No, a rat study with marginal results does not prove that cell phones cause cancer, no matter what Mother Jones and Consumer Reports say

The zombie story that cell phones cause cancer has risen from the grave yet again.

There are certain myths that are frustratingly resistant to evidence, science, and reason. Some of these are basically medical conspiracy theories, where someone (industry and/or big pharma and/or physicians and/or the government) has slam-dunk evidence for harm but conspires to keep it from you, the people. For example, despite decades worth of negative studies, the belief that vaccines are harmful, causing conditions ranging from autism to sudden infant death syndrome, to all varieties of allergies and autoimmune diseases, refuses to die. Fortunately, this myth is one that, after more than a decade of hammering by scientists, skeptics, and public health advocates, has finally taken on enough of the patina of a fringe belief that most mainstream news sources no longer feel obligated to include the antivaccine side in stories about vaccines for “balance.” It is a zombie myth, one that, no matter how often it is “killed,” always seems to rise again. Unfortunately, the same cannot be said for the myth that cell phones cause cancer, as some very credulous reporting late last week demonstrated in the form of headlines like this:

• “Game-Changing” Study Links Cellphone Radiation to Cancer (Mother Jones)
• Cellphone-Cancer Link Found in Government Study (WSJ)
• Questions abound after study links tumors to cellphone radiation (Science)
• Major US study links cellphone exposure to cancer — at least in rats (STAT)
• “It actually has me concerned, and I’m an expert”: Major cell phone radiation study reignites cancer debate (Salon.com and Scientific American, the latter under a less inflammatory, but still overblown headline.)
• Government study finds link between cell phones and cancer in rats (Yahoo! Finance and Consumer Reports
• Massive government study concludes cell phone radiation causes brain cancer (NaturalNews)

Yes, I know that NaturalNews.com is not a mainstream news site. Rather it’s a quack site run by Mike Adams. Just search this blog or my not-so-super-secret other blog for numerous posts about the contortions and abuse of science and medicine by Mr. Adams. I included his article, quite simply, to illustrate that some headlines from mainstream news articles on the study don’t sound all that different from Mike Adams. Also notice how many of these headlines leave out an important fact, namely that this study was not done with humans, but with rats.

Let me show you what I mean. Here’s a quote from the Consumer Reports article:

The results of this large, long-term study could dramatically shift the national debate over cell phone safety. The NTP’s website says that the results may be used by the Food and Drug Administration and the FTC in determining how best to protect consumers from the potential harms of radiation that comes from cell phones.

The CDC might also consider reinstating the cautions it pulled from its web site. (We’ve reached out to the agency for comment, and will update our story once we hear back from them).

Likewise, the cell phone industry may have to alter its stance. The wireless association trade group CTIA has maintained that cell phones are completely safe, and has fought to block San Francisco from passing laws that would require electronics retailers to notify consumers about the proper handling of cell phones.

From Mother Jones:

It’s the moment we’ve all been dreading. Initial findings from a massive federal study, released on Thursday, suggest that radio-frequency (RF) radiation, the type emitted by cellphones, can cause cancer.

The findings from a $25 million study, conducted over two-and-a-half years by the National Toxicology Program (NTP), showed that male rats exposed to two types of RF radiation were significantly more likely than unexposed rats to develop a type of brain cancer called a glioma, and also had a higher chance of developing the rare, malignant form of tumor known as a schwannoma of the heart.

Now, NaturalNews:

After decades of denials and attacks by the media which called people concerned about cell phone radiation “tin foil hat-wearing conspiracy theorists,” a massive, multi-year study funded by the federal government now concludes that yes, cell phone radiation causes brain cancer.

The study is published here and it’s entitled, “Report of Partial Findings from the National Toxicology Program Carcinogenesis Studies of Cell Phone Radiofrequency Radiation in Hsd: Sprague Dawley SD rats (Whole Body Exposures).”

“The findings, which chronicle an unprecedented number of rodents subjected to a lifetime of electromagnetic radiation, present some of the strongest evidence to date that such exposure is associated with the formation of rare cancers in at least two cell types in the brains and hearts of rats,” reports Scientific American.

