Bad UV map hurts melanoma risk screening
February 27, 2012
A few years ago I tried my hand at writing a few NIH grant proposals. I struck out. I had a couple applications scored but none were funded and they didn’t really come very close. In any case, while I was researching my topic I came across the map below on the National Cancer Institute’s website:
Along with this map and question they provide an explanation:
Question 1:
Does the patient live in the Northern, Central, or Southern United States?
Explanation
Melanoma is more common in people who live in areas that get large amounts of UV radiation from the sun. For example, melanoma is more common in California than in Minnesota, where the sun is not as strong.
“Gee, don’t forget your sunscreen if you cross the border from Utah into Nevada.”
Are you kidding me? Is this really the state of the art at the National Cancer Institute? Melanoma is serious business. According to NCI it killed about 8,700 people last year. Couldn’t we take the time to make a better map?
Now, let’s take a look at a map of annual solar radiation from the National Renewable Energy Lab (NREL):
See any differences between the two maps? Clearly, a neat and tidy south to north UV gradient does not exist, at least not in the US. And UV exposure doesn’t seem to pay much attention to state borders. Why didn’t the peer review process catch these Geography 101 errors? Perhaps a state-level map isn’t really appropriate for this type of screening? Probably there should be UV risk indices by nearest city or CBSA. Would it really be so much trouble to ask patients to select their city or enter a ZIP code rather than just current state of residence? And, while noting general medical history, couldn’t they determine the approximate number of years spent in residence per location to estimate cumulative lifetime UV exposure?
In the age of Google Maps on every laptop and smart phone, everyone thinks that they’re a cartographer. “Oh, making maps isn’t so hard. I painted the northern, central and southern states different colors. It was easy.”
Not really. There’s more to it than that.
My hope is by the time my 3-year old daughter decides on a career there will be as many people studying well-constructed maps and looking for spatial patterns in geographic data as we currently have staring into microscopes and looking for patterns of disease in skin tissue samples. John Snow should have convinced us long ago that both scales of analysis are critical for tackling public health problems.
11 Comments
This was a really interesting read! Thanks! Although, sadly, no matter which map I look at I am still in the high risk zone….and the proof is in the pudding as I had malignant melanoma ten years ago. Luckily, my dermatologist can “map” out my body better than the National Cancer Institute can provide an accurate map!
[…] exemples récents, voici un article qui a inspiré ce billet où l’utilisation de la carte comme support visuel vient brouiller le propos […]
[…] 10. Terrible map by Duke Climate Researchers. I had to call out these researchers for failing to even think about basic geography concepts when producing their temperature change map. The map by the National Cancer Institute attempting to illustrate melanoma risk is probably even wor…. […]
Melanoma rates are not at all correlated to the amount of UV radiation received geographically. This is simply not true!
The UV NREL map above is true. Compare that to the melanoma rates by state map below and see if you see any correlation at all:
http://www2.epa.gov/sites/production/files/sunwise/images/melanoma_incidence_usmap.gif
New Mexico and Arizona have some of the strongest sun and are in the second lowest quintile for melanoma rate diagnoses. On the other hand, Washington, Vermont, NH and RI are some of the grayest states and are in the top quintile for melanoma rates. There seems to be no correlation at all between UV intensity and melanoma rates.
It’s a fallacy.
Hi Greg. I won’t even bother to argue with you. Your lack of understanding as it relates to selecting an appropriate spatial scale for data analysis is precisely the problem this article attempts to illuminate. Thanks for helping to make the point. Best, Justin
No, I do understand your point very well. You stated it very clearly in your article: “UV exposure doesn’t seem to pay much attention to state borders.” Your point is that by ignoring more complete local data— for example, that western Kansas has a higher UV index than eastern Kansas— the NCI’s recommendations can be both useless and dangerous. Using broad data like “Kansas is in the central region, so you don’t need to worry about sunscreen,” if applied to all people living in Kansas, would not be appropriate to those living in western Kansas, since they have a UV index more similar to southern California. It might be, therefore, dangerous to them since they might go outside without wearing sunscreen.
I think you missed my point, though.
The point I’m trying to make is that there is a bigger issue here: The very premise of the map the NCI is supposed to be addressing…. that there is a correlation between UV exposure and melanoma risk…. is weak. You didn’t include such data here, and just assumed that that premise is accurate. The data is much more subtle than that. For example, the correlation between INTERMITTENT UV exposure and melanoma, or sunburn and melanoma, is very good. However, consistent UV exposure actually LOWERS your risk of dying from melanoma if you are diagnosed with it. I can send you the relevant MEDLINE abstracts if you care. Finally, if you look at the data even more closely, there is no correlation at all for Blacks and Hispanics with regards to UV exposure.
The maps I provided, while extremely broad and therefore open to the same criticism as the NCI map is open to, do, I argue, have some usefulness. Comparing melanoma rates between New Mexico and Arizona, which have high UV rates everywhere geographically, with New Hampshire and Rhode Island, which have very low UV indices everywhere, removes such local state variations from the argument.
Would it be better to compare, say, county by county melanoma rates with local UV indices? Of course! But there is still some usefulness in the map I provided.
You didn’t present that data, so this is no critique on your post. I’m attempting to point out that there is a more relevant issue here.
