Photo by Mike Blyth

While social media and news outlets were reacting, or in some cases overreacting, to a new rodent-based medical study on the unlikely link between cell phone use and brain cancer last month, two studies and an accompanying commentary were quietly published in Pediatrics that raised similar concerns. Rather than cell phone use, the proposed potential cause of pediatric cancer in these newly published papers was phototherapy, a common treatment for newborn jaundice that I use regularly and have written about before. My previous post has a full review of jaundice in the newborn, how it can potentially cause permanent brain damage, and why phototherapy is a safe and effective treatment in most cases.

But is phototherapy truly safe? Can exposure to a narrow spectrum of blue light increase the risk of cancer in young children? And if so, what type or types of cancer? This is exactly what the study authors set out to investigate using the power of “Big Data.” Time will eventually tell us if the authors’ conclusions are justified or if they will end up only serving as excellent future examples of the Texas Sharpshooter Fallacy.

But I am perhaps getting ahead of myself.

Phototherapy and cancer: Is it probable?

A proposed link between phototherapy and cancer isn’t new. The phototherapy used in hospitals and at home to treat severe jaundice in babies consists of blue light between 430 and 490 nm in length. It isn’t ultraviolet, but it’s close. Ultraviolet light is not ionizing radiation, like x-rays, but it does have both positive and negative health effects. Most well-known is the link between ultraviolet light, in the form of excessive sun or tanning bed exposure, and skin cancer, but we also rely on it for vitamin D production. In fact, ultraviolet light is the only carcinogen with known health benefits.

So phototherapy is close to ultraviolet light on the electromagnetic spectrum. I would love to be able to brush this off, but there is in vitro research going back to the 1970s showing that DNA damage does occur in peripheral blood cells such as circulating lymphocytes, and in vivo data more recently has supported this. Again, sunlight has been shown to do the same thing, although there is considerably greater evidence in support of its impact on skin cells and the related increase in the risk of skin cancer. Any link between sun exposure and non-skin cancer is highly controversial at best.

The most recent paper on the subject of phototherapy-induced DNA damage, published in January of this year, again showed that phototherapy alone, rather than the elevated levels of bilirubin causing jaundice, damages DNA in blood cells. It appears that the longer the exposure to the light the greater the evidence of damage, regardless of the intensity of the phototherapy. This is a real effect, so unlike cell phones and brain cancer there is some plausibility to an association between phototherapy and cancer. But our much more advanced epidemiological understanding of the health effects of sun exposure serves as a glaring counterbalance to the concerns raised in these new studies.

Investigations into a link between phototherapy and cancer in children have so far been a mixed bag. Epidemiological studies, likely because of the data on DNA damage in peripheral blood cells such as lymphocytes and monocytes, have primarily focused on the risk of leukemia. I personally think it’s a stretch knowing what we do about sun exposure as explained above. Some of these studies have shown no association whatsoever. Others have shown possible increased risk of leukemia, but not always the same types. No studies have been able to conclusively answer the question.

The new studies

In the first paper, researchers set out to look for an association between newborn phototherapy and the diagnosis of any cancer during the study period, rather than just leukemia. It is a retrospective cohort study that looked at nearly half a million children born in Northern California hospitals from 1995 to 2011. They weeded out significantly premature infants and babies that died prior to discharge home from the hospital, and follow-up began after 60 days of life. Using electronic records, they gathered data on phototherapy use, cancer diagnoses, and numerous other variables such as gender, race and ethnicity, maternal age, birth weight, blood levels of bilirubin, and diagnosis of a genetic disorder or congenital anomaly.

Of the 499,621 study subjects, 39,403 of which had undergone phototherapy, 711 were ultimately diagnosed with cancer at some point after 60 days of life. 60 of these children with cancer had been exposed to phototherapy, which works out to 25 cases of cancer per 100,000 person-years compared to 18/100,000 in the unexposed group. Prior to adjusting for potential confounding variables, they found statistically significant but tiny increased risk for any cancer, any leukemia, myelocytic leukemia specifically, other leukemia, and liver cancer. After adjustment, the apparent increased risk disappeared.

The authors downplay the likelihood that chance or bias, rather than confounding alone, played any significant role in their initial findings of a crude association. I wasn’t impressed with their logic. They also go to great lengths to explain that they are not ruling out the possibility that there truly is a risk associated with phototherapy, despite the negative results, particularly for myelogenous leukemia, but they admit that it would be low.

After pointing out the low risk of negative outcomes in most babies with jaundice, and implying that treatment guidelines are too aggressive, they call for providers to avoid unnecessary phototherapy. Infants with Down syndrome, who have a much higher baseline risk of leukemia, are singled out as a potentially vulnerable population. Newborns with Down syndrome are also at increased risk for severe jaundice, so this presents a conundrum if one were to incorporate the authors’ conclusions into daily practice. Much more likely is that genetic abnormalities such as Down syndrome increase the risk of both jaundice and leukemia regardless of phototherapy exposure. As we say repeatedly, correlation does not necessarily equal causation.