To be fair, NaturalNews includes Adams’ usual conspiracy-mongering about vaccines, GMOs, and the like, linking them all to “government coverups,” but when you are a mainstream publication like Consumer Reports or Mother Jones and your headlines and much of your text are not that far removed from something published on NaturalNews, you are doing it wrong. As Matthew Herper put it writing for Forbes about the reporting on this study, “Yesterday’s cell phone cancer scare scares me a little about the future of journalism.” In fact, if you look at some of the stories linked to above, you’ll note that many of them include notes at the end mentioning something like, “This article was updated to reflect criticism of the study’s conclusions by outside researchers.” That’s the press jumping first and being forced to backtrack under reasonable criticism. Unfortunately, none of them seem actually to make it very clear specifically how the stories were altered in response to criticism, which is bad.

Let’s take a look at some background, and then on to the study.

A little background: The science of the cell phones-cancer link

I can predict right now that someone will object to my starting this post by comparing the belief that cell phones cause cancer to the quack belief that vaccines cause autism, but I’m going to go one further. The idea that vaccines cause autism is actually more plausible than the idea that cell phones cause cancer—a lot more plausible. (Look for that sentence to be quote mined someday.) Moreover, I say this as someone who has criticized a physicist and famous skeptic for using a what I view as a simplistic “Cancer Biology 101” understanding of carcinogenesis to state that radio waves of the frequency used in cell phones cannot possibly cause cancer because of their low energy, which is very much insufficient to break chemical bonds in DNA, which is how many carcinogens cause cancer. I’ve even been criticized for perhaps being a little too open to the idea that radio waves can have significant biological effects that might even include causing cancer, and in one case I probably was.

Still, from the standpoint of basic science, specifically basic physics and biology, the likelihood that radio waves can cause cancer is incredibly unlikely, or, as I like to put it, not quite homeopathy-level implausible but damned implausible nonetheless. Indeed, from a biological standpoint, a strong link between cell phone use and brain cancer (or any other cancer) is not very plausible at all; in fact, it’s highly implausible. Cell phones do not emit ionizing radiation; they emit electromagnetic radiation in the microwave spectrum whose energy is far too low to cause the DNA damage that leads to mutations that lead to cancer. While it is possible that perhaps heating effects might contribute somehow to cancer, most cell phones, at least ones manufactured in the last decade or so, are low power radio transmitters. It is also necessary to acknowledge the possibility that there might be an as-yet-undiscovered biological mechanism by which low power radio waves can cause cancer, perhaps epigenetic or other, but the evidence there is very weak to nonexistent as well. Basically, based on what we know about carcinogenesis, a postulated link between cell phones and cancer is highly implausible.

In the absence of better basic science that nails down a heretofore-undiscovered potential biological mechanism by which exposure to radio waves could cause cancer, I have a hard time managing to muster any enthusiasm about recommending more studies than the ones that are already going on, particularly in light of various recent studies that we’ve examined that purport to find a link between cell phones and cancer but really do not, as described in these posts dating back to 2008, listed for your convenience if you want more in-depth information and discussion:

1. Cell Phones and Brain Tumors (Steve Novella)
2. Cell phones and cancer again, or: Oh, no! My cell phone’s going to give me cancer! (David Gorski)
3. Cell phones and cancer again, or: Oh, no! My cell phone’s going to give me cancer! (revisited) (David Gorski)
4. New Data on Cell Phones and Cancer (Steve Novella)
5. A Disconnect between cell phone fears and science (Lorne Trottier)
6. Critique of “Risk of Brain Tumors from Wireless Phone Use” (Lorne Trottier)
7. Are Cell Phones a Possible Carcinogen? An Update on the IARC Report (Lorne Trottier)
8. No, carrying your cell phone in your bra will not cause breast cancer, no matter what Dr. Oz says (David Gorski)
9. About that Cell Phone and Cancer Study (Steve Novella)

In other words, as a skeptic who’s probably the most open-minded (perhaps almost to the point of my brains falling out) to the claim that cell phones cause cancer, I still consider the claim, on basic science considerations alone, so incredibly implausible as to be an incredible, albeit not quite physically impossible, claim. I base this opinion on a preponderance of evidence that shows that brain cancer incidence is not increasing, inconsistent cell culture and animal studies that suffer from publication bias and when considered in the context of Bayesian prior plausibility are in fact negative, several epidemiological studies that failed to find a cell-phone cancer link, and the fact that the only epidemiological studies that claim to find a cell phone-cancer link have come from one group in Sweden whose principal investigator is known for being an expert witness in lawsuits against mobile phone companies.

Against that background, the background that journalists should have taken into account in writing about this study, let’s take a look at its results.

