504592
research-article2014
PENXXX10.1177/0148607113504592Journal of Parenteral and Enteral NutritionKoretz
Journal Club
JPEN Journal Club 3. When and Why to Randomize Ronald L. Koretz, MD1
Journal of Parenteral and Enteral Nutrition Volume 38 Number 3 March 2014 400–401 © 2014 American Society for Parenteral and Enteral Nutrition DOI: 10.1177/0148607113504592 jpen.sagepub.com hosted at online.sagepub.com
Keywords outcomes research/quality; nutrition support practice; research and diseases
When and Why to Randomize The question being addressed in the article being considered for discussion in this installment of JPEN Journal Club1 is straightforward. The investigators were interested in what happens to patients with such severe gastrointestinal (GI) involvement with systemic sclerosis that they are placed on home parenteral nutrition (PN). In other words, this was a natural history (prognosis) study. Such studies do not require control groups. Rather, individuals with a defined medical condition are identified and followed or, in the case of retrospective investigations, their medical records are assessed after the fact. Ideally, in natural history studies, all eligible patients are identified, and none are subsequently lost to follow-up. The advantage to doing such studies prospectively is that the investigator can select the particular outcomes of interest and make sure that that information is collected. Obviously, in retrospective studies, one is limited to the information that somebody else already collected. In this report, 12 patients were retrospectively identified from the medical records of 1 healthcare delivery system (University Health Network in Toronto, Ontario, Canada). The patients were followed for variable amounts of time; in fact, there was a remarkable disparity between the median followup (which is the follow-up that 50% of the group achieved) and the mean follow-up (which is the total number of months of follow-up divided by 12, the number of subjects)—namely, 16 months and almost 41 months. We are told, late in the article, that 1 patient, who happened to have the mildest GI disease, was followed for 270 months, so the average follow-up for the remaining 11 was 20 months, making the mean and median much closer together. Hence, going in we know that there is 1 outlier. We are told that there was an improvement in nutrition markers; this was largely manifested by around a 10%–15% increase in the body mass index at 6 months. While the serum albumin also increased comparably, Jawa et al1 did concede in the Discussion section that this was not a good nutrition marker.2 Functional status did not change. However, this was assessed by the Karnofsky score, a scale that has been validated for patients with cancer but not for those with scleroderma, so its meaning is not clear. The true functional status is even less clear because these subjective data were collected in
an unblinded fashion. Five of the 12 patients, or 42%, died within 2 years, although none of the deaths were attributed to the PN. (Of course, the certainty of this conclusion is unclear, since the investigators did not have all of the details.) In summary, while the PN may have prevented these patients from starving to death, the long-term prognosis of the disease appears to be poor. It should be appreciated that these data cannot necessarily be extrapolated to the entire home PN population. None of these patients had anatomical short bowel; the pathophysiologic problem was deranged motility (with potential bacterial overgrowth in the intestinal lumen), but the absorptive mucosa was largely still present. Although precise information was lacking, it is likely that many, perhaps all, of these patients were consuming nutrients enterally. In fact, we do not even know what percentage of the daily nutrient intake was provided intravenously. A subtle reading point can be taken away from Table 2. Those data show the average values of nutrition, interventional, and laboratory parameters at 0, 6, and 12 months. To be noted, however, is that there were 2 fewer subjects at the 12-month mark. Since data were missing for 2 patients, we cannot make any sense out of the 1-year data, since the mean values at 12 months are only derived from 10 of the 12 individuals. (For example, if the 2 dropouts had the most adverse values, even if the numbers remained the same in the other 10, the average would appear to be better.) What might have provided more insight would have been to present the data only for the 10 patients who were followed for the entire time. It was for this reason that I previously presented the body mass
From Olive View–UCLA Medical Center, David Geffen–UCLA School of Medicine, Sylmar and Los Angeles, California. Financial disclosure: None declared. Received for publication June 13, 2013; accepted for publication August 20, 2013. Corresponding Author: Ronald L. Koretz, MD, Emeritus Professor of Clinical Medicine, Olive View–UCLA Medical Center, David Geffen–UCLA School of Medicine, Sylmar and Los Angeles, CA. Email: [email protected]
