Changes

The study failed to reach statistical significance which should be unsurprising given its limitations. The results were positive, however, with HCQ having a P value of .21 for lower rate of ongoing symptoms and a P value of .29 for a lower rate of hospitalization.<ref>[https://pubmed.ncbi.nlm.nih.gov/32673060/ Hydroxychloroquine in Nonhospitalized Adults With Early COVID-19: A Randomized Trial] by Caleb P Skipper et. al. (2020 Jul 16) ''Annals of Internal Medicine''</ref>

The study failed to reach statistical significance which should be unsurprising given its limitations. The results were positive, however, with HCQ having a P value of .21 for lower rate of ongoing symptoms and a P value of .29 for a lower rate of hospitalization.<ref>[https://pubmed.ncbi.nlm.nih.gov/32673060/ Hydroxychloroquine in Nonhospitalized Adults With Early COVID-19: A Randomized Trial] by Caleb P Skipper et. al. (2020 Jul 16) ''Annals of Internal Medicine''</ref>

===Retrospective studies from New York==

===Retrospective studies from New York===

There is no question that retrospective studies are not the best science.  One has to use statistics to try and account for all kinds of variables such as delayed onset of treatment, not giving a patient a drug due to a patient's risk factors, and other variables such as differences in comorbidities.  Some of these factors can be controlled for in certain situations.  For example, if patients are seen by the same group of doctors, then differences such as when a particular drug is administered may not vary widely, but even in that situation other variables will persist such as screening for risk factors.  The main advantage of retrospective studies is that ethical concerns are limited to protecting patient privacy as no patients are being given placebos or potentially harmful drugs.   

There is no question that retrospective studies are not the best science.  One has to use statistics to try and account for all kinds of variables such as delayed onset of treatment, not giving a patient a drug due to a patient's risk factors, and other variables such as differences in comorbidities.  Some of these factors can be controlled for in certain situations.  For example, if patients are seen by the same group of doctors, then differences such as when a particular drug is administered may not vary widely, but even in that situation other variables will persist such as screening for risk factors.  The main advantage of retrospective studies is that ethical concerns are limited to protecting patient privacy as no patients are being given placebos or potentially harmful drugs.   

In the case of New York, three large, retrospective studies for hosptilized patients were conducted, two of which said HCQ treatment had no effect, and a third which said HCQ had a net benefit.

In the case of New York, three large, retrospective studies for hosptilized patients were conducted, two of which said HCQ treatment had no effect, and a third which said HCQ had a net benefit.

Joshua Geleris studied the patients who came to the New York–Presbyterian Hospital (NYP)–Columbia University Irving Medical Center (CUIMC).  Having the results come from one hospital could mean that some of the variables one might encounter with a multi-hospital analysis are controlled for, that doesn't hold true in this case.  While the hospital initially recommended dosing for HCQ at 400 mg, these recommendations were subsequently removed and treatment was left to the doctor's discretion.  Moreover, it was only until after patients reached the emergency room that they were given HCQ (see figure S3), and even then only 45.8% received it within 24 hours.  That is pretty late to be giving HCQ.  In any case, what did they find?

Joshua Geleris studied the patients who came to the New York–Presbyterian Hospital (NYP)–Columbia University Irving Medical Center (CUIMC).  Having the results come from one hospital could mean that some of the variables one might encounter with a multi-hospital analysis are controlled for, that doesn't hold true in this case.  While the hospital initially recommended dosing for HCQ at 400 mg, these recommendations were subsequently removed and treatment was left to the doctor's discretion.  Moreover, it was only until after patients reached the emergency room that they were given HCQ (see figure S3), and even then only 45.8% received it within 24 hours.  That is pretty late to be giving HCQ.  In any case, what did they find?