The National Toxicology Program study: Design

You will notice two points in nearly every story about this study, which was published at biorxiv.org. First is that the study was an expensive one. It cost $25 million. Second, it is frequently described with language like “one of the biggest and most comprehensive experiments into health effects from cellphones” and as “reigniting the debate” over cell phones and cancer. Personally, what I noticed about this study boiled down to two questions: How on earth can you spend $25 million on rat and mouse experiments? No, seriously. I do mouse and rat experiments myself in my lab, and, yes, they are expensive, but even with 90 rats per group, breeding rats, and repetitions I have a hard time figuring out how a study like this could cost $25 million. Hell, I could do a respectably large multicenter clinical trial for that much money. Indeed, I have a grant application in that proposes a reasonably sized pilot clinical trial for less than $2 million. I’m not saying that science isn’t expensive, but what’s going on here? My second question is this: Why was this release to the public instead of going through peer review first and being published in a decent journal? Basically, this is a report of a partial finding of a study not published in the peer-reviewed medical literature, with the report noting, “The findings in this report were reviewed by expert peer reviewers selected by the NTP and National Institutes of Health.” Something odd is going on here. As Matt Herper noted:

This paper was different. It was published on something called bioRxiv, a server for scientific papers run by Cold Spring Harbor Laboratory. It’s modeled on similar servers used by economists and physicists to share papers quickly, without the cumbersome process of going through peer review at medical journals. This is already par for the course for economists and physicists, but news in those fields tends to dribble out to the public. For medical studies, news can come in a giant, crushing wave.

After Microwave News released a story about the preliminary results of this study, according to Science the NTP investigators “decided to release the rat data before completing their analysis and writeup of the entire study, which isn’t scheduled to be finished until 2017, because of high public interest and the intriguing results.” See where that got them. This is one good reason why great care should be taken releasing preliminary results. What the NTP did borders on science by press release.

Now let’s look at the design of this study. One thing I found odd about it was that exposure to cell phone radiofrequency radiation (RFR) began in utero. Pregnant rats were exposed to 900 MHz GSM– or CDMA-modulated RFR beginning at on Gestation Day (GD) 5 and continuing through gestation. After birth, pups were exposed to the same RFR until weaning on postnatal day (PND) 21, at which point the mothers were removed and the exposure of 90 pups per sex per group was continued for up to 106 weeks. Pups remained group-housed from PND 21 until 24 they were individually housed on PND 35. All RFR exposures were “conducted over a period of approximately 18 hours using a continuous cycle of 10 minutes on (exposed) and 10 minutes off (not exposed), for a total daily exposure time of approximately 9 hours a day, 7 days/week.” Control rats were treated identically, except that they were not exposed to RFR.

Before doing this study with 90 rats per group, the NTP investigators did some pilot studies in order to establish that the RFR field strengths used didn’t raise the body temperatures of the rats, which could affect the results. They also did 28 day toxicology pilot studies before doing the massive experiment, which ended up with seven experimental groups:

1. Control (no RFR)
2. GSM 1.5 W/kg
3. GSM 3.0 W/kg
4. GSM 6.0 W/kg
5. CDMA 1.5 W/kg
6. CDMA 3.0 W/kg
7. CDMA 6.0 W/kg

At 90 rats per group, that’s 630 rats, but the rats were actually divided by sex, and 90 rats were used for each sex, which means that there were actually 1,260 rats. That’s a lot of rats.

The NTP study: Brain tumor results

The first thing that jumps out at me as I read the NTP report is this result:

At the end of the 2-year study, survival was lower in the control group of males than in all groups of male rats exposed to GSM-modulated RFR. Survival was also slightly lower in control females than in females exposed to 1.5 or 6 W/kg GSM-modulated RFR. In rats exposed to CDMA-modulated RFR, survival was higher in all groups of exposed males and in the 6 W/kg females compared to controls.

What I found particularly irritating about how this was reported was how difficult it was to find convenient, easy comparisons of the survival results for each group. Some Kaplan-Meier survival curves would have been really nice here, as we do in pretty much every medical paper in which survival is noted, including preclinical studies using rodents. If this paper were to have been submitted to a journal and I was asked to review it, I would insist on this because it takes way more effort to figure out what the authors found than it should. That being said, I can’t resist noting that some skeptics have been joking that in reality this study should have been reported with headlines like, “Cell phone radiation makes rats live longer.” True, saying that is just as misleading as many of the headlines about the study, but, then, that’s rather the point.