In the second paper, a historical cohort study, the same authors looked at an order of magnitude more children, around 5 million born from 1998 to 2007, whose data was culled from the California Office of Statewide Health Planning and Development. Although it’s a lot more babies, use of this deidentified data set requires many more assumptions of accuracy than in the previous study and it only goes out to one year of age. They did not include the significantly premature or infants who died within 60 days of birth, and attempted to gather data on the same potential confounding variables as in the previous study.

Of the 5,144,849 included study subjects, 1,100 were diagnosed with cancer in the first year of life, which yields a population incidence of 21.4/100,000. Of the 178,017 subjects who received phototherapy, 58 were diagnosed with cancer, resulting in an incidence in the exposed of 32.6/100,000. That is an unadjusted relative risk (RR) of 1.6 with a CI that nudges up against 1. It’s statistically significant, but I question how clinically meaningful it is. There were similar small unadjusted risk differences found with myeloid leukemia, kidney cancer, and “other cancers.”

After adjusting for potentially confounding variables, these findings persisted but were overall a bit less significant. They came up with a number needed to harm, by causing an additional case of any cancer, of 10,638. For children with Down syndrome, only 0.2% of the overall cohort, they settled on a number needed to harm of 7,812. Keep in mind that although this study included many more subjects, it was significantly limited by a lack of access to the same potential confounding variables as in the previous study. This may explain why the concerning findings didn’t melt away after adjustment.

They also found that jaundice itself was an independent risk factor for cancer, albeit a small one with an RR of 1.3, but this raises some serious concerns in my opinion. Increased risk of cancer caused by jaundice alone is not something that has been reported. In fact, studies (including one I linked to above) looking into the mutagenic properties of phototherapy have shown that bilirubin doesn’t even cause DNA damage.

We know that bilirubin is neurotoxic, hence the brain damage with sustained high levels in the blood, but it would be very unlikely to cause cancer. In this study, they found similar risk of various cancers in infants who got phototherapy as in those who were simply jaundiced. It’s all likely just statistical noise as this association wasn’t found in the first study, which had access to actual serum levels rather than just a diagnosis code.

In this study, the authors again call for caution when recommending phototherapy. They again point out that most children with even severe jaundice don’t suffer permanent disability, and call for providers to avoid phototherapy unless indicated based on treatment guidelines. The potential risk of cancer, it is implied, should enter into our risk/benefit assessments.

Of note in both studies was the finding that the percentage of infants undergoing phototherapy increased significantly during the study periods. In particular in the first study, which included more recent data from 2011, the use of phototherapy increased from roughly 3% of subjects to nearly 16%. In the second study, use only increased from 3% to a little over 4%. That’s a big difference and I’m not sure why there was such a large discrepancy. I can’t find recent data on this question, but the lower figures are much more consistent with my own personal experience.

It is very likely that there has been an increase in the use of phototherapy considering that updated AAP guidelines in 2004 standardized our approach to universally screening all infants prior to discharge from the newborn nursery and tightened up recommendations for post-discharge follow-up. Universal screening of babies will find more cases of jaundice that is less clinically apparent and more likely to be near or just at the threshold for initiating phototherapy. But what we have not seen is an increase in cancer in children, which speaks against a causal link with phototherapy.

Some perspective

Accompanying the above two studies in Pediatrics last month was a commentary, “Can Big Data Shed Light on the Origins of Pediatric Cancer?“, the primary author of which is a pediatric oncologist at the Dana-Farber Cancer Institute here in Boston. It both supports the ultimate conclusions of the two studies, which really just boil down to “follow the treatment guidelines for newborn jaundice,” and attempts to put their findings into perspective. She begins, however, by lamenting the difficulty in searching for environmental causes of pediatric cancer and the difficulty in performing helpful epidemiological studies on such rare diagnoses.

She writes about how the authors had to develop a scoring system to help take into account all the potential confounding variables that might increase the risk of cancer and the need for phototherapy. She presents this as a necessary aspect of studying such uncommon outcomes. While I don’t question this, I do wonder how many researcher degrees of freedom went into its use. Could this have played a role in why the second study found an association with cancer even after adjustment was performed?

Frazier goes on to point out, rightfully so, how few actual cases of cancer were observed in these studies. It’s great that pediatric cancer is so rare, but it should give pause to anyone making strong claims based on three cases of liver cancer and ten cases of myelogenous leukemia in a phototherapy exposed group of over 175,000 babies. And considering the deidentified nature of the larger study data set, it is possible that some of these were inappropriately labeled as having undergone phototherapy. It wouldn’t take very many errors of that nature to completely change the statistical results.