Downloaded from pen.sagepub.com at Australian National University on March 13, 2015
Koretz
401
index data only at 6 months, because that was the only followup point for all 12 patients. Through a PubMed search, I found 3 other published similar natural history studies in patients with systemic sclerosis believed to require PN infusions,3-5 including 2 cited in the Jawa et al1 study.3,4 All of the studies suggested that about half of the patients died within the first year or two from the underlying disease process despite improved body weight. It seems fair to believe that this study did describe the true natural history of patients with systemic sclerosis with such severe GI involvement that long-term PN is required. What if a different question had been posed by the investigators—namely, what was the efficacy of PN in this patient population? In general, efficacy trials, unlike natural history ones, do require control groups. However, on a few occasions, this is not the case. The classic example of such a case is the use of parachutes when jumping out of an airplane at 20,000 feet.6 Would you want to be enrolled in a placebo-controlled trial assessing the utility of parachutes? The time when controls are not needed is when the outcome is absolutely predictable in any given patient; if an intervention changes that outcome in a beneficial way, efficacy is established. Clinical examples of such conditions include cardiopulmonary resuscitation for cardiac arrest, hemodialysis for end-stage renal failure, and PN for patients with such severe long-term bowel failure that nutrition status cannot be maintained (our example). For most clinical conditions, the outcome is not so absolutely predictable in any given patient; while there are many diseases in which an adverse outcome is “likely” to occur, controls are still required. Even the parachute paradigm has limitations; for example, what if the individual is jumping from an airplane parked on the ground and the total distance of the fall is only 10 feet? One might legitimately wonder if the seconds used to try to deploy a parachute (assuming that it could even deploy in time) would be better employed in assuming a more defensive posture before hitting the ground. In such a scenario, a controlled trial is necessary. Once it has been determined that a control group is necessary, one must then consider the issue of randomization. The ideal control group differs from the treated one in only 1
variable—namely, the receipt or nonreceipt of the intervention. Any manner of defining controls other than randomization must change 2 variables. The first is, obviously, the receipt or nonreceipt of the intervention. The second is the reason (or reasons) why the intervention was or was not provided. Whatever the reason or reasons are (eg, historical controls from a time when the intervention was not available, comorbidities that reduce the likelihood that the intervention would be given, different healthcare sites), once there are 2 different variables, one cannot attribute any difference in outcome only to one of them. A nonrandomized control group can demonstrate an association and thereby create a hypothesis. However, it cannot prove that hypothesis. For the next installment, please read the following article: Silander E, Nyman J, Bove M, Johansson L, Larsson S, Hammerlid E. Impact of prophylactic percutaneous endoscopic gastrostomy on malnutrition and quality of life in patients with head and neck cancer: a randomized study. Head Neck. 2012;34:1-9. This is a randomized trial assessing the utility of prophylactic percutaneous endoscopic gastrostomies before beginning treatment of patients with head and neck malignancies. Join the conversation about this and other JPEN Journal Club articles in ASPENConnect (www.nutritioncare. org/connect).
References 1. Jawa H, Fernandes G, Saqui O, Allard JP. Home parenteral nutrition in patients with systemic sclerosis: a retrospective review of 12 cases. J Rheumatol. 2012;39:1004-1007. 2. Baron M, Hudson M, Steele R; Canadian Scleroderma Research Group. Is serum albumin a marker of malnutrition in chronic disease? The scleroderma paradigm. J Am Coll Nutr. 2010;29:144-151. 3. Ng SC, Clements PJ, Berquist WE, Furst DE, Paulus HE. Home central venous hyperalimentation in fifteen patients with severe scleroderma bowel disease. Arth Rheum. 1989;32:212-216. 4. Grabowski G, Grant JP. Nutrition support in patients with systemic scleroderma. JPEN J Parenter Enteral Nutr. 1989;13:147-151. 5. Brown M, Teubner A, Shaffer J, Herrick AL. Home parenteral nutrition— an effective and safe long-term therapy for systemic sclerosis-related intestinal failure. Rheumatology. 2008;47:176-179. 6. Smith GCS, Pell JP. Parachute use to prevent death and major trauma related to gravitational challenge: systematic review of randomized controlled trials. BMJ. 2003;327:1459-1461.
