Temperature, Humidity, and Risk of Recurrent Gout Attacks
Temperature, Humidity, and Risk of Recurrent Gout Attacks
We conducted an internet-based case-crossover study of triggers of recurrent gout attacks. The details of this study have been described previously. Specifically, we constructed a study website on an independent secure server in the Boston University School of Medicine domain. The study was advertised on the Google (Mountain View, California) search engine (www.google.com) by linking an advertisement to the search term "gout." Individuals who clicked on the study advertisement were directed to the study website and were asked for the following items: sociodemographic information, gout-related data (e.g., diagnosis of initial gout attack, age of onset, medication used for the treatment of gout, and the number of gout attacks in the last 12 months), and history of other diseases and medication use.
To be eligible for the study, a subject had to 1) report gout diagnosed by a physician, 2) have had a gout attack within the past 12 months, 3) be at least 18 years of age, 4) reside in the United States, 5) provide informed consent, and 6) agree to release medical records pertaining to gout diagnosis and treatment. To confirm a diagnosis of gout, we obtained medical records pertaining to the participant's gout history and/or a checklist of the features listed in the American College of Rheumatology (ACR) preliminary classification criteria for gout completed by the subject's physician. Two rheumatologists (T.N. and D.J.H.) reviewed all medical records and the checklists to determine whether participants had a diagnosis of gout according to the ACR criteria. Similar methods of gout diagnosis confirmation have been used in the Health Professionals Follow-Up Study. The study was approved by the institutional review board of Boston University Medical Campus, and subjects provided electronic informed consent.
Data were collected regarding the date of onset of the recurrent gout attack, the anatomical location of the attack, clinical symptoms and signs (e.g., maximal pain within 24 hours or redness), and medications used to treat the attack (e.g., colchicine, nonsteroidal antiinflammatory drugs, systemic corticosteroids, intraarticular corticosteroid injections). This method of identifying gout attacks is consistent with approaches used in acute and chronic gout trials and the provisional definition of gout attacks that includes only patient-reported elements.
At the time of recurrent gout attacks, participants were asked to provide zip codes for their location in the 0- to 24-hour period and the 25- to 48-hour period prior to that attack using an online questionnaire. Using these zip codes, we retrieved information on temperature (in °F) and relative humidity (as %) during the 0–24 hours and 25–48 hours prior to the gout attack through the National Oceanic and Atmospheric Administration website (www.noaa.gov), a publicly accessible national weather data bank. We obtained the same weather-related information for control periods by matching the weekday within a 35-day interval using a time-stratified case-crossover study design. Specifically, we divided the calendar into blocks consisting of 35-day intervals starting on January 1, 2003. Within a particular 35-day interval (i.e., block), the day the gout attack occurred was used as the index day to anchor the corresponding control periods (i.e., anchored to the same day in the other weeks contained within the same block), enabling matching of each case period to 4 control periods. For example, if the gout attack occurred on the second Monday of a block, then the first, third, fourth, and fifth Mondays in that 35-day interval were used to define the index day for each of 4 control periods. We chose 2 days preceding the index date as the effect period (i.e., the exposure window) because we hypothesized that the gout attack would not occur immediately after exposure to the specific weather factors.
The time-stratified case-crossover study design can control for cyclical variation of the underlying hazard of gout attacks according to day of the week. For example, even though weather variation does not depend on the day of the week, individuals may have different behaviors on weekdays versus weekends that could have bearing on the risk of gout attacks, such as drinking alcohol. In addition, temperature and humidity can vary greatly over a long period of time. Thus, by restricting the control periods to the same day of the week within a short period of time (e.g., 35 days), potential control selection bias and time trend of exposure are minimized. This design has been successfully applied in several studies to assess the effect of weather-related factors on the risk of acute disease conditions.