But what about cancer? What’s problematic about this study is that, even though there were 90 rats in each group, that’s actually a small number to detect meaningful differences in low frequency events. Here’s what I mean. The study reports increased incidences of tumors in the brains and hearts of the male rats. Specifically, there were noted to be increased incidences of malignant glioma and cardiac schwannomas in some of the groups exposed to GSM or CDMA RFR. Now here’s where the problem of small numbers comes in. The highest number of any single lesion in any single experimental group reported was 6, or 6.6% of the group. All numbers were in the single digits, and most were less than 3. Comparing such low frequency events between groups can be very problematic, particularly in the case low plausibility associations with multiple comparisons.

Let’s dive in a bit. In the male rats the number of gliomas detected for GSM RFR among 90 rats in each group was 3 (3.3%), 3 (3.3%), and 2 (2.2%) for 1.5 W/kg, 3 W/kg, and 6 W/kg, respectively, while for CDMA RFR the same numbers were 0 (0%), 0 (0%), and 3 (3.3%) for the same doses. Compare this to zero gliomas in the control group, and on the surface this looks alarming. There are a couple of problems. First, for GSM RFR, there doesn’t appear to be a dose-response, unless there is a threshold level that is, for example, below 1.5 W/kg. In fact, it’s a problem that the number of gliomas observed in the control group is zero, because according to historical controls in previous studies, that number should be around 2% (11/550 (2.0%), with a range from 0-8%), as is noted under Table 1. Examining the study that way, I have to be very concerned that there was something off about the control group, particularly given the lack of a clear dose-response effect and comparing to historical controls. Given that, color me very skeptical that this is a “real” result, particularly given that there was no statistically significant difference among the female rats, that there is no clear biological mechanism that would explain why GSM radiation would be more “carcinogenic” at the same exposure levels, and adding the very low prior plausibility that RFR causes cancer.

Bottom line: I believe that this result is almost certainly spurious and not indicative that cell phone RFR causes malignant gliomas in rats.

NTP Results: Schwannomas

In humans schwannomas are benign nerve sheath tumors (only 1% ever become malignant) composed of Schwann cells, the cell type that normally produces the insulating myelin sheath covering peripheral nerves. As noted in the introduction of the NTP study, there have been studies (that I consider unconvincing) linking cell phone RFR to acoustic neuromas, which are also known as vestibular schwannomas, thanks, presumably, to the proximity to the ear of the RFR source (the cell phone) when being used for a conversation. Why schwannomas would occur in nerves in the hearts of rats whose whole bodies were exposed to whole body cell phone RFR is puzzling and would be unexpected; that is, if these results are to be believed.

What the NTP investigators found in terms of the incidence of schwannomas was similar to what was found with respect to gliomas. There was no difference in incidence in females. In males, there was a statistically significant difference in schwannomas of the heart, but when the investigator looked for schwannomas elsewhere, they found:

In contrast to the significant increase in the incidence of schwannomas in the heart of exposed males, the incidence of schwannomas observed in other tissue sites of exposed males (GSM and CDMA modulations) was not significantly different than in controls (Table 5). Additionally, Schwann cell hyperplasia was not observed in any tissues other than the heart. The combined incidence of schwannomas from all sites was generally higher in GSM- and CDMA-modulated RFR exposed males, but not significantly different than in controls. The Schwann cell response to RFR appears to be specific to the heart of male rats.

In other words, the only significant result was an increase in schwannoma incidence in the hearts of male Sprague-Dawley rats exposed to cell phone RFR, while there was no difference between the control group and the exposed groups for total incidence of schwannoma at all sites. Another way of looking at this (at least from my point of view) is that, because the incidence of schwannoma at all sites did not differ between controls and the RFR-exposed groups of male rats, in the RFR-exposed group there was a redistribution of schwannoma to the heart. Considering that there is no known plausible biological mechanism to explain how RFR might somehow sensitize the heart to schwannoma formation only in males and given that high level of biological implausibility that RFR even causes schwannoma in the first place, again the most likely explanation is that this, too, is a spurious result.

The problems with this study

To its credit, the NTP included the peer reviews that it did ask for. One reviewer, a veterinarian named Diana Haines, commented mainly on the pathology and, although she agreed that the results should be considered “likely the result of exposures to GSM—and CDMA—modulated RFR,” she did have some caveats, particularly how well the scheme of exposing the rats beginning in utero represented actual human exposure (my assessment: not very well). Another reviewer, Maxwell Lee of the National Cancer Institute Laboratory of Cancer Biology and Genetics, analyzed the data and concluded that the association with glioma was significant for schwannomas but probably not for glioma. Of course, my retort is whether it’s biologically significant, something I sincerely doubt.