Downloaded from pen.sagepub.com at Australian National University on March 13, 2015
research-article2014
PENXXX10.1177/0148607113504592Journal of Parenteral and Enteral NutritionKoretz
Journal Club
JPEN Journal Club 3. When and Why to Randomize Ronald L. Koretz, MD1
Journal of Parenteral and Enteral Nutrition Volume 38 Number 3 March 2014 400–401 © 2014 American Society for Parenteral and Enteral Nutrition DOI: 10.1177/0148607113504592 jpen.sagepub.com hosted at online.sagepub.com
Keywords outcomes research/quality; nutrition support practice; research and diseases
When and Why to Randomize The question being addressed in the article being considered for discussion in this installment of JPEN Journal Club1 is straightforward. The investigators were interested in what happens to patients with such severe gastrointestinal (GI) involvement with systemic sclerosis that they are placed on home parenteral nutrition (PN). In other words, this was a natural history (prognosis) study. Such studies do not require control groups. Rather, individuals with a defined medical condition are identified and followed or, in the case of retrospective investigations, their medical records are assessed after the fact. Ideally, in natural history studies, all eligible patients are identified, and none are subsequently lost to follow-up. The advantage to doing such studies prospectively is that the investigator can select the particular outcomes of interest and make sure that that information is collected. Obviously, in retrospective studies, one is limited to the information that somebody else already collected. In this report, 12 patients were retrospectively identified from the medical records of 1 healthcare delivery system (University Health Network in Toronto, Ontario, Canada). The patients were followed for variable amounts of time; in fact, there was a remarkable disparity between the median followup (which is the follow-up that 50% of the group achieved) and the mean follow-up (which is the total number of months of follow-up divided by 12, the number of subjects)—namely, 16 months and almost 41 months. We are told, late in the article, that 1 patient, who happened to have the mildest GI disease, was followed for 270 months, so the average follow-up for the remaining 11 was 20 months, making the mean and median much closer together. Hence, going in we know that there is 1 outlier. We are told that there was an improvement in nutrition markers; this was largely manifested by around a 10%–15% increase in the body mass index at 6 months. While the serum albumin also increased comparably, Jawa et al1 did concede in the Discussion section that this was not a good nutrition marker.2 Functional status did not change. However, this was assessed by the Karnofsky score, a scale that has been validated for patients with cancer but not for those with scleroderma, so its meaning is not clear. The true functional status is even less clear because these subjective data were collected in
an unblinded fashion. Five of the 12 patients, or 42%, died within 2 years, although none of the deaths were attributed to the PN. (Of course, the certainty of this conclusion is unclear, since the investigators did not have all of the details.) In summary, while the PN may have prevented these patients from starving to death, the long-term prognosis of the disease appears to be poor. It should be appreciated that these data cannot necessarily be extrapolated to the entire home PN population. None of these patients had anatomical short bowel; the pathophysiologic problem was deranged motility (with potential bacterial overgrowth in the intestinal lumen), but the absorptive mucosa was largely still present. Although precise information was lacking, it is likely that many, perhaps all, of these patients were consuming nutrients enterally. In fact, we do not even know what percentage of the daily nutrient intake was provided intravenously. A subtle reading point can be taken away from Table 2. Those data show the average values of nutrition, interventional, and laboratory parameters at 0, 6, and 12 months. To be noted, however, is that there were 2 fewer subjects at the 12-month mark. Since data were missing for 2 patients, we cannot make any sense out of the 1-year data, since the mean values at 12 months are only derived from 10 of the 12 individuals. (For example, if the 2 dropouts had the most adverse values, even if the numbers remained the same in the other 10, the average would appear to be better.) What might have provided more insight would have been to present the data only for the 10 patients who were followed for the entire time. It was for this reason that I previously presented the body mass
From Olive View–UCLA Medical Center, David Geffen–UCLA School of Medicine, Sylmar and Los Angeles, California. Financial disclosure: None declared. Received for publication June 13, 2013; accepted for publication August 20, 2013. Corresponding Author: Ronald L. Koretz, MD, Emeritus Professor of Clinical Medicine, Olive View–UCLA Medical Center, David Geffen–UCLA School of Medicine, Sylmar and Los Angeles, CA. Email: [email protected]
