Summary
Background Long COVID and myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) disproportionately affect females, suggesting modulation by sex hormones. We sought to investigate whether symptom severity is influenced by changes in sex hormones over the menstrual cycle, or by hormonal contraception.
Methods We carried out a retrospective analysis of menstrual and symptom data, prospectively collected via the Visible app from individuals with long COVID, ME/CFS, or both, who had regular menstrual cycles, between 7th September 2022 and 6th March 2024. Mixed-effects models were used to examine associations between symptom severity, menstrual cycle phase and contraception type.
Findings 948 users were included; 100% of users were female and 92.6% identified as women. The most tracked symptoms were fatigue (99·5% of users), brain fog (88·3%), headaches (85·1%) and muscle aches (78·6%). All menstrual cycle phases showed a modest, but significant, improvement compared to the menstrual phase, most markedly the early luteal (IRR = 0·963, 95% CI: 0·958–0·968), but also the follicular (IRR = 0·985, 95% CI: 0·981–0·990) and late luteal phases (IRR = 0·980, 95% CI: 0·974– 0·985). Crashes (sudden and severe worsening of symptoms following exertion) were significantly more frequent during menstruation than in other phases. Users of combined hormonal contraception (n=70) had a statistically significant reduction in overall symptom score (OR = 0·827, 95% CI: 0·690–0·992) and crash incidence (OR = 0·548, 95% CI: 0·350–0·856) compared to those not using contraception (n=786).
Interpretation Menstruation is associated with worsened symptoms in long COVID and ME/CFS. Users of combined hormonal contraception report a lower symptom burden than non-users, suggesting a modulatory role of ovarian hormones. These findings could empower menstruating people living with long COVID and ME/CFS to anticipate cyclical changes in symptoms and plan their activities accordingly, and could also inform their use of contraception.
Funding UK Medical Research Council, NIHR, Community Jameel, the Eric and Wendy Schmidt Fund for Strategic Innovation, Bill & Melinda Gates Foundation, Borne, Action Medical Research, and Genesis Research Trust.
Introduction
Four years after the first identification of novel coronavirus SARS-CoV-2, the spread of coronavirus disease 2019 (COVID-19) continues to profoundly impact people’s lives. Whilst most SARS-CoV-2 infections resolve within 4 weeks, over 10% progress to a chronic multisystemic condition known as ‘long COVID’.1 The Winter COVID-19 Infection study, conducted between November 2023 and March 2024, estimated that 2 million people currently live with long COVID in the UK.2 Of those who reported having long COVID, 71·1% reported having long COVID for over a year, 51·3% over 2 years and 30·6% over 3 years.2 Long COVID symptoms are wide-ranging, but commonly include fatigue, brain fog, shortness of breath, insomnia and post-exertional malaise,3 and can significantly reduce quality of life.4
A substantial fraction of long COVID cases also meet the diagnostic criteria for myalgic encephalomyelitis (ME), also known as chronic fatigue syndrome (CFS).5–8 Whilst not all cases of ME/CFS have an identifiable infectious trigger, many are preceded by a viral infection such as glandular fever,9 or Ross River virus.10 The prevalence of ME/CFS is estimated to be between 0.59% and 1.3%,11–13 and around a quarter of people living with with ME/CFS are housebound. 14 Despite the profound impact ME/CFS can have on quality of life, research has been consistently underfunded.15 Given the historic neglect of ME/CFS and the recent emergence of long COVID, as well as the complexity of both diseases, many questions about their pathophysiology remain unanswered.
