Can an algorithm replace “the pill”?

Natural Cycles is a leader in the innovative field of digital contraceptives and fertility tracking devices, but can its success be sustained after news of unexpected pregnancies among users?

Natural Cycles is an algorithm-based digital contraceptive tool, the first of its kind to receive FDA approval [1]. Numerous reports of unexpected pregnancies among users, however, have raised questions about the role of technologies like this one in driving important personal health decisions.

The app is built on an algorithm that leverages a user’s self-reported basil body temperature and period cycle days to predict when the user is ovulating, improving the reliability of its predictions over a learning term of few cycles [2]. The app pinpoints a user’s fertile days with a high degree of precision [3], signaling to the user with a color-coded screen whether having sex on a given day has the potential to result in pregnancy or not. Developed in Sweden by Co-Founder Dr. Elina Berglund, a former particle physicist at CERN, the algorithm is designed to account for uncertainties like sickness and alcohol consumption as it customizes the probability of conceiving for each unique user [4].

Clinical studies have shown that Natural Cycles has a “perfect use” failure rate of 1% and a “typical use” failure rate of 6.8% [3]. Compared to the most commonly used methods of contraception like hormonal birth control pills, IUDs, and condoms, Natural Cycle’s offers a similar failure rate without the discomfort of taking hormone-based contraceptives or relying on condoms [5]. Fundamentally, Natural Cycles has improved upon the age old “rhythm method” by replacing burdensome tracking with a digital app and a smart algorithm. As critics readily point out, it has not, however, been able to correct for human inaccuracies in gathering and reporting data [6].

After a hospital in Stockholm reported in early 2018 that 37 out of 668 requested abortions since September had been users reliant on Natural Cycles [5], public skepticism took over headlines about the company. Natural Cycles spokespeople responded: “no contraception is 100 percent effective, and unwanted pregnancies are an unfortunate risk with any contraception” [5]. A recently completed review by Swedish investigators confirmed this assertion [5].

At the request of Swedish regulators, Natural Cycles is responding to the crisis by clarifying the risks associated with using the product [7]. The UK’s Advertising Standards Authority has similarly warned that the company’s 99% “perfect use” accuracy claim is going too far [7]. Natural Cycles is thus changing its claims to focus on the 93% accuracy for “typical use”, recognizing the role human reporting plays in driving the bulk of the variation in the algorithm’s efficacy [7].

The Ava wearable device and companion app [9]
In the longer term, Natural Cycles plans to focus on comparative trials that assess the efficacy of the product in relation to traditional methods of birth control [4]. However, with several years needed to carry out these studies, the intermediate priority is product development and international expansion [4]. This starts with additional research into women’s health and markers of fertility, both of which could lead to an improved algorithm and potential adjacent products [4].

As Natural Cycles continues to grow, it should focus on two key opportunities:

  1. Short Term: Build closer relationships with physicians to build a network of advocates for the product
  2. Medium Term: Invest in a wearable that can automatically gather a user’s data, thereby limiting the risk of reporting error

Building greater buy-in with physicians and the medical community is essential to managing the downside risk of recent media coverage. Scathing comments from doctors about the role of human error in this product are persistent throughout coverage, despite the product having a “typical use” failure rate similar to that of traditional birth control. As the rollout of autonomous vehicles has shown, the public expects products built on machine learning to have a “typical use” failure rate closer to 0%, a much higher standard than human-led precedents. Building more support among doctors could help change public perception as well as create an organic future promotion channel for the app.

In the medium term, Natural Cycles can invest in a wearable that automatically gathers a user’s temperature as well as other relevant indicators like pulse rate. Companies like Zurich-based Ava are already pioneering wearables that do exactly this, successfully gathering a wider range of data to predict a user’s fertility [8]. This is essential to reducing the “typical use” failure rate to a level closer to the “perfect use” rate. Doing so could place Natural Cycles clearly ahead of hormonal birth control and condoms as a method of contraception. Can an algorithm replace the pill? I’d bet on it.

As we think about this product and future ones like it, should machine learning-based digital health products have a higher standard for efficacy in order to be approved by the FDA? Or should we treat this the same way we treat existing contraceptives?

(Word Count: 784)

______________________________________________________________________________________________________

[1] “FDA allows marketing of first direct-to-consumer app for contraceptive use to prevent pregnancy,” FDA press release (August 10, 2018).

