(1.243–1.31) in Cohort 2. This constitutes 35 and 27% increase in all serious infection rates measured, which would translate to millions of more infected women worldwide. Moreover, the 95% CIs of both RRs are extremely narrow with P < .00001.
6 CONCLUSIONS
Women at high risk for infections such as the immunosuppressed, diabetic or those consuming medications with a risk for immunosuppression, may be at a particularly high risk for infections when using OCs. The clinical implications discussed above would result in a high economical burden due to higher medication use, hospitalizations, loss of work days and long-term sequelae. While our study considered 5 common infections, it is conceivable that similar mechanisms will increase the rates of other serious infections secondary to the use of OCs. These too will have to be studied further.
As with any new adverse drug reaction signal, these results should be examined by other groups and critically, by the manufacturers of OCs, especially in view of large existing basic research on such effects. More caution should be practiced in women with higher risks of infection due to systemic conditions. Then, one may consider in more detail a comparison of the risk by different OCs, especially as it seems that the risk with progesterone-only is substantially lower.
OCs are used extensively in almost all populations, among other aspects, due to their perceived naive adverse effect profile. Hence, increasing risk of infections is likely to have major clinical and economic implications. More research is required, both epidemiological and biological, to further understand the mechanisms and implications of the increased rates of infections during OC use. Although OCs had a substantial positive clinical and economical impact on women over recent decades, these findings may have major implications to millions of women worldwide and this should lead to renewed discussion on the safety of the pill.
COMPETING INTEREST
None of the authors declare any competing interests.
CONTRIBUTORS
Y.R. conceived the project and wrote the first draft; A.R. conducted statistics; H.R. and S.R. conducted investigations and analysis; G.H. oversaw the analysis; G.K. oversaw the project and cowrote the first draft.
RESEARCH ETHICS APPROVAL
The study was approved by Assuta Hospital's Research Ethics Committee (#96–8010)
STRENGTHS AND LIMITATIONS OF THIS STUDY
- A major strength is being a large population-based cohort.
- Our study demonstrated incidence rates similar to rates found in other studies, corroborating the reliability of our data.
- A potential weakness is that some women might have had infections that were not documented in the patient's electronic health record; however, suboptimal documentation of these infections should have spread evenly between the 2 study periods.
Open Research
DATA AVAILABILITY STATEMENT
Data available on request due to privacy/ethical restrictions.