Long COVID and ME/CFS both affect approximately twice as many females as males16,17 and higher oestrogen may be beneficial in females with acute COVID-19.18,19 Whilst the pathophysiology of long COVID and ME/CFS are still unclear, a dysregulated immune response may play a role, with the discovery of autoantibodies in long COVID patients supporting the suggestion that long COVID could be an autoimmune disease. 6,20 Sex hormones (oestrogen, progesterone and testosterone) modulate sex differences in immune function21 and phases of the menstrual cycle correlate with symptoms of chronic, and particularly autoimmune, conditions,22 summarised in Figure 1. Symptoms of rheumatoid arthritis, inflammatory bowel disease, lupus, multiple sclerosis and asthma all worsen at or shortly before menstruation.22 Hormonal contraception alters the cyclical fluctuations in ovarian hormones across the menstrual cycle and has an association with altered disease susceptibility and severity, 22 with users experiencing improved symptoms of multiple sclerosis. 23
(A) A schematic representation of the menstrual cycle illustrates relative changes in pituitary and ovarian hormones, follicle development, and endometrial lining. Correlated changes in disease severity and infections are shown in the top panel, red indicates disease symptoms worsen/susceptibility to infection increases and green indicates disease symptoms improve/response to infection improves. Created in Biorender.com.34 (B) Screenshots from the Visible app illustrate how users track individual disease symptoms, menstruation and crashes. Data was collected from 7th September 2022 to 6th March 2024. (C) Flow diagram to illustrate the data cleaning process of users. Red boxes show exclusions of users, blue boxes show the number of users (n) after exclusions. (D) Flow diagram to illustrate the data cleaning process of cycles after (C). Yellow boxes show cycle exclusions, turquoise boxes show the number of cycles (n) after exclusions. (E) The decision tree for assigning menstrual cycle phase to each day.
Cross-sectional surveys have suggested that long COVID symptoms may also vary with the menstrual cycle,3,24 but these approaches can be affected by recruitment and recall bias. Many individuals track both menstrual cycles and disease symptoms using apps, entering data in real-time as part of their healthcare routine. Data collected in this way are less likely to be affected by recruitment bias than other methods of data collection.25 Visible is an app designed for people with chronic health conditions (such as long COVID and ME/CFS) to track aspects of their health including symptoms, biometric measurements and other factors such as menstruation.26 In this study, we used data from the Visible app to test two hypotheses: 1) Long COVID and ME/CFS symptoms are worse during the menstrual and premenstrual phases, improving in the mid-luteal phase, and 2) stabilising ovarian hormone levels with hormonal contraception is associated with improved symptoms.
Evidence before this study
In acute COVID-19, there is evidence to suggest that oestrogen may have a protective effect. A Swedish cohort study of post-menopausal women infected with SARS-CoV-2 (Sund et al., 2022) found that women with high oestrogen had a lower risk of death compared to controls and women with low oestrogen had higher risk of death compared to controls. Furthermore, a population-based matched cohort study of UK women found that users of the combined hormonal contraceptive pill (n=295,689) had lower rates of COVID-19 and a reduction in hospital attendance (Costeira et al., 2021). To our knowledge, the impact of the menstrual cycle on ME/CFS symptoms had yet to be investigated prior to this study and research into long COVID and the menstrual cycle to date has been limited. Arguably one of the most influential studies in the general field of long COVID is from Davis et al. 2021 who reported the results of a large international patient-led survey. In addition to characterising symptom prevalence, Davis et al., found that a third of menstruating participants (n =1792) reported relapses of their long COVID symptoms immediately before or during their period. In a smaller UK survey (Newson et al., 2021, n = 460), 62% of female respondents with long COVID reported that their symptoms of long COVID were worse on the days before their period. However, the participant demographics were not reported, so it is unclear whether this is a representative sample.
Added value of this study
This study is the first to describe changes to long COVID and ME/CFS symptoms across the menstrual cycle, using prospectively collected app data. Symptoms tended to be worse during the menstrual phase and less severe during the early luteal phase, in line with our hypothesis and with previous findings. This study is also the first to investigate the effect of hormonal contraception on symptoms of long COVID and ME/CFS. Use of combined hormonal contraception was associated with significantly ameliorated symptoms and a reduced risk of crashes.