We first examined the relation of mean temperature to the risk of recurrent gout attacks using conditional logistic regression. Specifically, we examined the associations of the mean temperatures over 0- to 24-hour, 25- to 48-hour, and 0- to 48-hour exposure windows. For each exposure window, we divided mean temperature into the following 7 categories: less than 30°F, 30–39°F, 40–49°F, 50–59°F, 60–69°F, 70–79°F, and 80°F or higher. (To convert degrees Fahrenheit to degrees Celsius, subtract 32, multiply by 5, and divide by 9.) We used the category that contained the median temperature (50–59°F) as the reference group because it is very rare that the temperature would change from the lowest category (<30°F) to the highest category (>80°F) within a consecutive 35-day period. Because the 0- to 48-hour exposure window appeared to generate more robust effect estimates for the risk of recurrent gout attacks in terms of providing maximal risk ratios, those results are presented here. To better depict the dose-response relationship between mean temperature and the risk of gout attacks, we used quadratic spline regression to smooth the dose-risk curve.
We then examined relative humidity, defined as the amount of atmospheric moisture (i.e., water vapor) present relative to the amount that would be present if the air were saturated at any given time and temperature. We divided mean relative humidity into the following 6 categories: less than 40% (very dry), 40%–49%, 50%–59%, 60%–74% (reference category containing the median relative humidity), 75%–84%, and 85% or more (very humid), and we used the same approach as described above to evaluate its association with the risk of recurrent gout attacks. Because a J-shaped dose-response relationship between humidity and the risk of recurrent gout attacks was noted in the spline curve, we added a quadratic term for humidity in the regression model.
Finally, we evaluated the combined association of temperature and relative humidity with the risk of recurrent gout attacks. To obtain robust effect estimates, we collapsed average temperature and humidity into 3 categories each (≤49°F, 50–69°F, and ≥70°F for temperature; <60%, 60%–74%, and ≥75% for humidity). In sensitivity analyses, we examined the association of temperature and relative humidity with the first attack during study follow-up, as well as with attacks affecting the lower extremities.
Methods
Study Subjects
We conducted an internet-based case-crossover study of triggers of recurrent gout attacks. The details of this study have been described previously. Specifically, we constructed a study website on an independent secure server in the Boston University School of Medicine domain. The study was advertised on the Google (Mountain View, California) search engine (www.google.com) by linking an advertisement to the search term "gout." Individuals who clicked on the study advertisement were directed to the study website and were asked for the following items: sociodemographic information, gout-related data (e.g., diagnosis of initial gout attack, age of onset, medication used for the treatment of gout, and the number of gout attacks in the last 12 months), and history of other diseases and medication use.
To be eligible for the study, a subject had to 1) report gout diagnosed by a physician, 2) have had a gout attack within the past 12 months, 3) be at least 18 years of age, 4) reside in the United States, 5) provide informed consent, and 6) agree to release medical records pertaining to gout diagnosis and treatment. To confirm a diagnosis of gout, we obtained medical records pertaining to the participant's gout history and/or a checklist of the features listed in the American College of Rheumatology (ACR) preliminary classification criteria for gout completed by the subject's physician. Two rheumatologists (T.N. and D.J.H.) reviewed all medical records and the checklists to determine whether participants had a diagnosis of gout according to the ACR criteria. Similar methods of gout diagnosis confirmation have been used in the Health Professionals Follow-Up Study. The study was approved by the institutional review board of Boston University Medical Campus, and subjects provided electronic informed consent.
Ascertainment of Recurrent Gout Attacks
Data were collected regarding the date of onset of the recurrent gout attack, the anatomical location of the attack, clinical symptoms and signs (e.g., maximal pain within 24 hours or redness), and medications used to treat the attack (e.g., colchicine, nonsteroidal antiinflammatory drugs, systemic corticosteroids, intraarticular corticosteroid injections). This method of identifying gout attacks is consistent with approaches used in acute and chronic gout trials and the provisional definition of gout attacks that includes only patient-reported elements.