{{Hydroxychloroquine-treated patients were more severely ill at baseline than those who did not receive hydroxychloroquine (median ratio of partial pressure of arterial oxygen to the fraction of inspired oxygen, 223 vs. 360). Overall, 346 patients (25.1%) had a primary end-point event (180 patients were intubated, of whom 66 subsequently died, and 166 died without intubation). In the main analysis, there was no significant association between hydroxychloroquine use and intubation or death (hazard ratio, 1.04, 95% confidence interval, 0.82 to 1.32). Results were similar in multiple sensitivity analyses.<ref>[https://www.nejm.org/doi/10.1056/NEJMoa2012410?url_ver=Z39.88-2003&rfr_id=ori:rid:crossref.org&rfr_dat=cr_pub%20%200pubmed Observational Study of Hydroxychloroquine in Hospitalized Patients with Covid-19] by Joshua Geleris et.al (June 18, 2020) ''The New England Journal of Medicine''</ref>}} In short, the researchers found that if you give HCQ in an ER preferentially to patients who were more severely ill at baseline, there is no significant difference between those given HCQ and those who weren't given it.  It is difficult to figure out how to interpret that result. Is there no effect in an ER setting as evidenced by the similarity of outcomes between the two groups?  Does it mean that there was a net positive effect to HCQ because the sicker patients weren't dying as often as one might otherwise expect?

{{quotebox|Hydroxychloroquine-treated patients were more severely ill at baseline than those who did not receive hydroxychloroquine (median ratio of partial pressure of arterial oxygen to the fraction of inspired oxygen, 223 vs. 360). Overall, 346 patients (25.1%) had a primary end-point event (180 patients were intubated, of whom 66 subsequently died, and 166 died without intubation). In the main analysis, there was no significant association between hydroxychloroquine use and intubation or death (hazard ratio, 1.04, 95% confidence interval, 0.82 to 1.32). Results were similar in multiple sensitivity analyses.<ref>[https://www.nejm.org/doi/10.1056/NEJMoa2012410?url_ver=Z39.88-2003&rfr_id=ori:rid:crossref.org&rfr_dat=cr_pub%20%200pubmed Observational Study of Hydroxychloroquine in Hospitalized Patients with Covid-19] by Joshua Geleris et.al (June 18, 2020) ''The New England Journal of Medicine''</ref>}} In short, the researchers found that if you give HCQ in an ER preferentially to patients who were more severely ill at baseline, there is no significant difference between those given HCQ and those who weren't given it.  It is difficult to figure out how to interpret that result. Is there no effect in an ER setting as evidenced by the similarity of outcomes between the two groups?  Does it mean that there was a net positive effect to HCQ because the sicker patients weren't dying as often as one might otherwise expect?

The short answer?  Nobody really knows because there are too many variables not being accounted for.

The short answer?  Nobody really knows because there are too many variables not being accounted for.

The study divided patients into ambulatory and hospitalized sub-groups.  Ambulatory was defined as "patients whose encounter was an office visit, emergency department (ED) visit, or telehealth/telemedicine."  Hospitalized was defined as "inpatients and ambulatory patients who were subsequently admitted to the hospital."  This subdivision is important as it helps control for two of the most important factors, namely the timing of HCQ use and the overall health of the patients using the drug.  Because of its sheer size, the study was able to do this subdivision without impacting its ability to achieve statistically significant results... even the subdivisions were larger than the studies conducted by Geleris and Rosenberg.  The subgroups were then further subdivided into survivors and non-survivors, with the end date being April 17, 2020 (meaning a patient had to have died or discharged by that date to be included in the study). Then they did a whole bunch of statistical stuff (Cox Regression analysis, inverse probability weighting (IPTW) based on propensity scoring) to determine what the statistical results were.<ref name=MountSanai>[https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7325642/#MOESM1 Risk Factors for Mortality in Patients with COVID-19 in New York City] by Takahisa Mikami (2020 Jun 30) ''Journal of General Internal Medecine''</ref>  So what did they find?         

The study divided patients into ambulatory and hospitalized sub-groups.  Ambulatory was defined as "patients whose encounter was an office visit, emergency department (ED) visit, or telehealth/telemedicine."  Hospitalized was defined as "inpatients and ambulatory patients who were subsequently admitted to the hospital."  This subdivision is important as it helps control for two of the most important factors, namely the timing of HCQ use and the overall health of the patients using the drug.  Because of its sheer size, the study was able to do this subdivision without impacting its ability to achieve statistically significant results... even the subdivisions were larger than the studies conducted by Geleris and Rosenberg.  The subgroups were then further subdivided into survivors and non-survivors, with the end date being April 17, 2020 (meaning a patient had to have died or discharged by that date to be included in the study). Then they did a whole bunch of statistical stuff (Cox Regression analysis, inverse probability weighting (IPTW) based on propensity scoring) to determine what the statistical results were.<ref name=MountSanai>[https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7325642/#MOESM1 Risk Factors for Mortality in Patients with COVID-19 in New York City] by Takahisa Mikami (2020 Jun 30) ''Journal of General Internal Medecine''</ref>  So what did they find?         