Implications of all the available evidence
The finding that long COVID and ME/CFS symptoms follow a cyclical pattern over the menstrual cycle not only gives us new insights into the role of sex hormones in these diseases but could also empower menstruating patients to schedule more strenuous activities during phases of the cycle when they are likely to have milder symptoms, and be less vulnerable to crashes (sudden and severe worsening of symptoms following exertion). The finding that combined hormonal contraceptives as associated with reduced symptoms could also inform the use of contraception in people living with long COVID and ME/CFS.
Methods
Data source and collection
A retrospective analysis of symptom severity and menstrual cycle phase was conducted on prospectively collected data from users of the Visible app. 26 On registration, users enter their birth year, gender and whether they consider themselves to have long COVID, ME/CFS or another condition. After consenting to participate, users were asked to complete a survey on their use of hormonal contraception, pregnancy status, breastfeeding status, number of pregnancies, and number of births. Each day, users were prompted to self-report their commonly experienced symptoms on a 4-point Likert scale [None, Mild, Moderate, Severe] and whether they are currently experiencing a crash (appendix, p23). The following definition was provided: “Crashes usually occur as a part of post-exertional symptom exacerbation and affect your ability to carry out your usual activities. They normally last a few days.’’ Reports entered between 7th September 2022 to 8th March 2024 were collected.
Data cleaning and menstrual cycle phase coding
All data cleaning and subsequent analysis was conducted in RStudio version 2024.12.0+467. 27 Users who did not complete the survey questions, or who were breastfeeding, pregnant or outside the age range 18-45 were excluded. Under-18s were not included due to the additional ethical approval that would have needed to be obtained when analysing data from children; over 45s were excluded due to an increased likelihood of being perimenopausal. Any users who did not have either long COVID or ME/CFS, who did not track menstrual data or who did not report a single day where they were menstruating were also excluded. A menstrual cycle was counted from the first day of one menstrual period to the first day of the next. The first day of a period was assigned based on the International Federation of Gynecology and Obstetrics (FIGO) definition of a normal period28 where the day fulfilled all the following criteria: users reports being on period; user did not report being on their period the previous 5 days; the following day and the day after the user either reported being on their period or did not make a report; user did not report being on their period 9 and 10 days after.
Menstrual cycles that contained >2 consecutive days of missing menstrual tracking data or were outside of the normal length of a cycle (24-38 days, as defined by FIGO28) were removed from the dataset. Exclusion of users and cycles is summarised in Figure 1C-D. For all remaining cycles, any missing menstrual tracking data was imputed (‘on period’ or ‘not on period’) so that menstrual cycle phases could be assigned. Data was imputed from the previous day with data, with the following exception: missing data on the 9th or 10th day of the cycle was imputed as ‘not on period’, assuming duration of menses does not exceed the normal 8 days. 28 The period end date was assigned where the day fulfilled all the following criteria: user assigned as being on their period; user assigned as being on their period at least the previous 2 days; user assigned as not being on their period the following day.
After defining the last day of the period, a menstrual cycle phase was assigned to each day, according to the decision tree outlined in Figure 1E. This follows the assumption that the follicular phase is variable, and the luteal phase consistently accounts for the final 14 days of each cycle.29 The luteal phase was further divided into the early luteal (14-8 days before the first day of period (D1), late luteal (7-3 days before D1) and premenstrual phase (2 days before D1).
Once menstrual cycle phases had been assigned to each day of all included cycles, any symptom and sleep tracking associated with that day were integrated into the dataset. Retrospective symptom reporting (symptom reported >2 days after date) was removed to reduce the risk of recall bias.
