Skip to main content

INSIGHTS BLOG > A Critical Review of a Study on Cannabis and Heart Failure

A Critical Review of a Study on Cannabis and Heart Failure

Written on 14 November 2026

Ruth Fisher, PhD. by Ruth Fisher, PhD

Slavin et al. (2026)[1] recently published a study with the objective “To determine the association between legal recreational cannabis sales (hereafter ‘legalisation’) and cardiovascular hospitalisations.”

The researchers obtained data on cardiovascular hospitalizations (CVH) by US census region. The researchers compared CVH for the Western Region (Alaska, Arizona, California, Colorado, Hawaii, Idaho, Montana, Nevada, New Mexico, Oregon, Utah, Washington, and Wyoming) to CVH in all other US census regions during the Jan 2012 through Dec 2018 period and concluded that “Legalisation was associated with increased hospitalisations for AF [atrial fibrillation], HF [heart failure] and MI [myocardial infarction].”

I believe two methodological flaws in particular, but also several others, render the results invalid.

1. The Regional Comparisons Fail to Establish the Objective

Slavin et al. reason that states in the Western region of the US were the first to legalize recreational cannabis activity, so a comparison of CVH in the Western region to CVH in all other regions during the early period of cannabis legalization would provide a “natural experiment” for testing the impact of cannabis activity on CVH. For their experiment to be valid, it must be the case that both of the following two Necessary Assumptions hold:

  • Cannabis was legal and widely available in the Western region during the study period, and
  • Cannabis was not legal or widely available in the other regions.

A. Recreational Cannabis

Figure 1 provides a map of the US with the Western region shaded in blue and with the month and year recreational sales began in states that legalized activity during the period of study, 2014 – 2018. Figure 2 provides a graph — reproduced from Slavin et al. — showing the portion of the total regional population residing in states that legalized recreational cannabis during the period of study, where 2 years prior to recreational legalization, 2012 – 2013, were included as a control.

Figure 1

1 us map early rec

Figure 2

2 pop state

Rollouts in states that have legalized cannabis activity have generally been slow, taking at least a year after retail sales have begun, but generally several years, for sufficient numbers of dispensaries to open up before legal cannabis becomes available to a good portion of the state population. Yet, Slavin et al.’s analysis assumes both that

  • The entire state population had reasonable access to legal cannabis at the time first sales began, and 
  • The entire population had access during the entire first year of sales, even if sales began during the middle or near the end of the year. 

Neither of these assumptions is valid. But even if we assume both of these to be the case, the information indicates that only a relatively small portion of people in the Western region — some 20% — had access to legal cannabis during all but the last year of study. In my view, this invalidates the first Necessary Assumption and thus the comparisons the researchers are making. It is presumably to overcome this problem that leads Slavin et al. to assume people in states bordering on legal states also have access to legal cannabis. Making this assumption increases access in the Western region from a mere 20% during the majority of the period to the vast majority of the population during most of the period, while leaving access in non-Western regions at minimal levels (Figure 3), as the researchers note:

By January 2018, 79.3% of the population of the Western region was living in a state with legal recreational sales and 98.6% of the population was either living in a state with recreational legalization or a border state. By contrast, there were no legal sales outside the Western region until legalisation in Massachusetts in late November 2018 with only a negligible volume of sales before the end of the year.

Figure 3

3 pop border state

While it is true that some people in border states do purchase from nearby dispensaries in legal states, it is not the case that the entire population of the border state will do so. In fact, most people will not travel too far – either within a state or to a neighboring state – to buy legal cannabis. A 2022 report by Bruce Barcott and Beau Whitney on this subject[2] noted 

A 2016 Access Development study found that 87% of consumers typically travel 15 minutes or less to make everyday purchases. The more frequent the purchases, the less time consumers are willing to travel... We estimate that cannabis consumers are willing to travel between 12 to 14 minutes from home to make a legal purchase.

Looking at the map in Figure 1 shows how comical Slavin et al.’s assumptions are. To get to their high coverage, they assume, for example, that

  • In 2014, when Washington had legalized cannabis but Idaho had not, that people from, say, southeastern Idaho would travel up to Washington to buy legal cannabis (~ 9 hours from Pocatello, ID to Pullman, WA);
  • In 2015, when Oregon had legalized cannabis but California had not, that people in, say, Los Angeles or San Diego would travel to Oregon to purchase legal cannabis (~ 10 hours from LA, CA to Medford, OR or ~ 12.5 hours from San Diego, CA to Medford, OR). 
  • In 2014, when Colorado had legalized cannabis but Arizona had not, that people in, say, Phoenix, Arizona would travel to Colorado to buy legal cannabis (~ 16 hours from Phoenix, AZ to Durango, CO).