Downloaded from pen.sagepub.com at Australian National University on March 13, 2015
Koretz
401
index data only at 6 months, because that was the only followup point for all 12 patients. Through a PubMed search, I found 3 other published similar natural history studies in patients with systemic sclerosis believed to require PN infusions,3-5 including 2 cited in the Jawa et al1 study.3,4 All of the studies suggested that about half of the patients died within the first year or two from the underlying disease process despite improved body weight. It seems fair to believe that this study did describe the true natural history of patients with systemic sclerosis with such severe GI involvement that long-term PN is required. What if a different question had been posed by the investigators—namely, what was the efficacy of PN in this patient population? In general, efficacy trials, unlike natural history ones, do require control groups. However, on a few occasions, this is not the case. The classic example of such a case is the use of parachutes when jumping out of an airplane at 20,000 feet.6 Would you want to be enrolled in a placebo-controlled trial assessing the utility of parachutes? The time when controls are not needed is when the outcome is absolutely predictable in any given patient; if an intervention changes that outcome in a beneficial way, efficacy is established. Clinical examples of such conditions include cardiopulmonary resuscitation for cardiac arrest, hemodialysis for end-stage renal failure, and PN for patients with such severe long-term bowel failure that nutrition status cannot be maintained (our example). For most clinical conditions, the outcome is not so absolutely predictable in any given patient; while there are many diseases in which an adverse outcome is “likely” to occur, controls are still required. Even the parachute paradigm has limitations; for example, what if the individual is jumping from an airplane parked on the ground and the total distance of the fall is only 10 feet? One might legitimately wonder if the seconds used to try to deploy a parachute (assuming that it could even deploy in time) would be better employed in assuming a more defensive posture before hitting the ground. In such a scenario, a controlled trial is necessary. Once it has been determined that a control group is necessary, one must then consider the issue of randomization. The ideal control group differs from the treated one in only 1
variable—namely, the receipt or nonreceipt of the intervention. Any manner of defining controls other than randomization must change 2 variables. The first is, obviously, the receipt or nonreceipt of the intervention. The second is the reason (or reasons) why the intervention was or was not provided. Whatever the reason or reasons are (eg, historical controls from a time when the intervention was not available, comorbidities that reduce the likelihood that the intervention would be given, different healthcare sites), once there are 2 different variables, one cannot attribute any difference in outcome only to one of them. A nonrandomized control group can demonstrate an association and thereby create a hypothesis. However, it cannot prove that hypothesis. For the next installment, please read the following article: Silander E, Nyman J, Bove M, Johansson L, Larsson S, Hammerlid E. Impact of prophylactic percutaneous endoscopic gastrostomy on malnutrition and quality of life in patients with head and neck cancer: a randomized study. Head Neck. 2012;34:1-9. This is a randomized trial assessing the utility of prophylactic percutaneous endoscopic gastrostomies before beginning treatment of patients with head and neck malignancies. Join the conversation about this and other JPEN Journal Club articles in ASPENConnect (www.nutritioncare. org/connect).
References 1. Jawa H, Fernandes G, Saqui O, Allard JP. Home parenteral nutrition in patients with systemic sclerosis: a retrospective review of 12 cases. J Rheumatol. 2012;39:1004-1007. 2. Baron M, Hudson M, Steele R; Canadian Scleroderma Research Group. Is serum albumin a marker of malnutrition in chronic disease? The scleroderma paradigm. J Am Coll Nutr. 2010;29:144-151. 3. Ng SC, Clements PJ, Berquist WE, Furst DE, Paulus HE. Home central venous hyperalimentation in fifteen patients with severe scleroderma bowel disease. Arth Rheum. 1989;32:212-216. 4. Grabowski G, Grant JP. Nutrition support in patients with systemic scleroderma. JPEN J Parenter Enteral Nutr. 1989;13:147-151. 5. Brown M, Teubner A, Shaffer J, Herrick AL. Home parenteral nutrition— an effective and safe long-term therapy for systemic sclerosis-related intestinal failure. Rheumatology. 2008;47:176-179. 6. Smith GCS, Pell JP. Parachute use to prevent death and major trauma related to gravitational challenge: systematic review of randomized controlled trials. BMJ. 2003;327:1459-1461.
Downloaded from pen.sagepub.com at Australian National University on March 13, 2015