Descriptive statistics and statistical testing
Overall study population characteristics are summarised in Table 1. For each user, for each date, the scores of each tracked symptom (all on a scale of 0-4) were summed to give an overall symptom score and provide a single, comprehensive measure of a user’s overall health, reflecting the cumulative burden of symptoms. Incorporating the total number of symptoms experienced in addition to severity of symptoms is important, as for many individuals with long COVID, navigating the volume of symptoms feels overwhelming and unmanageable.30 Gravidity was assessed as a binary variable (nulligravid or multigravid), and the Mann-Whitney U-Test was used to test the association between gravidity and mean symptom score. Kruskal-Wallis tests were used to assess the association between mean symptom score and disease type, contraception type, quarters post-joining and age category. Post-hoc pairwise comparisons were conducted with Dunn’s tests, and p-values were adjusted using the Bonferroni method to control for Type 1 error. For menstrual cycle phase, the mean of all overall symptom scores across all dates in each cycle phase was calculated for each individual. The Friedman test was used to compare symptom score means across phases. Post-hoc pairwise comparisons were performed using the Nemenyi test. For R package details see appendix (p23). For all variables described above, the mean of the mean overall symptom scores is presented as scatter plots in Figure 2, with pairwise significant comparisons shown on the plot (p<0.05). The percentage of days classified as a crash was calculated out of the total days with data and displayed as a scatter plot with mean +/-standard error of the mean (SEM).
(A) Distribution of mean symptom scores is shown as a histogram, binwidth = 0·5. (B-G) Mean symptom score was calculated by first summing the scores of each tracked symptom (all on a scale of 0-4) and secondly taking the mean across all days for each individual. Scatter plots show mean symptom score split by menstrual phase (B), disease type (C), contraception type (D), quarters post-joining (E), gravidity (F) and age category (G). Mean symptom score is displayed as mean +/-SEM. The Friedman test plus Nemenyi pairwise comparisons was used to test the significance between mean symptom score and menstrual phase, due to the repeated measures nature of the data. The Mann-Whitney U-Test was used to test the significance between gravidity and mean symptom score, due to the binary nature of the variable. Unpaired Kruskal-Wallis tests was used to test significance for all other variables, Dunn’s test was used for post-hoc comparisons and p values were adjusted using the Bonferroni method. P > 0.05 are considered not significant, P ≤ 0·05 are given 1 asterisk (*), P ≤ 0·01 are given 2 asterisks (**), P ≤ 0·001 are given 3 asterisks (***), P ≤ 0·0001 are given 4 asterisks (****). (H) Scatter plot showing mean percentage of days classified as a crash, displayed as mean +/-SEM.
Mixed-effects modelling
Two separate generalized linear mixed-effects models were used to evaluate the effect of relevant variables on different outcomes: overall symptom score and crash occurrence. First, a negative binomial regression mixed-effects model was used to investigate overall symptom score. Fixed effects included menstrual phase, age category, disease type, contraception type, nulligravidity, and time since joining. ‘Individual’ was considered a random effect, which is recommended in menstrual cycle research31 to reliably estimate between-person differences in within-person changes across the cycle.31–33 Second, to evaluate the effect of the same set of variables on crash occurrence, a logistic regression (binomial family with a logit link) mixed-effects model was used. Fixed effects included menstrual phase, disease type, contraception type, quarters post-joining, nulligravidity and age category. ‘Individual’ was considered a random effect, to account for within-subject variability. Details of both models can be found in appendix (p24-25).
Ethics
All app users were given the option to consent to the use of their anonymised data for this study by checking a box in the “Research” section of the Visible app. Ethical approval for this study was obtained from the Research Governance and Integrity Team (study number 277771) at Imperial College London.
Role of the funding source
The funders of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report.
Results
Of the 2678 users who consented for their data to be used in the study period (7th September 2022 – 6th March 2024), 948 remained after data cleaning (Figure 1C-D, Table 1). Just under half the cohort identified as having long COVID (45·5%), around a third as having ME/CFS (34·7%) and a fifth (19·8%) as having both long COVID and ME/CFS. Almost all users identified as women (92·6%) and as not disabled (99·6%). The majority of users were not on hormonal contraception (82·9%), and the remainder were split fairly evenly between oestrogen and progestin combined hormonal contraception (combined pill (7·0%) or patch (0·4%), hereafter called “combined hormonal contraception”) and progestin-only contraceptives (intrauterine system (5·5%) or progestogen-only pill (2·3%) or contraceptive implant (1·9%)). Around two-thirds of users had never been pregnant (65·1%) or given birth (72·5%). The median user age was 37 (interquartile range = 10, distribution shown in appendix (p3). There were no significant differences between the characteristics of study participants in each of the disease groups appendix (p3).