Ascertainment of Weather-related Factors
At the time of recurrent gout attacks, participants were asked to provide zip codes for their location in the 0- to 24-hour period and the 25- to 48-hour period prior to that attack using an online questionnaire. Using these zip codes, we retrieved information on temperature (in °F) and relative humidity (as %) during the 0–24 hours and 25–48 hours prior to the gout attack through the National Oceanic and Atmospheric Administration website (www.noaa.gov), a publicly accessible national weather data bank. We obtained the same weather-related information for control periods by matching the weekday within a 35-day interval using a time-stratified case-crossover study design. Specifically, we divided the calendar into blocks consisting of 35-day intervals starting on January 1, 2003. Within a particular 35-day interval (i.e., block), the day the gout attack occurred was used as the index day to anchor the corresponding control periods (i.e., anchored to the same day in the other weeks contained within the same block), enabling matching of each case period to 4 control periods. For example, if the gout attack occurred on the second Monday of a block, then the first, third, fourth, and fifth Mondays in that 35-day interval were used to define the index day for each of 4 control periods. We chose 2 days preceding the index date as the effect period (i.e., the exposure window) because we hypothesized that the gout attack would not occur immediately after exposure to the specific weather factors.
The time-stratified case-crossover study design can control for cyclical variation of the underlying hazard of gout attacks according to day of the week. For example, even though weather variation does not depend on the day of the week, individuals may have different behaviors on weekdays versus weekends that could have bearing on the risk of gout attacks, such as drinking alcohol. In addition, temperature and humidity can vary greatly over a long period of time. Thus, by restricting the control periods to the same day of the week within a short period of time (e.g., 35 days), potential control selection bias and time trend of exposure are minimized. This design has been successfully applied in several studies to assess the effect of weather-related factors on the risk of acute disease conditions.
Statistical Analysis
We first examined the relation of mean temperature to the risk of recurrent gout attacks using conditional logistic regression. Specifically, we examined the associations of the mean temperatures over 0- to 24-hour, 25- to 48-hour, and 0- to 48-hour exposure windows. For each exposure window, we divided mean temperature into the following 7 categories: less than 30°F, 30–39°F, 40–49°F, 50–59°F, 60–69°F, 70–79°F, and 80°F or higher. (To convert degrees Fahrenheit to degrees Celsius, subtract 32, multiply by 5, and divide by 9.) We used the category that contained the median temperature (50–59°F) as the reference group because it is very rare that the temperature would change from the lowest category (<30°F) to the highest category (>80°F) within a consecutive 35-day period. Because the 0- to 48-hour exposure window appeared to generate more robust effect estimates for the risk of recurrent gout attacks in terms of providing maximal risk ratios, those results are presented here. To better depict the dose-response relationship between mean temperature and the risk of gout attacks, we used quadratic spline regression to smooth the dose-risk curve.
We then examined relative humidity, defined as the amount of atmospheric moisture (i.e., water vapor) present relative to the amount that would be present if the air were saturated at any given time and temperature. We divided mean relative humidity into the following 6 categories: less than 40% (very dry), 40%–49%, 50%–59%, 60%–74% (reference category containing the median relative humidity), 75%–84%, and 85% or more (very humid), and we used the same approach as described above to evaluate its association with the risk of recurrent gout attacks. Because a J-shaped dose-response relationship between humidity and the risk of recurrent gout attacks was noted in the spline curve, we added a quadratic term for humidity in the regression model.
Finally, we evaluated the combined association of temperature and relative humidity with the risk of recurrent gout attacks. To obtain robust effect estimates, we collapsed average temperature and humidity into 3 categories each (≤49°F, 50–69°F, and ≥70°F for temperature; <60%, 60%–74%, and ≥75% for humidity). In sensitivity analyses, we examined the association of temperature and relative humidity with the first attack during study follow-up, as well as with attacks affecting the lower extremities.