The most important thing, though, with all of the hoopla about melanoma rates, is that UV exposure is correlated quite well with overall LOWER rates of ALL cancers. (Specifically, breast, prostate, ovarian and colon.) Here is a better map with ovarian cancer rates by economic region.
http://observationcloud.files.wordpress.com/2011/09/ovarian-cancer-rates.png
Here’s and article on breast cancer rates and latitude:
http://thehealthycow.blogspot.com/2011/03/side-note-on-sunlight-and-vitamin-d.html
https://lh3.googleusercontent.com/-4PGO6wWeh4E/TXKIKT9PFGI/AAAAAAAAAD0/A6HtJLYqBUU/s1600/breast+cancer+by+city.jpg
The Vitamin D Council explains the science very well here:
https://www.vitamindcouncil.org/blog/vitamin-d-uv-esposure-and-skin-cancer-in-a-nutshell/
Yes, 8700 people a year in the US die of melanoma. However, 14,000 die of ovarian cancer, and 40,000 of breast, and 27,000 of prostate. Many of these may have been prevented with more UV exposure. The benefit-risk ratio for UV exposure is very good. Even if UV exposure does increase rates of melanoma (the data of which is not very convincing to me,) it likely decreases the rates of many other forms of cancer.
That is all.
My recollection is the relationship between UV exposure and skin cancer, including melanoma, is fairly robust if you separate out rates by race. I think I did some analysis with County level data but it’s been 5+ years and I don’t recall details. I’ll see if I can dig it up. You may be right about a connection between low levels of VitD and ovarian/breast/prostate cancer – I have no idea – but I don’t think you have demonstrated sufficient basis to dismiss the connection between UV and melanoma as a fallacy. In any case, thanks for the information. Best wishes, Justin
Additionally, there’s this map which shows total cancer rates by state.
The states with the highest UV radiation seem to have the lowest overall cancer rates.
This makes sense since UV radiation actually LOWERS rates of breast, prostate and colon cancers!
http://www.cdc.gov/cancer/dcpc/images/2010_all_map_incidence.jpg
This might be a better map showing overall melanoma rates per capita by state:
http://www.cdc.gov/cancer/skin/images/2010_skin_map_incidence.jpg
Ummm, since when is “radiation” only about “Ultra Violet” waves? The NREL map does not state it is representative of “UV” radiation across the Country. It is referring to ALL solar radiation and was more than likely created for some other use than skin cancer aid. The light (solar) spectrum goes from ultra violet to visible light to infrared with various other wave lengths along the way.
I can go outdoors after 6 p.m. in Sacramento, CA for example, and during the summer months the UV index (according to the EPA) is on average <2. There is plenty of light though… stays light until about 9 p.m. actually… but I certainly won't be getting a tan at 6 p.m. because there is almost no UV light waves present. Most everyone learned in school that the highest UV rays are present between 10 a.m. and 4 p.m., and those are the hours to be especially mindful to not overexpose oneself, and especially not burn, which HAS been associated with higher incidence of Melanoma, regardless of which State a person lives.
I was diagnosed with Melanoma in July 2011. I follow the EPA's map and their "forecast" of daily UV index published on their website mostly so I can plan my week without having to be excessively exposed to high UV radiation — one day during the week noon to 1 p.m. may be a "7" while three days later the forecast may be a "9" which is more intense and can cause overexposure more quickly (and that's important to note… it's a forecast, just like the weather, subject to change).
In any case, I "get" that it's extremely difficult to make a map of an entire country that would be so specific as to show the exact UV index on any given day at any given time, so the EPA does break it down into zones. You should check out their map and see what you think. If nothing else, it provides me with something that makes me feel better having a type of information regarding sun exposure. Being diagnosed with Cancer slaps a person down. It causes anxiety, depression, and even confusion. By having a tool that at least seems logical as far as sun-intensity goes, I can go about my day with less anxiety and stress. I actually had a Dermatologist tell me once that anytime I go outside without the need for a flashlight would be dangerous for me. She was either on some sort of power trip, was an egomaniac playing with the lives of her patients, or she was just plain stupid. I reported her for her arrogant remarks and she no longer works for the hospital where she had tenure for over 26 years. I don't know how many lives she screwed with, but having information readily available online, maps to consult, and study results to ponder, certainly helped me.
Until there is an invention that I can hang on the patio wall outside that will tell me EXACTLY what the UV index is in my backyard, I'll take what I can get. Because I DO believe that OVEREXPOSURE during the hours where UV radiation is strongest isn't healthy and can do more damage in a shorter amount of time. I still go outside. I still garden… it's one of my favorite things to do. I developed a Melanoma because I used to be a competitive swimmer and was sunburned more times than I can count and it finally caught up with me at age 53. Melanoma can also run in families though, children who have never been sunburned have developed Melanoma, so either of your maps makes absolutely no difference to them whatsoever.
So can a MAP, regardless of how accurate it is with regard to Ultra Violet light exposure on any given day of the year at any location, make a positive impact on the number of Melanoma diagnoses each year, especially when sun-exposure isn't always the cause of the Melanoma?
My observation is that you're both right in some respects… and you're both wrong in others :o)
Hi Viki, thank you for sharing your perspective! You correctly point out the need for information that operates on various spatial scales (where have I lived since birth vs where will I go today) and temporal scales (how many years did I live in Sacramento vs how long can I stay outside this afternoon). My intent in posting the NREL solar radiation map was to illustrate NCI’s over-simplification when it comes to assessing risk based on residential history. Best wishes to you and thanks again for sharing your thoughts! Best, Justin