To determine how overall disease severity changed over the cycle and with demographic characteristics, we examined the daily sum of all symptom ratings, the ‘symptom score’, with higher scores indicating higher severity. For each individual, we calculated the mean symptom score across all relevant days. The mean symptom score was lowest in the early luteal phase, significantly higher in the follicular (p=0·0043) and late luteal phase (p= 0·0037) and still higher in the menstrual (p=1·5e-08) and premenstrual phases (p= 4·6e-05) (Figure 2B). Mean symptom scores were highest in users with both long COVID and ME/CFS and were lowest in users with long COVID alone (Figure 2C). Mean symptom scores from users with different contraception types, time since joining, gravidity and age categories did not significantly differ in this initial analysis (Figure 2D-G), although there was a near significant difference (p=0.076) between users of no hormonal contraception compared to combined hormonal contraception (p=0·076) (Figure 2D).
On multivariable mixed-effects regression modelling, as symptom score was positively skewed and over-dispersed (Figure 2A), we used negative binomial regression and incorporated random intercepts for each user. The association between menstrual cycle phase and symptom score remained significant adjusting for contraception type, disease type, age category, gravidity and time since joining the app (Table 2). Symptom scores were significantly higher during menstruation than in all other cycle phases. Compared to the menstrual phase, the early luteal phase showed the greatest reduction in symptom scores (incidence rate ratio (IRR)=0·963, p=5·77E-46); the follicular and late luteal phases also showed significant reductions (IRR=0·985, p=6·49E-09 and IRR=0·980, p=4·96E-13, respectively); and the smallest reduction was observed in the premenstrual phase (IRR=0·992, p=0·0425). In the regression model, symptom scores were significantly lower in users of combined hormonal contraception (IRR=0·872, p=0·041), in comparison to users not on hormonal contraception. Users on progestin-only contraception also had lower symptom scores compared to users not on hormonal contraception, but the difference was not significantly different (IRR=0·985, p=0·848). In comparison to those with long COVID, users with ME/CFS had significantly higher overall symptom scores (IRR=1·21, p=0·000408), and users with both long COVID and ME/CFS had even higher overall symptom scores (IRR=1·52, p=6·11E-11). Compared to reports submitted in the first three months since joining the app, symptom scores significantly increased over time, and by the fifth quarter IRR = 1·06, p =7·8E-49. This was associated with an increase in the number of symptoms tracked over time (appendix, p4). Age and gravidity did not show a significant association with symptom scores.
The most tracked symptoms across all users were fatigue (99·5%), brain fog (88·3%), headaches (85·1%) and muscle aches (78·6%) (appendix, p4). These symptoms were frequently co-reported with each other and with the following symptoms: dizziness, nausea, muscle weakness, joint pain, shortness of breath, palpitations, light sensitivity, noise sensitivity and memory issues (appendix, p5-6). Restricting the analysis to the first three cycles enabled us to conduct Pearson’s chi-squared tests as there were equal numbers of cycles for each user, and in this analysis, 25/36 individual symptoms differed significantly between menstrual cycle phases (appendix, p7-9). However, this approach may overlook long-term trends and variations in symptoms over time so we also built individual regression models for each symptom (appendix, p10-21). In this analysis 32/36 symptoms showed a significant association between symptom severity and menstrual cycle phase, with 22 of these being worse in the menstrual phase and better in the luteal phase (appendix, p10-21).