Any conclusions of a causal relationship between phototherapy and cancer in children, according to Frazier, are a “judgement call.” She lists and then attempts to jibe these findings with Hill’s criteria for causation, first stating that the exposure in this case is plausible. I disagree. The plausibility is perhaps greater than homeopathy but it is extremely low in my opinion.

She claims that “most of the case-control studies have suggested an association between phototherapy and childhood cancer, principally acute myelogenous leukemia.” Again, I disagree. Some have, some haven’t. It’s mixed, which is what would be expected with preliminary studies looking at rare outcomes when no true relationship exists. The best of these smaller studies, which looked at 55,000 Danish children, showed no association.

She mentions an analogy with UV light and its proven ability to cause skin cancer, although the far-UV found in sunlight reaching us through the atmosphere isn’t ionizing radiation as she implies. But as I discussed earlier, this doesn’t support a link to cancers of the blood or solid organs and doesn’t meet Hill’s criteria in my opinion. Again, sun exposure causes skin cancer only, and risk is generally felt to be primarily associated with the number of episodes where severe sunburn has occurred. Babies are not burned by phototherapy in any way.

Frazier points out a dose-related increased risk of myelogenous leukemia seen in the first study I discussed. But this is based on only two cases in children who were exposed to higher doses of phototherapy. And again, after adjusting for potentially confounding variables, any association between cancer and phototherapy disappeared.

She then points out the fact that rates of cancer in children are not increasing, even the cancers that the study authors are most concerned about. In her final statement, Frazier recognizes “that the information is imperfect” and that front line providers face a dilemma. Phototherapy, she writes, might not be harmless and we should weigh the risks and benefits before placing orders.

Phototherapy vs cancer vs jaundice

Despite the fact that I think these two studies are largely nonsense, and I do not agree with any recommendations to discuss the risk of cancer with parents of children who require phototherapy as part of the informed consent process, I agree with their overall conclusions that we shouldn’t start phototherapy on kids who don’t meet treatment criteria. Phototherapy may not cause cancer in my opinion, but it is not risk free. No medical intervention ever is.

When an infant is placed under the lights, it means they are not in their mother or father’s arms. It usually means that they must stay in the hospital or be readmitted and hospitalization is not a benign intervention. It exposes patients to hospital bugs and medicalizing babies has the potential for establishing a false parental impression of vulnerability.

Phototherapy can thus potentially interfere with bonding. It can also interfere with early breastfeeding attempts and potentially lead to failure. This isn’t likely, particularly if the mother is supported in her efforts by a team that includes evidence-based lactation consultants. It is also helpful, once bilirubin levels are clearly responding to treatment, to relax the strict protocols about time allowed out from under the lights for feeding and snuggling.

Weighing risks and benefits of any intervention is paramount in the practice of medicine. In the context of newborn jaundice, we must consider the risk of untreated jaundice and the potential for permanent injury to a child’s brain. It is rare but it is considered a “never event” in pediatrics. Thus the gain on our approach to starting treatment is already cranked up pretty high and we don’t need to push it even further without good reason, such as a very rapid rate of rise in bilirubin in a child with a pathologic etiology of their jaundice.

How did the media handle the story?

At Science-Based Medicine, we frequently discuss the mishandling of nuanced medical findings by the media and institutional press releases that seem more interested in drumming up interest than in the accuracy of their reporting. I have to say that in this case, it wasn’t that bad – although I must add that this story was largely ignored, which I found a bit surprising considering it is the perfect set up for overreaction. Baby cancer! But cell phones are sexier I guess.

As expected, The Daily Mail botched their coverage of the studies, calling phototherapy a “leukemia and kidney cancer risk” right out of the gates. They right the ship towards the end, but the damage is already done. A United Press International article is better, but includes a terrible quote from epidemiologist and study author Thomas Newman that implies phototherapy may cause cancer in 1 out of every 1,000 treated babies. The two press releases on the publications are also fairly bland, although the one from UCSF focuses on children with Down syndrome and, I believe, isn’t clear enough that it is likely a case of correlation rather than causation.

Conclusion: Phototherapy is still safer than jaundice

Pediatric cancer is rare, so determining what potential environmental factors if any may play a role is exceedingly difficulty. I certainly understand the drive to make use of our increasing access to “Big Data” in order to attempt to tease them out. That being said, I don’t think that this approach was very helpful in building a case against newborn phototherapy.

There is no reason to differentiate any risk of negative health effects related to phototherapy from that of sun exposure, which is not associated with non-skin cancers. Plausibility of phototherapy causing leukemia is there but it is extremely low, and even lower for solid organ cancers such as in the kidney or liver. These studies do not justify changing the medical approach to science-based jaundice management. And I worry now about parents refusing phototherapy or avoiding recommended newborn follow-up care because of unfounded fears of cancer.