[2] Natural Cycles, “How it Works,” https://www.naturalcycles.com/en-us/birthcontrol/howitworks/, accessed November 2018.

[3] E. Berglund Scherwitzl, O. Lundberg, H. Kopp Kallner, K. Gemzell Danielsson, J. Trussell, R. Scherwitzl, “Perfect-use and typical-use Pearl Index of a contraceptive mobile app,” Contraception, no. 90 (August 24, 2017): 420-425, via Google Scholar, accessed November 2018.

[4] Natasha Lomas, “Natural Cycles Gets $30M for its EU Certified ‘Digital Contraception’,” TechCrunch, November 9, 2017, https://techcrunch.com/2017/11/09/natural-cycles-gets-30m-for-its-eu-certified-digital-contraception/, accessed November 2018.

[5] Arielle Pardes, “In Contraceptive Tech, the App’s Guess is as Good as Yours,” Wired, January 19, 2018, https://www.wired.com/story/natural-cycles-contraceptive-apps/, accessed November 2018.

[6] Alexandra Sifferlin, “Can an App Prevent Pregnancy?” TIME, August 15, 2018, http://time.com/5365564/fertility-apps-contraception/, accessed November 2018.

[7] Natasha Lomas, “Natural Cycles Contraception App Told to Clarify Pregnancy Risks,” TechCrunch, September 17, 2018, https://techcrunch.com/2018/09/17/natural-cycles-contraception-app-told-to-clarify-pregnancy-risks/, accessed November 2018.

[8] Mohaned Shilaih, Valérie de Clerck, Lisa Falco, Florian Kübler & Brigitte Leeners, “Pulse Rate Measurement During Sleep Using Wearable Sensors, and its Correlation with the Menstrual Cycle Phases, A Prospective Observational Study,” Scientific Reports, no. 7 (May 2, 2017): 1294, via Nature.com, accessed November 2018.

[9] Daniel Winkler, “As More Women Choose to Have Babies Later, a Fertility Fitbit May Help,” NBC News, September 26, 2016, https://www.nbcnews.com/mach/technology/fertility-fitbit-more-women-choose-have-babies-later-life-n654411, accessed November 2018.

[Title Image Citation] Berenice Magistretti, “Natural Cycles is First Contraceptive App to get EU Approval,” VentureBeat, February 9, 2017, https://venturebeat.com/2017/02/09/natural-cycles-is-first-contraceptive-app-to-get-eu-approval/, accessed November 2018.

 

 

Previous:

Chariot: Changing Transit through Crowdsourced Co-Creation

Next:

Crowd-sourcing the Secret of Life: 23andMe and Open Innovation

11 thoughts on “Can an algorithm replace “the pill”?

  1. Sophia – Very interesting article on a product I hadn’t heard about before. I am glad to learn though that the company is focusing on the 93% rate instead of the 99% rate, so as not to oversell the consumer. I question the long-term position of this company, since a lower success rate could hurt the brand or a higher success rate would just entice more competition. I don’t know if this product has enough staying power to resist greater competition. As to your question of FDA success rates, there is already such a broad range in terms of the success rates of contraceptives in the market, I think that technology-based products should not have to reach a higher hurdle level. However, I do think there should be more education in the market about the actual levels of efficacy, those ranges, and what can impact that range higher or lower. I hope that the company provides clear and honest education for the consumer on these answers.

  2. Sophia – this is fantastically informative! I think you’re right on with your point that the public expects more our of machine learning than they do out of humans, so any small errors will result in horrible press, which is truly unfortunate. I think that what’s needed here is a culture shift. Right now people are very attached to the pill because it’s what they’re used to. But as we know, the pill is only around 93% effective (or less). Not to mention it has possible side effects that can be dangerous. People are often afraid of change, but with a societal shift – led by doctors and experts – can really change the industry and people’s lives. While I think it’s unfair, I do believe that learning-based digital health products will need a higher standard for efficacy to start. The reason I say this is because people need time to get used to this shift, and and negative press around it may scare people off. However, over time, I think these types of use of contraception will prove to be much more accurate, useful, and healthier. Thanks for an awesome article!