Approximately 85% of those affected by long COVID experience episodic symptoms that can rapidly fluctuate from periods of symptom stability to severe exacerbations, resulting in significant functional declines.3 These periods of severe worsening symptoms are often called “crashes” by those living with complex chronic illnesses,35 who may choose, where possible, to avoid scheduling strenuous activities at times when they are more vulnerable to a crash. The frequency of crashes was higher during menses, compared to all other phases, particularly the early luteal phase (p = 0·0002680) (appendix, p22). The mean percentage of days with crash was highest in the menstrual phase and decreased over the cycle, plateauing between the early luteal and premenstrual phases (Figure 2H). Adjusting for contraception type, disease type, age category, gravidity and time since joining the app using a mixed-effects logistic regression model, crashes remained less frequent in every phase compared to menstruation (Table 3). The late luteal phase showed the greatest reduction in crashes (odds ratio (OR)=0·888, p=5·87E-05) compared to the menstrual phase. The follicular, early luteal and premenstrual phase also showed significant reductions (OR=0·947, p=0·0367 and OR=0·900, p=0·000109 and OR=0·919, p=0·0333, respectively).
Crashes were significantly less frequent in users on combined hormonal contraception (OR=0·548, p=0·0409), compared to users not on hormonal contraception. Users on progestin-only contraception also had fewer crashes compared to users not on hormonal contraception but this was not significantly different (OR=0·864, p=0·446). In comparison to those with long COVID, users with ME/CFS had significantly more crashes (OR=1·47, p=0·00291) and users with both long COVID and ME/CFS had higher odds of crash (OR=1·85, p=4·4E-05). Age and gravidity were not significantly associated with crashes.
Discussion
This study is the first to describe changes to long COVID and ME/CFS symptoms across the menstrual cycle, using prospectively collected app data. In line with previous reports, overall symptom scores and crash incidence varied by disease, with the lowest scores reported in individuals with long COVID, higher scores in individuals with ME/CFS and the highest scores in individuals with both.7 The most tracked symptoms and clustering are broadly in line with Davis et al. 2021 who found that fatigue and brain fog were the most frequent symptoms in long COVID patients after 6 months of the study, alongside post-exertional malaise/crashes3.
When daily symptom ratings were summed to form an overall symptom score, there were small but significant changes between phases: symptoms were most severe in the menstrual and premenstrual phases and least severe in the early luteal phase. We saw a similar pattern with crash incidence, and these findings are in line with our hypothesis and with previous findings 3,17. When disease type, age, reproductive history and contraception type were adjusted for, the menstrual cycle phase still had a small, but significant impact on symptom score and crashes. Use of combined hormonal contraception was associated with significantly ameliorated symptoms and a reduced risk of crashes. This warrants further investigation as combined hormonal contraception is generally considered safe, cheap and is already widely available – making it a promising approach to ameliorate long COVID and ME/CFS symptoms.
The major strength of this study is its use of data prospectively collected from a large cohort of users of the Visible app. Users enter data in real time as part of their personal healthcare routines, reducing recall and recruitment bias. The availability of a large volume of data enabled us to examine changes over the menstrual cycle in granular detail, since sufficient user-days were assigned to even the shortest phase (premenstrual) for statistical analysis. The comprehensive list of 36 symptoms tracked allows a thorough investigation into the effect of menstrual cycle phase on long COVID and ME/CFS. The use of 4-point Likert scales for symptom severity ensured consistency in data reporting and analysis. This approach also allowed us to calculate an ‘overall symptom score’ as a single comprehensive indicator of a user’s health on any given day. We also examined individual symptoms and crashes to provide a more detailed picture of symptom severity.
The use of app data also carries some limitations. Our data is a convenience sample. Individuals with limited access to technology or digital literacy are unlikely to be app users, and individuals who choose to use the app and participate in the study might also have different characteristics or health behaviours compared to those who do not, potentially limiting the applicability of the results. Importantly, our study also relies on self-reported disease status, which may introduce inaccuracies into disease group assignments, and daily symptoms, which may introduce biases such as over-reporting or under-reporting of symptoms, and inaccuracies in tracking. Different users might also interpret and rate symptoms differently. However, to address this issue, we included ‘individual’ as a random effect in the model. The userbase of the app also means that there are no healthy controls, which prevents us from determining whether changes in long COVID and ME/CFS symptoms are due to the ovarian hormones affecting the disease, or whether the observed change in symptoms reflects normal variation in wellbeing across the menstrual cycle. However, our finding that combined hormonal contraception is associated with lower symptom scores, does support the hypothesis that ovarian hormones directly affect disease symptoms.
