New research says average cycle isn't 28 days (and water is wet, etc)

[mcg1102]

A great new paper of interest to the sub came out this week, and I wanted to draw attention to it and discuss it.

Original research paper here

A variety of popular press articles about the paper here

Title: Real-world menstrual cycle characteristics of more than 600,000 menstrual cycles

What did they do? This is a study from Natural Cycles and their academic collaborators. They analyzed data from 124,648 users and 612,613 ovulatory cycles on BBT, OPKs, and bleeding patterns.

What did they find? A lot of cool stuff! One of the most important headline findings is that the average cycle isn’t the “textbook” one:

The mean follicular phase length was 16.9 days (95% CI: 10–30) and mean luteal phase length was 12.4 days (95% CI: 7–17).

So the average user ovulates around CD17, and this is true even if you look at people with average cycle lengths from 25-30 days — those people have an average ovulation day of CD15.

They also found that both cycle length and menstrual bleeding length decreased with age. Older users ovulate earlier than younger ones, but their luteal phases are not shorter.

A critically important finding in their study is that the “classic” 14-day luteal phase isn’t even the average luteal phase — that the average LP is more like 12 days.

What are the strengths? Did you see the part where I said it was SIX HUNDRED THOUSAND CYCLES? That’s awesome. Natural Cycles has a lot of users who are temping to avoid pregnancy, so they are motivated to enter a temp every day and be consistent in their temping habits. Previous studies, on which virtually all of our information is based, have generally used something like 100-200 subjects.

What are the limitations? This is data from real people using the Natural Cycles app, so temp data was collected by users at home, with all the typical weirdness that you know can happen if you frequent Temping Tuesday or /r/TFABChartStalkers. They didn’t confirm ovulation with ultrasound imaging, which is the gold standard, but which obviously wouldn’t allow them to analyze such a huge number of cycles.

What’s another thing that warms devbio’s cold, dark heart? They have an entire supplemental information section devoted to further nerdery, including comparing their results with the oft-discussed Ecochard paper and others in the field. Overall, I feel pretty convinced by their dataset.

TL;DR: If a calendar-based app is the only way you’re timing a) sex and b) when to take a pregnancy test, you’re gonna have a bad time.

 

[exinanition]

@mcg1102 We truly do not deserve you DevBio

 

[missourian101]

@mcg1102 Super interesting!! Okay so I just skimmed the article and I am very surprised that they reported average cycle length (and cycle phase lengths) based on the mean rather than the median. Since there is some minimum time length for each cycle phase, and we all know that cycles can get extremely long, the distribution of lengths is likely highly skewed with a long positive tail, which would pull the mean up. So even if the median (“most typical” or “most commonly observed”) cycle length is 28 days (14/14), the mean might be higher. Most scientists are aware that in skewed distributions, the median is more representative of the population than the mean. Anyone catch something that I didn’t? Why are they using mean instead of median?

 

[mcg1102]

@missourian101 I don't know, and I do wish they had presented medians and histograms, totally agreed. It's clear that the mean cycle length is skewed rightward by people with longer cycles -- they have about twice as many people in the 31-35-day cycle range than the 15-20 and 21-24-day cycle range combined. The modal cycle range is 25-30 days, and among that group, the mean cycle length is 27.6 days.

In the supplemental information, it looks like the LP length is pretty normally-ish distributed, and the modes are 13 and 14 days. It's also clear that they do have more people in the left tail than previous datasets do.

 

[mcg1102]

@missourian101 ...incidentally, and sorry for the double-reply, I just realized another paper I've been kicking around recently (this one), which involved about 1000 subjects, also found a modal LP of 12 days (see Figure 3), and a modal day of ovulation of CD16 for 28-day-cycle-having subjects (see Table 1).

 

[emanuela]

@missourian101 It does say that cycles over 90 days were excluded and less than 1% were over 50 days, so I doubt the longer cycles skewed things all that much in actuality.

 

[childfreechristianaggie]

@missourian101 Because the mean is the average number.....that's exactly the one they should be using.

 

[sisi]

@mcg1102 This is awesome! And totally accords with my intuitions just based on hanging around here and chartstalking - it has just never seemed like CD 14 or earlier is common enough for it to be the average, and luteal phases longer than ~13 days also seem relatively rare. I love when crowd-sourced information gets subjected to a nice analysis like this!

 

[mcg1102]

@sisi Yeah, agreed -- it's very clear from being around here that shorter-than-14-day LPs are much more common than the conventional wisdom would suggest.

 

[prissanna]

@mcg1102 Look at me with an incredibly average cycle. Who knew!? Thanks for the info.

 

[cristal19]

@mcg1102 Thanks for sharing!! This is really interesting. I checked back at the TFAB BFP stats compiled and the mean O day was 18, median 16. The mean and median age from TFAB is 30, like this study. It's cool to see it align.

Since only BFPs are recorded, we don't have the typical luteal phase on TFAB. I did subtract O day from average cycle length and luteal phase length mean comes to 14 and median comes to 13 days.

One observation from using both NC and FF was that NC followed FAM temping rules a bit more strictly. It seemed possible that NC didn't to use OPK data in the same way that FF does, and they don't use any CF data. My O date prediction would usually be predicted as a day or two later on NC than on FF.

 

[slainsparten]

@cristal19 TFAB has cool research too!

 

[emanuela]

@mcg1102 Welp, saving this post!

This is really, really cool. And also not surprising in the least, honestly! It lines up so much better with what I've personally seen over the years here. Especially with LP length - I see 11-13 way more than I see 14+.

I wonder if FF has done any sort of similar analyses like they have for average day of first positive test etc?

 

[mcg1102]

@emanuela They really should. I respect the hell out of FF's data policy, but dude, if NC has hundreds of thousands of cycles, FF probably has tens of millions by now.

 

[emanuela]

@mcg1102 Like, even if it was just something they release on their site instead of publishing in journals etc., it would be really cool to see.

 

[katrina2017]

@emanuela I don't know if this is what you are referring to, but the "pregnancy monitor" tells you Pregnancy Test Probability but their data is significantly limited to those who post BFPs in their gallery as opposed to data from all documented charts.

 

[lida09]

@mcg1102 That's super interesting! Any idea why cycle length decreases with age?

 

[mcg1102]

@lida09 It's because you tend to ovulate earlier as you get older. As you get older, and your ovarian reserve (the number of eggs you have left in the bucket) gets smaller, your levels of AMH (which is produced by the reserve follicles) get lower. AMH inhibits FSH, so as AMH gets lower, early-cycle FSH gets higher. Higher levels of FSH early in the cycle accelerate follicle selection, which causes earlier ovulation.

 

[lida09]

@mcg1102 So should someone with short cycles have reason to be concerned about ovarian reserve? And are women with longer (but still regular) cycles more likely to be fertile?

 

[sisi]

@lida09 DevBio linked one of my old comments (under an old username) but just want to add that I was definitely concerned that my early ovulation (CD 11-13 usually, as early as CD 9 a couple of times) might be linked to low ovarian reserve. However, I had my AMH tested and it was 4.7 ng/ml at age 33, which is somewhere between the 85th-90th percentile, so there goes that explanation!

My guess would be that it's relative to individuals so that as you get older, your ovulation might get earlier, but that doesn't really mean the absolute value tells you anything. So for example Person A ovulated on CD 17 on average when younger and ovulates CD 13 due to ovarian aging, but Person B ovulated CD 13 when younger and ovulates CD 10 due to ovarian aging. So just knowing that you ovulate CD 13 doesn't tell you whether you are like Person A where that indicates ovarian aging or Person B where it indicates the opposite.