A third reviewer, Aleksandra M. Michalowski, also of the NCI Laboratory of Cancer Biology and Genetics echoed my complaint about the presentation of the overall survival of each group and actually did what the authors should have done and constructed a table of the median survival of each group, which showed that the median survival in the males receiving 6 W/kg CDMA RFR was 8% longer, which makes me want to change my sarcastic headline to “Cell phone radiation makes male rats live 8% longer on average.” In any case, she was concerned about the potential for bias and agreed that the glioma data were marginal, although she thought the schwannoma data were likely indicative of carcinogenesis. Multiple reviewers also noted that the survival of the control group of male rats was poorer than most historical controls (only 28% were still alive at two years, compared to a mean of 47%, range 24% to 72%), which might have skewed the results.

Perhaps the most comprehensive critique came from Michael S. Lauer of the Office of Extramural Research at the NIH, who noted potential problems with blinding, use of the intent-to-treat principle, and the like. He even did some simulations and power analyses and concluded that:

Based on these inputs, the recommendations in Table 13 of the FDA guidance document, and a sample size of 90 rats in each group, I find very low power (<5%, see Appendix 2). Even allowing for a risk ratio of 5.0 (a level that is clinically unlikely), the power for 2-sided alpha=0.005, k=3 and low lethality is only ~14% (see Appendix 2).

And:

The low power implies that there is a high risk of false positive findings², especially since the epidemiological literature questions the purported association between cell phone exposure and cancer.

He was lead to conclude that he was “unable to accept the authors’ conclusions,” adding:

I suspect that this experiment is substantially underpowered and that the few positive results found reflect false positive findings. The higher survival with RFR, along with the prior epidemiological literature, leaves me even more skeptical of the authors’ claims.

So do I. So does Aaron Carroll, a.k.a. The Incidental Economist, who notes:

Where to begin? I didn’t see any sample size calculation, nor any discussion of what they expected to see. One of the reviewers did a power calculation for them (page 37) and found that based on 90 rats per group, the power was about 14%. This means that false positives are very likely. The cancer difference was only seen in females, not males. The incidence of brain cancer in the exposed groups was well within the historical range. There’s no clear dose response. Why schwannomas? Schwannomas in other locations than the heart were not significantly different. These are rats. I don’t know how this compares to real world exposure. And one more thing – the survival of male rats in the control group was relatively low, and if these tumors developed later in life, this could be the whole reason for the difference.

Yep.

I become even more skeptical by taking a Bayesian approach to the analysis and considering the very low prior probability of a positive result based on what we know about biology coupled with the multiple outcomes examined. Taking these issues into account, I agree with Lauer that the results reported are almost certainly due to chance and are not indicative of a real biological effect. There are just so many red flags in the study that should have told journalists that there’s a lot less there than meets the eye. I could tell this, and I’m not even a statistician.

Lessons to be learned

I had considered not blogging this, given how much digital ink had been spilled on the study within 24 hours after its being announced, but sometimes there is an advantage to having to wait three days before writing about a science story. It provides time to see how the reporting develops. To their credit, at least Scientific American, STAT, and Mother Jones altered their articles after criticism and noted that they had done so, with the results being much less alarmist stories. The same can’t be said about Consumer Reports or The WSJ.

I realize that news of the preliminary results of this NTP study had been reported by Microwave News (and in a very alarmist manner, I might add), and that that was likely the impetus for the NTP to rush to publish a preliminary report in a non-peer-reviewed source. I can even understand why the NTP might have done it. After all, imagine the conspiracy mongering that would have gone on after this report if the NTP had said that it wasn’t going to publish the results until it had been accepted to a peer-reviewed journal. I can picture the headlines: “Government hides evidence of cell-phone cancer link.” Still, I wish the investigators had waited, rather than to publish an incomplete study that was not peer-reviewed. Given that this is only a partial result, I anticipate a lot more seemingly-positive results that will likely be no more convincing than this publication was.

Matthew Herper, in criticizing how the press handled this story, noted that it scared him, correctly noting this about how all the caveats and red flags about this study were lost in the shuffle:

Those caveats should have come with the study when it came out. Instead, readers were told why they should be scared before they found out the reasons they should calm down.

I fear that this is not the last set of results from the NTP study that will be released to the public in this manner.