While our model adjusts for several variables, there may still be unmeasured confounders. Within-individual analysis means that the association between symptom severity and menstrual cycle phase is likely to be robust to this, but this could certainly affect the relationship between contraceptive use and symptoms. Furthermore, the demographic survey questions were completed when the user first enrolled and were not updated over the study. Therefore, any changes in contraception use during the study would not have been captured. To minimise the cognitive burden of participation on app users, we designed the survey to be as short as possible, but this meant we did not collect information on other geographic location, ethnicity, socioeconomic status, body-mass index, education or other health conditions including gynaecological disorders. Nor were we able capture daily contraception information, including emergency contraception use, or any other hormonal intake (e.g. testosterone or hormone replacement therapy (HRT)).
The results presented here support the idea that menstrual cycle phase influences long COVID and ME/CFS symptoms. Broadly, people living with long COVID or ME/CFS may notice that their symptoms are worse during the perimenstrual period and better in the early luteal phase. The knowledge that they are more likely to crash during menstruation could improve quality of life, by allowing them to schedule strenuous activities, which might cause a crash and cannot be carried out during a crash, for other phases of the menstrual cycle. Furthermore, the association between combined hormonal contraceptives, lower symptom scores and fewer crashes points towards a role for ovarian hormones in modulating symptoms and highlights the therapeutic potential of oestrogen-containing contraceptives in long COVID and ME/CFS.
Declarations
Funding
AG is funded by the UK Medical Research Council (MRC) [grant number MR/W00710X/1]. CM acknowledges funding from the MRC Centre for Global Infectious Disease Analysis (MR/X020258/1), the NIHR (NIHR200908); a philanthropic donation from Community Jameel supporting the work of the Jameel Institute; The Eric and Wendy Schmidt Fund (G-22-63345). JTH acknowledges funding from the Bill & Melinda Gates Foundation (INV-005289). VM acknowledges funding from Borne, Action Medical Research (GN2971), the UK Medical Research Council (MR/X006875/1) and Genesis Research Trust. The funders of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. For the purpose of open access, the author has applied a ‘Creative Commons Attribution’ (CC BY) licence to any Author Accepted Manuscript version arising from this submission.
Availability of data and materials
While all data used in this analysis were anonymised, the individual-level nature of the data used risks individuals being identified or being able to self-identify if it is released publicly. However, the app data can be requested by bona fide researchers for specified scientific purposes by emailing info{at}makevisible.com.
Code availability
https://github.com/AbiGoodship/effect_of_ovarian_hormones_on_long_COVID_and_ME_symptoms
Competing interests
During the course of the study HL was employed by, and RP was a paid contractor for Visible Health Inc., the company that owns and operates the Visible app. Some of the study participants had paid for access to certain Visible features, though not access to symptom tracking or menstrual cycle tracking functionality which is free for all Visible app users. This was a study across both the free and paid user base. All other authors declare no competing interests.
Authors’ Contributions
VM and CM conceptualised the study. RP and HL curated the data. JH, AG and CM developed the methodology. AG formally analysed and visualised the data. CM and VM supervised the study. AG wrote the original manuscript draft. All authors were responsible for the methodology and review and editing of the manuscript. All authors discussed, edited, and approved the final version of the manuscript. All authors had final responsibility for the decision to submit the manuscript for publication.
References
Subject Area
Reviews and Context
1
Comment
0
TRIP Peer Reviews
0
Community Reviews
0
Automated Services
0
Blogs/Media
Author Videos