While some people do travel long distances to buy legal cannabis, not enough do so to support the validity of Slavin et al.’s analysis.

B. Medical Cannabis

Slavin et al.’s objective was “To determine the association between legal recreational cannabis sales (hereafter ‘legalisation’) and cardiovascular hospitalisations.” Why did Slavin et al. focus only on recreational cannabis, and not also include medical cannabis? It’s not clear to me that the cardiovascular system can distinguish recreational from medical use and suffer an event only or predominantly if use is recreational.

Recognizing medical cannabis activity established prior to and during the period of study helps support Slavin et al.’s first Necessary Assumption, but it invalidates their second Necessary Assumption. Figure 4 presents a crude adjustment to Figure 1 (AI couldn’t quite get everything right) to include dates that both medical and recreational cannabis sales began in each state. As seen in the figure, medical activity was abundant in the Midwest and in the Northeast during the study period. As such, it’s difficult to blame cannabis activity on increases in CVH in the Western region relative to CVH in other regions during the period. That is, the inclusion of medical cannabis use in the analysis likely invalidates the study’s findings.

Figure 4

Signature: 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

2. The Study Fails to Identify the Majority of Cannabis Users in the Study Population

Slavin et al. used diagnoses codes (ICD codes) to identify cannabis users and distinguish them from non-users in their hospital discharge data:

“We obtained data [from 50,042,564 hospital discharges] on hospital discharge diagnoses and patient demographics from the NIS from 2012 to 2018.”

“Hospital discharge diagnoses were identified using ICD-­ 9 (International Classification of Diseases, 9th revision) codes before 1 October 2015, and ICD-­ 10 (10th revision) codes on or after 1 October 2015,” where “Hospitalisations with cannabis-­ related diagnoses served as our positive control to ensure that we could detect the effects of legalisation within our sample.

A. Diagnosis Codes Fail to Capture the Majority of Cannabis Users

ICD codes, by definition, identify diseases, not functional activity. ICD codes exist for cannabis abuse, dependence, and other dysfunctional use, but there are no codes for functional cannabis use. More specifically, Slavin et al. list 4 cannabis codes used in their analysis to identify cannabis users, 304.3, 305.2, F12, and T40.7. Here are the definitions for these codes:

  • 2014 ICD-9-CM Diagnosis Code 304.3: Cannabis dependence[3]
  • 2015 ICD-9-CM Diagnosis Code 305.2: Nondependent cannabis abuse[4]
  • F12: Cannabis related disorders[5]
  • 2027 ICD-10-CM Diagnosis Code T40.7: Poisoning by, adverse effect of and underdosing of cannabis (derivatives)[6]

Only a small portion of all cannabis users are diagnosed with dysfunctional use, so using these codes as a means to identify all cannabis users will fail to achieve that objective. Curiously, Slavin et al. provided no information in their study as to the number of cannabis users they identified in their hospital discharges data using ICD codes. 

As a point of reference, Benedict et al. (2016) used as a similar methodology as that used by Slavin et al. to study the association between cannabis use and the prevalence of certain fungal infections in health insurance claims data.[7] Using ICD coded to identify cannabis users in their analysis of over 21 million claims, Benedict et al. identified 0.25% of patients in the claims database as being cannabis users (Figure 5). 

Figure 5

5 fungus study

In their discussion, Benedict et al. noted that “cannabis use is likely greatly underrepresented by ICD codes, supported by the finding that <0.3% of our study population had cannabis use codes, whereas ≈9% of the US population reported using cannabis in the past month.” Using Benedict et al.’s estimate of a 9% prevalence of cannabis use suggests they captured 0.3% / 9% = 3% of the likely cannabis users in their claims database. Notably, Benedict et al.'s recognition that they likely missed the vast majority of cannabis users in their study population did not stop them from using the outcomes from their study to claim that cannabis use is associated with fungal infections, but I believe that under-capturing the vast majority of cannabis users invalidates their findings. If Slavin et al. similarly failed to capture the vast majority of cannabis users in their database – and there’s no reason to think otherwise – then I believe their findings would likewise be invalid.