  3. Wonderful analysis of this technology. The public’s concern with the discrepancy of perfect use vs. typical use failure rate (6.8%) is a major issue in all medical treatments. Healthcare researchers attempt to discern the efficacy of by using “intention to treat” analysis when interpreting outcomes data in clinical trials. Put simply, the analysis considers the outcomes of everyone randomized to the treatment arm not just those who strictly adhered to the protocol. There is also a trend toward conducting pragmatic clinical trials where treatments are given in the natural environment (not perfectly controlled likely traditional studies) to better determine the real-world impact of an intervention.

  4. Growing up in Catholic schools, I learned about “natural” family planning as a means of birth control. It is really great to see a health/tech company catering towards the market that uses this method. I think a wearable is likely a hurdle to far for the time being as it just feels too… public. Birth control is an intensely private and personal decision and I doubt enough women would want to advertise their own preferences.

  5. Sophia, I love this. I hadn’t heard about this yet, and appreciate you discussing the issues. To you question about whether algorithms should need to meet higher standards than drugs to be accepted by the FDA, my hope would be no. In all cases of drugs, doctors make the best advice for the patient and I would hope this would be no different. Some woman are not able to take the pill, or some other form of contraception, and I would like them to have opportunities and benefits from alternatives too.

  6. Thanks for sharing – this is super interesting! I think that the conversation you started around whether technology should be held to higher standards, or even the standard of perfection, is interesting. In the case of Uber, I can understand it, as if you have the responsibility of driving out of someone’s control, I would expect that person to only accept perfection. However, in this case, I wonder whether the better standard is just if the technology can match the best alternative. Thanks again – very informative read!

  7. Sophia, it’s a very interesting topic to read, thanks for sharing! For me, I think there’s probably an “emotional skepticism” towards algorithm compared to pills, which we are used to for so long time and the mechanisms are pretty transparent. Taking this into consideration, I do think a higher standard on accuracy is needed if it seeks for an expanded use from the public.

  8. Fantastic and informative article! It feels like Natural Cycles is a great example of the classic struggle we have seen with predictive models in general and machine learning in particular in class of Garbage In Garbage Out- in this case the fact that inaccurate data input dramatically reduces the efficacy of the product through not fault of its own, but rather because its predictions are then less effective in assisting people. While I don’t believe that tech necessarily needs to be held to a higher standard, I would argue that in the case of something that is a medical treatment of sorts and is highly reliant on the patient engaging in accurate or predictable behavior (similar to the idea of a pill needing to be taken daily)- the effectiveness rates advertised should not be the ones achieved in Randomized Control Trials, but rather the bear case of an average user whose information entries may be off. To me this represents a more accurate comparison then with alternate choices such as IUD which take away the element of human behavior. I really like the idea of wearables as a good solution to this human input issue- we know from behavioral economics that the less opt in choices you need a human to make, the more like a certain function will be carried out successfully.

  9. This is so interesting; I had heard about applications that help users maximize their success with the rhythm method but did not know that new innovations were incorporating machine learning to drive higher accuracy. You make a really interesting point about the risk associated with this technology’s dependence on user-input data. This reminds me of the machine learning case we covered in class, in which we highlighted the criticality of feeding the right data into an algorithm to get the right outcomes or learnings in return. Given that the functionality of this product really depends so heavily on how the user feeds it data, how can the efficacy of the technology be evaluated? And given that the algorithms continue to learn and change over time, does that require constant re-evaluation on the part of some regulatory body? It is difficult to picture how efficacy can accurately be measured over time, which begs the question for me – is the company liable in any way for those individuals who experienced accidental pregnancies? In those cases, are we able to understand whether there was a user failure vs. a technology failure?

  10. I love the idea of this contraceptive. As someone who struggled with taking hormone-based contraceptives (mood swings, sudden outbursts) I think this is a really powerful product that gives women control over what they choose to put in their bodies. Although I understand the 7% failure rate is a problem, as long as the business is upfront about the risks I believe women should be able to choose. The FDA should treat these just as any other product and should not require higher levels of efficacy. It seems as though the wearable is a necessary next step, leading to more accurate source of data and better learnings. I believe this efficacy rate will come down as the product develops!

  11. Sophia – Very interesting article on a product I hadn’t heard about before !! I think that is a great push for changing people’s view on such a sensitive topic. My only question is how will customers shift into this new technology? People are so used to the pill today and shifting from it might take time. Thank you for raising this topic Sophia, very interesting.

Leave a comment