{{A total of 858 of 6493 (13.2%) patients in our total cohort died: 52/2785 (1.9%) ambulatory patients and 806/3708 (21.7%) hospitalized patients. Cox proportional hazard regression modeling showed an increased risk of in-hospital mortality associated with age older than 50 years (hazard ratio [HR] 2.34, CI 1.47–3.71), systolic blood pressure less than 90 mmHg (HR 1.38, CI 1.06–1.80), a respiratory rate greater than 24 per min (HR 1.43, CI 1.13–1.83), peripheral oxygen saturation less than 92% (HR 2.12, CI 1.56–2.88), estimated glomerular filtration rate less than 60 mL/min/1.73m2 (HR 1.80, CI 1.60–2.02), IL-6 greater than 100 pg/mL (HR 1.50, CI 1.12–2.03), D-dimer greater than 2 mcg/mL (HR 1.19, CI 1.02–1.39), and troponin greater than 0.03 ng/mL (HR 1.40, CI 1.23–1.62). Decreased risk of in-hospital mortality was associated with female sex (HR 0.84, CI 0.77–0.90), African American race (HR 0.78 CI 0.65–0.95), and hydroxychloroquine use (HR 0.53, CI 0.41–0.67).<ref name=MountSanai/>}} The confidence interval for HCQ in this study ranged from .41 to .67, which indicates that HCQ was associated with decreased risk of in-hospital mortality.  If we use the .53 hazard ratio, people given HCQ were 47% less likely to die.  As shown by the supplementary figure four of the study, the survival of patients was significantly higher for patients taking HCQ. (P < 0.001).  That is a highly significant finding and a strong indicator that HCQ works.  There were no statistical results presented for HCQ among the ambulatory patients, though it was reported that the majority of them did not receive the drug.<ref name=MountSanai/>

{{quotebox|A total of 858 of 6493 (13.2%) patients in our total cohort died: 52/2785 (1.9%) ambulatory patients and 806/3708 (21.7%) hospitalized patients. Cox proportional hazard regression modeling showed an increased risk of in-hospital mortality associated with age older than 50 years (hazard ratio [HR] 2.34, CI 1.47–3.71), systolic blood pressure less than 90 mmHg (HR 1.38, CI 1.06–1.80), a respiratory rate greater than 24 per min (HR 1.43, CI 1.13–1.83), peripheral oxygen saturation less than 92% (HR 2.12, CI 1.56–2.88), estimated glomerular filtration rate less than 60 mL/min/1.73m2 (HR 1.80, CI 1.60–2.02), IL-6 greater than 100 pg/mL (HR 1.50, CI 1.12–2.03), D-dimer greater than 2 mcg/mL (HR 1.19, CI 1.02–1.39), and troponin greater than 0.03 ng/mL (HR 1.40, CI 1.23–1.62). Decreased risk of in-hospital mortality was associated with female sex (HR 0.84, CI 0.77–0.90), African American race (HR 0.78 CI 0.65–0.95), and hydroxychloroquine use (HR 0.53, CI 0.41–0.67).<ref name=MountSanai/>}} The confidence interval for HCQ in this study ranged from .41 to .67, which indicates that HCQ was associated with decreased risk of in-hospital mortality.  If we use the .53 hazard ratio, people given HCQ were 47% less likely to die.  As shown by the supplementary figure four of the study, the survival of patients was significantly higher for patients taking HCQ. (P < 0.001).  That is a highly significant finding and a strong indicator that HCQ works.  There were no statistical results presented for HCQ among the ambulatory patients, though it was reported that the majority of them did not receive the drug.<ref name=MountSanai/>

==Scientific studies that say hydroxychloroquine has a statistically significant net benefit==   

==Scientific studies that say hydroxychloroquine has a statistically significant net benefit==