B. Diagnosis Codes Capture Unrepresentative Cannabis Users

Noteworthy is the fact that individuals diagnosed with a cannabis use disorder (CUD) do not represent other cannabis users. Specifically, individuals diagnosed with CUD tend to also use many other legal, illicit, and prescription substances, including, for example, alcohol, cocaine, amphetamines, and opioids, and they tend to have multiple other comorbidities, notably mental health problems.[8] Even if Slavin et al,’s study were valid for the cannabis users they identified in their study – which the previous sections suggest is not the case – their study would only establish an association between cannabis use disorder and CVH, but not an association between functional cannabis use and CVH.

In short, I believe Slavin et al.’s reported study outcomes are not valid, since

  • The Necessary Assumption that legal recreational cannabis was reasonably accessible in the Western region during the study period but not in non-Western regions does not hold;
  • The study fails to account for legal medical cannabis, which was reasonably available both in the Western region as well as in other non-Western regions during the study period;
  • The study likely fails to capture the majority of cannabis users in the study population; and
  • The cannabis users identified in the study population do not represent the vast majority of cannabis users.

3. Alternative Explanations

Slavin et al. compared CVH in Western vs. non-Western regions during a specific period in time and attributed any differences to the legalization of recreational cannabis use. This assumes that the only change in the environment during the study period was the legalization of recreational cannabis activity in the Western region. That is, it assumes there were no other public policy changes, no changes in the economy, no natural or man-made disasters, no social or behavioral changes, etc. that could have caused their differential CVH. A quick search of the literature quickly revealed two potential causes of Slavin et al.’s findings: wildfires, and meth use.

A. Wildfires

There was unusually high wildfire activity in the Western region during the period of study, especially in 2015, 2017, and 2018. A study of wildfires in the Western US during 2017-2018 found extreme levels of particle pollution:

In the summers of 2017–2018, record‐breaking wildfires, combined with stable weather conditions, resulted in daily particle concentration of 35 to 175 μg/m 3 across western U.S. sites. These particle pollution extremes are twice as severe as long‐term average conditions.[9]

A study of the health effects of 2015 California wildfires further indicates an association between smoke exposure and the same types of cardiovascular events Slavin et al. found:

Researchers reviewed more than one million ER visits in northern and central California during intense wildfires in the summer of 2015… They found that smoke exposure was associated with increased rates of ER visits, not just for breathing trouble, but also ischemic heart disease, irregular heart rhythm, heart failure, pulmonary embolism and stroke.[10]

Wildfire activity could thus have contributed to Slavin et al.’s findings.

B. Increases in Use of Methamphetamines

Heart failure hospitalizations due to meth use (MethHF) increased dramatically during the 2008 – 2018 period:

Methods: Adult patients (≥18 years old) with HF as primary hospitalization diagnosis between 2008 and 2018 were included in this study.

Results: Age-adjusted MethHF hospitalizations per 100 000 increased by 585% from 4.1 in 2008 to 28.1 in 2018, while non-MethHF hospitalizations decreased by 6.0% from 342.3 in 2008 to 321.6 in 2018. The rate of MethHF hospitalization increase more than doubled that of a negative control group with urinary tract infection and meth-related secondary diagnoses...

Conclusions: MethHF hospitalizations increased sharply during the study period and contributed significantly to the HF hospitalization burden in California.[11]

The MethHF cases were disproportionately concentrated in the Western region:

After stratifying the 2014 HF hospitalization data by national census division, there was a heavy concentration of MethHF hospitalizations along the Pacific coast… (CA, HI, OR, WA…), with the prevalence of MethHF hospitalizations nearly 500 times that of the Middle Atlantic region (NJ, NY, PA…). The West Mountain region also had a high prevalence of methamphetamine abuse among HF hospitalizations (MT, WY, CO, UT, NV, AZ, NM…) [12]

Meth activity could thus also have contributed to Slavin et al.’s findings.

While wildfires and meth activity provide plausible alternatives to legalization of recreational cannabis activity, there are most likely other changes in the environment during the period that also could have contributed to Slavin et al.’s findings. But I’ve made my point. 

Final Note

As a final note, it saddens me to see such flawed research being published in noted peer-reviewed journals, by researchers at some of the nation’s top universities, and funded by NIH grants. No doubt, this study will be widely cited as evidence of the dangers of cannabis use, further perpetuating the longstanding stigma against cannabis. Perhaps one day the stigma will finally be overcome, with society recognizing the true potential of cannabis. Unfortunately, today is not that day.

References